wassname/Open-Assistant
1
0
Fork 0
mirror of https://github.com/wassname/Open-Assistant.git synced 2026-07-09 00:20:03 +08:00
Code Issues Packages Projects Releases Wiki Activity

[fix] Merge main branch update

Browse Source
This commit is contained in:
theblackcat102
2023-01-20 06:18:43 +00:00
parent 6cd62e3d48 5a65dd9440
commit c255148dc6
124 changed files with 4101 additions and 1134 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/workflows/build-postgres.yaml
+17
View File
@@ -0,0 +1,17 @@
name: Build OASST Postgres image
on:
push:
branches:
- main
paths:
- docker/oasst-postgres/**
jobs:
build-postgres:
uses: ./.github/workflows/docker-build.yaml
with:
image-name: oasst-postgres
context: ./docker/oasst-postgres
dockerfile: docker/oasst-postgres/Dockerfile
build-args: ""
.github/workflows/deploy-to-node.yaml
+9 -1
View File
@@ -32,7 +32,15 @@ jobs:
WEB_EMAIL_SERVER_PASSWORD: ${{ secrets.DEV_WEB_EMAIL_SERVER_PASSWORD }}
WEB_EMAIL_SERVER_PORT: ${{ secrets.DEV_WEB_EMAIL_SERVER_PORT }}
WEB_EMAIL_SERVER_USER: ${{ secrets.DEV_WEB_EMAIL_SERVER_USER }}
WEB_NEXTAUTH_SECRET: ${{ secrets.DEV_WEB_NEXTAUTH_SECRET }}
WEB_NEXTAUTH_SECRET: ${{ secrets.NEXTAUTH_SECRET }}
S3_BUCKET_NAME: ${{ secrets.S3_BUCKET_NAME }}
S3_REGION: ${{ secrets.S3_REGION }}
AWS_ACCESS_KEY: ${{ secrets.AWS_ACCESS_KEY }}
AWS_SECRET_KEY: ${{ secrets.AWS_SECRET_KEY }}
MAX_ACTIVE_TREES: ${{ vars.MAX_ACTIVE_TREES }}
MAX_TREE_DEPTH: ${{ vars.MAX_TREE_DEPTH }}
GOAL_TREE_SIZE: ${{ vars.GOAL_TREE_SIZE }}
SKIP_TOXICITY_CALCULATION: ${{ vars.SKIP_TOXICITY_CALCULATION }}
steps:
- name: Checkout
uses: actions/checkout@v2
ansible/deploy-to-node.yaml
+23 -3
View File
@@ -57,8 +57,9 @@
- name: Create postgres containers
community.docker.docker_container:
name: "oasst-{{ stack_name }}-postgres-{{ item.name }}"
image: postgres:15
image: ghcr.io/laion-ai/open-assistant/oasst-postgres
state: started
pull: true
recreate: "{{ (stack_name == 'dev') | bool }}"
restart_policy: always
network_mode: "oasst-{{ stack_name }}"
@@ -66,6 +67,13 @@
POSTGRES_USER: postgres
POSTGRES_PASSWORD: "{{ postgres_password }}"
POSTGRES_DB: postgres
S3_BUCKET_NAME:
"{{ lookup('ansible.builtin.env', 'S3_BUCKET_NAME') }}"
AWS_ACCESS_KEY_ID:
"{{ lookup('ansible.builtin.env', 'AWS_ACCESS_KEY') }}"
AWS_SECRET_ACCESS_KEY:
"{{ lookup('ansible.builtin.env', 'AWS_SECRET_KEY') }}"
AWS_DEFAULT_REGION: "{{ lookup('ansible.builtin.env', 'S3_REGION') }}"
volumes:
- "oasst-{{ stack_name }}-postgres-{{ item.name
}}:/var/lib/postgresql/data"
@@ -99,8 +107,18 @@
RATE_LIMIT: "{{ 'false' if stack_name == 'dev' else 'true' }}"
DEBUG_SKIP_EMBEDDING_COMPUTATION: "true"
DEBUG_SKIP_TOXICITY_CALCULATION:
"{{ 'true' if stack_name == 'dev' else 'false' }}"
"{{ lookup('ansible.builtin.env', 'SKIP_TOXICITY_CALCULATION') |
default('true', true) }}"
OFFICIAL_WEB_API_KEY: "{{ web_api_key }}"
TREE_MANAGER__MAX_ACTIVE_TREES:
"{{ lookup('ansible.builtin.env', 'MAX_ACTIVE_TREES') |
default('10', true) }}"
TREE_MANAGER__MAX_TREE_DEPTH:
"{{ lookup('ansible.builtin.env', 'MAX_TREE_DEPTH') | default('5',
true) }}"
TREE_MANAGER__GOAL_TREE_SIZE:
"{{ lookup('ansible.builtin.env', 'GOAL_TREE_SIZE') | default('15',
true) }}"
ports:
- "{{ backend_port }}:8080"
@@ -136,7 +154,9 @@
FASTAPI_KEY: "{{ web_api_key }}"
NEXTAUTH_SECRET:
"{{ lookup('ansible.builtin.env', 'WEB_NEXTAUTH_SECRET') }}"
NEXTAUTH_URL: http://web.{{ stack_name }}.open-assistant.io/
NEXTAUTH_URL:
"{{ 'https://open-assistant.io/' if stack_name == 'production' else
('https://web.' + stack_name + '.open-assistant.io/') }}"
ports:
- "{{ website_port }}:3000"
command: bash wait-for-postgres.sh node server.js
ansible/pgbackrest.conf
+24
View File
@@ -0,0 +1,24 @@
[oasst]
pg1-path=/var/lib/postgresql/data
[global]
repo1-retention-full=3
repo1-type=s3
repo1-path=/oasst-prod
repo1-s3-region=us-east-1
repo1-s3-endpoint=s3.amazonaws.com
# repo1-s3-bucket=$S3_BUCKET_NAME
# repo1-s3-key=$AWS_ACCESS_KEY
# repo1-s3-key-secret=$AWS_SECRET_KEY
# Force a checkpoint to start backup immediately.
start-fast=y
# Use delta restore.
delta=y
# Enable ZSTD compression.
compress-type=zst
compress-level=6
log-level-console=info
log-level-file=debug
backend/alembic/versions/2023_01_19_2153-7f0a28a156f4_switch_to_timestamp_with_tz.py
+47
View File
@@ -0,0 +1,47 @@
"""switch to timestamp with tz
Revision ID: 7f0a28a156f4
Revises: 0964ac95170d
Create Date: 2023-01-19 21:53:01.107137
"""
import sqlalchemy as sa
from alembic import op
# revision identifiers, used by Alembic.
revision = "7f0a28a156f4"
down_revision = "0964ac95170d"
branch_labels = None
depends_on = None
def upgrade() -> None:
# ### commands auto generated by Alembic - please adjust! ###
op.alter_column(table_name="user_stats", column_name="modified_date", type_=sa.DateTime(timezone=True))
op.alter_column(table_name="user_stats", column_name="base_date", type_=sa.DateTime(timezone=True))
op.alter_column(table_name="journal_integration", column_name="last_run", type_=sa.DateTime(timezone=True))
op.alter_column(table_name="message_embedding", column_name="created_date", type_=sa.DateTime(timezone=True))
op.alter_column(table_name="message_reaction", column_name="created_date", type_=sa.DateTime(timezone=True))
op.alter_column(table_name="message_toxicity", column_name="created_date", type_=sa.DateTime(timezone=True))
op.alter_column(table_name="message", column_name="created_date", type_=sa.DateTime(timezone=True))
op.alter_column(table_name="task", column_name="created_date", type_=sa.DateTime(timezone=True))
op.alter_column(table_name="task", column_name="expiry_date", type_=sa.DateTime(timezone=True))
op.alter_column(table_name="text_labels", column_name="created_date", type_=sa.DateTime(timezone=True))
op.alter_column(table_name="user", column_name="created_date", type_=sa.DateTime(timezone=True))
# ### end Alembic commands ###
def downgrade() -> None:
# ### commands auto generated by Alembic - please adjust! ###
op.alter_column(table_name="user_stats", column_name="modified_date", type_=sa.DateTime(timezone=False))
op.alter_column(table_name="user_stats", column_name="base_date", type_=sa.DateTime(timezone=False))
op.alter_column(table_name="journal_integration", column_name="last_run", type_=sa.DateTime(timezone=False))
op.alter_column(table_name="message_embedding", column_name="created_date", type_=sa.DateTime(timezone=False))
op.alter_column(table_name="message_reaction", column_name="created_date", type_=sa.DateTime(timezone=False))
op.alter_column(table_name="message_toxicity", column_name="created_date", type_=sa.DateTime(timezone=False))
op.alter_column(table_name="message", column_name="created_date", type_=sa.DateTime(timezone=False))
op.alter_column(table_name="task", column_name="created_date", type_=sa.DateTime(timezone=False))
op.alter_column(table_name="task", column_name="expiry_date", type_=sa.DateTime(timezone=False))
op.alter_column(table_name="text_labels", column_name="created_date", type_=sa.DateTime(timezone=False))
op.alter_column(table_name="user", column_name="created_date", type_=sa.DateTime(timezone=False))
# ### end Alembic commands ###
backend/main.py
+78 -80
View File
@@ -20,6 +20,7 @@ from oasst_backend.models import message_tree_state
from oasst_backend.prompt_repository import PromptRepository, TaskRepository, UserRepository
from oasst_backend.tree_manager import TreeManager
from oasst_backend.user_stats_repository import UserStatsRepository, UserStatsTimeFrame
from oasst_backend.utils.database_utils import CommitMode, managed_tx_function
from oasst_shared.exceptions import OasstError, OasstErrorCode
from oasst_shared.schemas import protocol as protocol_schema
from pydantic import BaseModel
@@ -120,7 +121,8 @@ if settings.RATE_LIMIT:
if settings.DEBUG_USE_SEED_DATA:
@app.on_event("startup")
def seed_data():
@managed_tx_function(auto_commit=CommitMode.COMMIT)
def create_seed_data(session: Session):
class DummyMessage(BaseModel):
task_message_id: str
user_message_id: str
@@ -134,73 +136,73 @@ if settings.DEBUG_USE_SEED_DATA:
try:
logger.info("Seed data check began")
with Session(engine) as db:
api_client = api_auth(settings.OFFICIAL_WEB_API_KEY, db=db)
dummy_user = protocol_schema.User(id="__dummy_user__", display_name="Dummy User", auth_method="local")
ur = UserRepository(db=db, api_client=api_client)
tr = TaskRepository(db=db, api_client=api_client, client_user=dummy_user, user_repository=ur)
pr = PromptRepository(
db=db, api_client=api_client, client_user=dummy_user, user_repository=ur, task_repository=tr
)
tm = TreeManager(db, pr)
api_client = api_auth(settings.OFFICIAL_WEB_API_KEY, db=session)
dummy_user = protocol_schema.User(id="__dummy_user__", display_name="Dummy User", auth_method="local")
with open(settings.DEBUG_USE_SEED_DATA_PATH) as f:
dummy_messages_raw = json.load(f)
ur = UserRepository(db=session, api_client=api_client)
tr = TaskRepository(db=session, api_client=api_client, client_user=dummy_user, user_repository=ur)
pr = PromptRepository(
db=session, api_client=api_client, client_user=dummy_user, user_repository=ur, task_repository=tr
)
tm = TreeManager(session, pr)
dummy_messages = [DummyMessage(**dm) for dm in dummy_messages_raw]
with open(settings.DEBUG_USE_SEED_DATA_PATH) as f:
dummy_messages_raw = json.load(f)
for msg in dummy_messages:
task = tr.fetch_task_by_frontend_message_id(msg.task_message_id)
if task and not task.ack:
logger.warning("Deleting unacknowledged seed data task")
db.delete(task)
task = None
if not task:
if msg.parent_message_id is None:
task = tr.store_task(
protocol_schema.InitialPromptTask(hint=""), message_tree_id=None, parent_message_id=None
)
else:
parent_message = pr.fetch_message_by_frontend_message_id(
msg.parent_message_id, fail_if_missing=True
)
conversation_messages = pr.fetch_message_conversation(parent_message)
conversation = prepare_conversation(conversation_messages)
if msg.role == "assistant":
task = tr.store_task(
protocol_schema.AssistantReplyTask(conversation=conversation),
message_tree_id=parent_message.message_tree_id,
parent_message_id=parent_message.id,
)
else:
task = tr.store_task(
protocol_schema.PrompterReplyTask(conversation=conversation),
message_tree_id=parent_message.message_tree_id,
parent_message_id=parent_message.id,
)
tr.bind_frontend_message_id(task.id, msg.task_message_id)
message = pr.store_text_reply(
msg.text,
msg.task_message_id,
msg.user_message_id,
review_count=5,
review_result=True,
check_tree_state=False,
)
if message.parent_id is None:
tm._insert_default_state(
root_message_id=message.id, state=msg.tree_state or message_tree_state.State.GROWING
)
db.commit()
dummy_messages = [DummyMessage(**dm) for dm in dummy_messages_raw]
logger.info(
f"Inserted: message_id: {message.id}, payload: {message.payload.payload}, parent_message_id: {message.parent_id}"
for msg in dummy_messages:
task = tr.fetch_task_by_frontend_message_id(msg.task_message_id)
if task and not task.ack:
logger.warning("Deleting unacknowledged seed data task")
session.delete(task)
task = None
if not task:
if msg.parent_message_id is None:
task = tr.store_task(
protocol_schema.InitialPromptTask(hint=""), message_tree_id=None, parent_message_id=None
)
else:
logger.debug(f"seed data task found: {task.id}")
parent_message = pr.fetch_message_by_frontend_message_id(
msg.parent_message_id, fail_if_missing=True
)
conversation_messages = pr.fetch_message_conversation(parent_message)
conversation = prepare_conversation(conversation_messages)
if msg.role == "assistant":
task = tr.store_task(
protocol_schema.AssistantReplyTask(conversation=conversation),
message_tree_id=parent_message.message_tree_id,
parent_message_id=parent_message.id,
)
else:
task = tr.store_task(
protocol_schema.PrompterReplyTask(conversation=conversation),
message_tree_id=parent_message.message_tree_id,
parent_message_id=parent_message.id,
)
tr.bind_frontend_message_id(task.id, msg.task_message_id)
message = pr.store_text_reply(
msg.text,
msg.task_message_id,
msg.user_message_id,
review_count=5,
review_result=True,
check_tree_state=False,
)
if message.parent_id is None:
tm._insert_default_state(
root_message_id=message.id, state=msg.tree_state or message_tree_state.State.GROWING
)
session.flush()
logger.info("Seed data check completed")
logger.info(
f"Inserted: message_id: {message.id}, payload: {message.payload.payload}, parent_message_id: {message.parent_id}"
)
else:
logger.debug(f"seed data task found: {task.id}")
logger.info("Seed data check completed")
except Exception:
logger.exception("Seed data insertion failed")
@@ -220,48 +222,44 @@ def ensure_tree_states():
@app.on_event("startup")
@repeat_every(seconds=60 * settings.USER_STATS_INTERVAL_DAY, wait_first=False)
def update_leader_board_day() -> None:
@managed_tx_function(auto_commit=CommitMode.COMMIT)
def update_leader_board_day(session: Session) -> None:
try:
with Session(engine) as session:
usr = UserStatsRepository(session)
usr.update_stats(time_frame=UserStatsTimeFrame.day)
session.commit()
usr = UserStatsRepository(session)
usr.update_stats(time_frame=UserStatsTimeFrame.day)
except Exception:
logger.exception("Error during leaderboard update (daily)")
@app.on_event("startup")
@repeat_every(seconds=60 * settings.USER_STATS_INTERVAL_WEEK, wait_first=False)
def update_leader_board_week() -> None:
@managed_tx_function(auto_commit=CommitMode.COMMIT)
def update_leader_board_week(session: Session) -> None:
try:
with Session(engine) as session:
usr = UserStatsRepository(session)
usr.update_stats(time_frame=UserStatsTimeFrame.week)
session.commit()
usr = UserStatsRepository(session)
usr.update_stats(time_frame=UserStatsTimeFrame.week)
except Exception:
logger.exception("Error during user states update (weekly)")
@app.on_event("startup")
@repeat_every(seconds=60 * settings.USER_STATS_INTERVAL_MONTH, wait_first=False)
def update_leader_board_month() -> None:
@managed_tx_function(auto_commit=CommitMode.COMMIT)
def update_leader_board_month(session: Session) -> None:
try:
with Session(engine) as session:
usr = UserStatsRepository(session)
usr.update_stats(time_frame=UserStatsTimeFrame.month)
session.commit()
usr = UserStatsRepository(session)
usr.update_stats(time_frame=UserStatsTimeFrame.month)
except Exception:
logger.exception("Error during user states update (monthly)")
@app.on_event("startup")
@repeat_every(seconds=60 * settings.USER_STATS_INTERVAL_TOTAL, wait_first=False)
def update_leader_board_total() -> None:
@managed_tx_function(auto_commit=CommitMode.COMMIT)
def update_leader_board_total(session: Session) -> None:
try:
with Session(engine) as session:
usr = UserStatsRepository(session)
usr.update_stats(time_frame=UserStatsTimeFrame.total)
session.commit()
usr = UserStatsRepository(session)
usr.update_stats(time_frame=UserStatsTimeFrame.total)
except Exception:
logger.exception("Error during user states update (total)")
backend/oasst_backend/api/v1/admin.py
+133 -1
View File
@@ -1,7 +1,17 @@
from datetime import datetime
from uuid import UUID
import pydantic
from fastapi import APIRouter, Depends
from loguru import logger
from oasst_backend.api import deps
from oasst_backend.config import Settings, settings
from oasst_backend.models import ApiClient, User
from oasst_backend.prompt_repository import PromptRepository
from oasst_backend.tree_manager import TreeManager
from oasst_backend.utils.database_utils import CommitMode, managed_tx_function
from oasst_shared.schemas.protocol import SystemStats
from oasst_shared.utils import ScopeTimer, unaware_to_utc
router = APIRouter()
@@ -13,7 +23,7 @@ class CreateApiClientRequest(pydantic.BaseModel):
admin_email: str | None = None
@router.post("/api_client")
@router.post("/api_client", response_model=str)
async def create_api_client(
request: CreateApiClientRequest,
root_token: str = Depends(deps.get_root_token),
@@ -29,3 +39,125 @@ async def create_api_client(
)
logger.info(f"Created api_client with key {api_client.api_key}")
return api_client.api_key
@router.get("/backend_settings/full", response_model=Settings)
async def get_backend_settings_full(api_client: ApiClient = Depends(deps.get_trusted_api_client)) -> Settings:
logger.info(
f"Backend settings requested by trusted api_client {api_client.id} (admin_email: {api_client.admin_email}, frontend_type: {api_client.frontend_type})"
)
return settings
class PublicSettings(pydantic.BaseModel):
"""Subset of backend settings which can be retrieved by untrusted API clients."""
PROJECT_NAME: str
API_V1_STR: str
DEBUG_USE_SEED_DATA: bool
DEBUG_ALLOW_SELF_LABELING: bool
DEBUG_SKIP_EMBEDDING_COMPUTATION: bool
DEBUG_SKIP_TOXICITY_CALCULATION: bool
DEBUG_DATABASE_ECHO: bool
USER_STATS_INTERVAL_DAY: int
USER_STATS_INTERVAL_WEEK: int
USER_STATS_INTERVAL_MONTH: int
USER_STATS_INTERVAL_TOTAL: int
@router.get("/backend_settings/public", response_model=PublicSettings)
async def get_backend_settings_public(api_client: ApiClient = Depends(deps.get_api_client)) -> PublicSettings:
return PublicSettings(**settings.dict())
class PurgeResultModel(pydantic.BaseModel):
before: SystemStats
after: SystemStats
preview: bool
duration: float
@router.post("/purge_user/{user_id}", response_model=PurgeResultModel)
async def purge_user(
user_id: UUID,
preview: bool = True,
ban: bool = True,
api_client: ApiClient = Depends(deps.get_trusted_api_client),
) -> str:
assert api_client.trusted
@managed_tx_function(CommitMode.ROLLBACK if preview else CommitMode.COMMIT)
def purge_tx(session: deps.Session) -> tuple[User, SystemStats, SystemStats]:
pr = PromptRepository(session, api_client)
stats_before = pr.get_stats()
user = pr.user_repository.get_user(user_id)
tm = TreeManager(session, pr)
tm.purge_user(user_id=user_id, ban=ban)
session.expunge(user)
return user, stats_before, pr.get_stats()
timer = ScopeTimer()
user, before, after = purge_tx()
timer.stop()
if preview:
logger.info(
f"PURGE USER PREVIEW: '{user.display_name}' (id: {str(user_id)}; username: '{user.username}'; auth-method: '{user.auth_method}')"
)
else:
logger.warning(
f"PURGE USER: '{user.display_name}' (id: {str(user_id)}; username: '{user.username}'; auth-method: '{user.auth_method}')"
)
logger.info(f"{before=}; {after=}")
return PurgeResultModel(before=before, after=after, preview=preview, duration=timer.elapsed)
@router.post("/purge_user/{user_id}/messages", response_model=PurgeResultModel)
async def purge_user_messages(
user_id: UUID,
purge_initial_prompts: bool = False,
min_date: datetime = None,
max_date: datetime = None,
preview: bool = True,
api_client: ApiClient = Depends(deps.get_trusted_api_client),
) -> str:
assert api_client.trusted
min_date = unaware_to_utc(min_date)
max_date = unaware_to_utc(max_date)
@managed_tx_function(CommitMode.ROLLBACK if preview else CommitMode.COMMIT)
def purge_user_messages_tx(session: deps.Session):
pr = PromptRepository(session, api_client)
stats_before = pr.get_stats()
user = pr.user_repository.get_user(user_id)
tm = TreeManager(session, pr)
tm.purge_user_messages(
user_id, purge_initial_prompts=purge_initial_prompts, min_date=min_date, max_date=max_date
)
session.expunge(user)
return user, stats_before, pr.get_stats()
timer = ScopeTimer()
user, before, after = purge_user_messages_tx()
timer.stop()
if preview:
logger.info(
f"PURGE USER MESSAGES PREVIEW: '{user.display_name}' (id: {str(user_id)}; username: '{user.username}'; auth-method: '{user.auth_method}')"
)
else:
logger.warning(
f"PURGE USER MESSAGES: '{user.display_name}' (id: {str(user_id)}; username: '{user.username}'; auth-method: '{user.auth_method}')"
)
logger.info(f"{before=}; {after=}")
return PurgeResultModel(before=before, after=after, preview=preview, duration=timer.elapsed)
backend/oasst_backend/api/v1/frontend_messages.py
+1 -1
View File
@@ -45,7 +45,7 @@ def get_tree_by_frontend_id(
"""
pr = PromptRepository(db, api_client)
message = pr.fetch_message_by_frontend_message_id(message_id)
tree = pr.fetch_message_tree(message.message_tree_id)
tree = pr.fetch_message_tree(message.message_tree_id, reviewed=False)
return utils.prepare_tree(tree, message.message_tree_id)
backend/oasst_backend/api/v1/leaderboards.py
+20
View File
@@ -6,6 +6,7 @@ from oasst_backend.models import ApiClient
from oasst_backend.user_stats_repository import UserStatsRepository, UserStatsTimeFrame
from oasst_shared.schemas.protocol import LeaderboardStats
from sqlmodel import Session
from starlette.status import HTTP_204_NO_CONTENT
router = APIRouter()
@@ -19,3 +20,22 @@ def get_leaderboard(
) -> LeaderboardStats:
usr = UserStatsRepository(db)
return usr.get_leaderboard(time_frame, limit=max_count)
@router.post("/update/{time_frame}", response_model=None, status_code=HTTP_204_NO_CONTENT)
def update_leaderboard_time_frame(
time_frame: UserStatsTimeFrame,
api_client: ApiClient = Depends(deps.get_trusted_api_client),
db: Session = Depends(deps.get_db),
) -> LeaderboardStats:
usr = UserStatsRepository(db)
return usr.update_stats(time_frame=time_frame)
@router.post("/update", response_model=None, status_code=HTTP_204_NO_CONTENT)
def update_leaderboards_all(
api_client: ApiClient = Depends(deps.get_trusted_api_client),
db: Session = Depends(deps.get_db),
) -> LeaderboardStats:
usr = UserStatsRepository(db)
return usr.update_all_time_frames()
backend/oasst_backend/api/v1/messages.py
+5 -1
View File
@@ -7,6 +7,7 @@ from oasst_backend.api.v1 import utils
from oasst_backend.models import ApiClient
from oasst_backend.prompt_repository import PromptRepository
from oasst_shared.schemas import protocol
from oasst_shared.utils import unaware_to_utc
from sqlmodel import Session
from starlette.status import HTTP_204_NO_CONTENT
@@ -29,6 +30,9 @@ def query_messages(
"""
Query messages.
"""
start_date = unaware_to_utc(start_date)
end_date = unaware_to_utc(end_date)
pr = PromptRepository(db, api_client)
messages = pr.query_messages(
username=username,
@@ -78,7 +82,7 @@ def get_tree(
"""
pr = PromptRepository(db, api_client)
message = pr.fetch_message(message_id)
tree = pr.fetch_message_tree(message.message_tree_id)
tree = pr.fetch_message_tree(message.message_tree_id, reviewed=False)
return utils.prepare_tree(tree, message.message_tree_id)
backend/oasst_backend/api/v1/stats.py
+32
View File
@@ -2,6 +2,7 @@ from fastapi import APIRouter, Depends
from oasst_backend.api import deps
from oasst_backend.models import ApiClient
from oasst_backend.prompt_repository import PromptRepository
from oasst_backend.tree_manager import TreeManager, TreeManagerStats, TreeMessageCountStats
from oasst_shared.schemas import protocol
from sqlmodel import Session
@@ -15,3 +16,34 @@ def get_message_stats(
):
pr = PromptRepository(db, api_client)
return pr.get_stats()
@router.get("/tree_manager/state_counts", response_model=dict[str, int])
def get_tree_manager__state_counts(
db: Session = Depends(deps.get_db),
api_client: ApiClient = Depends(deps.get_trusted_api_client),
):
pr = PromptRepository(db, api_client)
tm = TreeManager(db, pr)
return tm.tree_counts_by_state()
@router.get("/tree_manager/message_counts", response_model=list[TreeMessageCountStats])
def get_tree_manager__message_counts(
only_active: bool = True,
db: Session = Depends(deps.get_db),
api_client: ApiClient = Depends(deps.get_trusted_api_client),
):
pr = PromptRepository(db, api_client)
tm = TreeManager(db, pr)
return tm.tree_message_count_stats(only_active=only_active)
@router.get("/tree_manager", response_model=TreeManagerStats)
def get_tree_manager__stats(
db: Session = Depends(deps.get_db),
api_client: ApiClient = Depends(deps.get_trusted_api_client),
):
pr = PromptRepository(db, api_client)
tm = TreeManager(db, pr)
return tm.stats()
backend/oasst_backend/api/v1/tasks.py
+24 -1
View File
@@ -1,4 +1,4 @@
from typing import Any
from typing import Any, Optional
from uuid import UUID
from fastapi import APIRouter, Depends
@@ -36,6 +36,8 @@ def request_task(
try:
pr = PromptRepository(db, api_client, client_user=request.user)
pr.ensure_user_is_enabled()
tm = TreeManager(db, pr)
task, message_tree_id, parent_message_id = tm.next_task(request.type)
pr.task_repository.store_task(task, message_tree_id, parent_message_id, request.collective)
@@ -48,6 +50,27 @@ def request_task(
return task
@router.post("/availability", response_model=dict[protocol_schema.TaskRequestType, int])
def tasks_availability(
*,
user: Optional[protocol_schema.User] = None,
db: Session = Depends(deps.get_db),
api_key: APIKey = Depends(deps.get_api_key),
):
api_client = deps.api_auth(api_key, db)
try:
pr = PromptRepository(db, api_client, client_user=user)
tm = TreeManager(db, pr)
return tm.determine_task_availability()
except OasstError:
raise
except Exception:
logger.exception("Task availability query failed.")
raise OasstError("Task availability query failed.", OasstErrorCode.TASK_AVAILABILITY_QUERY_FAILED)
@router.post("/{task_id}/ack", response_model=None, status_code=HTTP_204_NO_CONTENT)
def tasks_acknowledge(
*,
backend/oasst_backend/config.py
+4
View File
@@ -55,6 +55,8 @@ class TreeManagerConfiguration(BaseModel):
mandatory_labels_prompter_reply: Optional[list[protocol_schema.TextLabel]] = [protocol_schema.TextLabel.spam]
"""Mandatory labels in text-labeling tasks for prompter replies."""
rank_prompter_replies: bool = False
class Settings(BaseSettings):
PROJECT_NAME: str = "open-assistant backend"
@@ -67,6 +69,7 @@ class Settings(BaseSettings):
POSTGRES_PASSWORD: str = "postgres"
POSTGRES_DB: str = "postgres"
DATABASE_URI: Optional[PostgresDsn] = None
DATABASE_MAX_TX_RETRY_COUNT: int = 3
RATE_LIMIT: bool = True
REDIS_HOST: str = "localhost"
@@ -79,6 +82,7 @@ class Settings(BaseSettings):
DEBUG_ALLOW_SELF_LABELING: bool = False # allow users to label their own messages
DEBUG_SKIP_EMBEDDING_COMPUTATION: bool = False
DEBUG_SKIP_TOXICITY_CALCULATION: bool = False
DEBUG_DATABASE_ECHO: bool = False
HUGGING_FACE_API_KEY: str = ""
backend/oasst_backend/database.py
+1 -1
View File
@@ -5,4 +5,4 @@ from sqlmodel import create_engine
if settings.DATABASE_URI is None:
raise OasstError("DATABASE_URI is not set", error_code=OasstErrorCode.DATABASE_URI_NOT_SET)
engine = create_engine(settings.DATABASE_URI)
engine = create_engine(settings.DATABASE_URI, echo=settings.DEBUG_DATABASE_ECHO, isolation_level="REPEATABLE READ")
backend/oasst_backend/journal_writer.py
+2 -3
View File
@@ -4,6 +4,7 @@ from uuid import UUID
from oasst_backend.models import ApiClient, Journal, Task, User
from oasst_backend.models.payload_column_type import PayloadContainer, payload_type
from oasst_backend.utils.database_utils import CommitMode, managed_tx_method
from oasst_shared.utils import utcnow
from pydantic import BaseModel
from sqlmodel import Session
@@ -80,6 +81,7 @@ class JournalWriter:
message_id=message_id,
)
@managed_tx_method(CommitMode.FLUSH)
def log(
self,
*,
@@ -115,7 +117,4 @@ class JournalWriter:
)
self.db.add(entry)
if commit:
self.db.commit()
return entry
backend/oasst_backend/models/journal.py
+1 -1
View File
@@ -50,6 +50,6 @@ class JournalIntegration(SQLModel, table=True):
)
description: str = Field(max_length=512, primary_key=True)
last_journal_id: Optional[UUID] = Field(foreign_key="journal.id", nullable=True)
last_run: Optional[datetime] = Field(sa_column=sa.Column(sa.DateTime(), nullable=True))
last_run: Optional[datetime] = Field(sa_column=sa.Column(sa.DateTime(timezone=True), nullable=True))
last_error: Optional[str] = Field(nullable=True)
next_run: Optional[datetime] = Field(nullable=True)
backend/oasst_backend/models/message.py
+3 -1
View File
@@ -30,7 +30,9 @@ class Message(SQLModel, table=True):
api_client_id: UUID = Field(nullable=False, foreign_key="api_client.id")
frontend_message_id: str = Field(max_length=200, nullable=False)
created_date: Optional[datetime] = Field(
sa_column=sa.Column(sa.DateTime(), nullable=False, server_default=sa.func.current_timestamp(), index=True)
sa_column=sa.Column(
sa.DateTime(timezone=True), nullable=False, server_default=sa.func.current_timestamp(), index=True
)
)
payload_type: str = Field(nullable=False, max_length=200)
payload: Optional[PayloadContainer] = Field(
backend/oasst_backend/models/message_embedding.py
+1 -1
View File
@@ -17,5 +17,5 @@ class MessageEmbedding(SQLModel, table=True):
# In the case that the Message Embedding is created afterwards
created_date: Optional[datetime] = Field(
sa_column=sa.Column(sa.DateTime(), nullable=False, server_default=sa.func.current_timestamp())
sa_column=sa.Column(sa.DateTime(timezone=True), nullable=False, server_default=sa.func.current_timestamp())
)
backend/oasst_backend/models/message_reaction.py
+3 -1
View File
@@ -19,7 +19,9 @@ class MessageReaction(SQLModel, table=True):
sa_column=sa.Column(pg.UUID(as_uuid=True), sa.ForeignKey("user.id"), nullable=False, primary_key=True)
)
created_date: Optional[datetime] = Field(
sa_column=sa.Column(sa.DateTime(), nullable=False, server_default=sa.func.current_timestamp(), index=True)
sa_column=sa.Column(
sa.DateTime(timezone=True), nullable=False, server_default=sa.func.current_timestamp(), index=True
)
)
payload_type: str = Field(nullable=False, max_length=200)
payload: PayloadContainer = Field(sa_column=sa.Column(payload_column_type(PayloadContainer), nullable=False))
backend/oasst_backend/models/message_toxicity.py
+1 -1
View File
@@ -20,5 +20,5 @@ class MessageToxicity(SQLModel, table=True):
# In the case that the Message Embedding is created afterwards
created_date: Optional[datetime] = Field(
sa_column=sa.Column(sa.DateTime(), nullable=False, server_default=sa.func.current_timestamp())
sa_column=sa.Column(sa.DateTime(timezone=True), nullable=False, server_default=sa.func.current_timestamp())
)
backend/oasst_backend/models/payload_column_type.py
+2
View File
@@ -48,6 +48,8 @@ def payload_column_type(pydantic_type):
class PayloadJSONBType(TypeDecorator, Generic[T]):
impl = pg.JSONB()
cache_ok = True
def __init__(
self,
json_encoder=json,
backend/oasst_backend/models/task.py
+2 -2
View File
@@ -20,9 +20,9 @@ class Task(SQLModel, table=True):
),
)
created_date: Optional[datetime] = Field(
sa_column=sa.Column(sa.DateTime(), nullable=False, server_default=sa.func.current_timestamp()),
sa_column=sa.Column(sa.DateTime(timezone=True), nullable=False, server_default=sa.func.current_timestamp()),
)
expiry_date: Optional[datetime] = Field(sa_column=sa.Column(sa.DateTime(), nullable=True))
expiry_date: Optional[datetime] = Field(sa_column=sa.Column(sa.DateTime(timezone=True), nullable=True))
user_id: Optional[UUID] = Field(nullable=True, foreign_key="user.id", index=True)
payload_type: str = Field(nullable=False, max_length=200)
payload: PayloadContainer = Field(sa_column=sa.Column(payload_column_type(PayloadContainer), nullable=False))
backend/oasst_backend/models/text_labels.py
+3 -1
View File
@@ -17,7 +17,9 @@ class TextLabels(SQLModel, table=True):
)
user_id: UUID = Field(sa_column=sa.Column(pg.UUID(as_uuid=True), sa.ForeignKey("user.id"), nullable=False))
created_date: Optional[datetime] = Field(
sa_column=sa.Column(sa.DateTime(), nullable=False, server_default=sa.func.current_timestamp(), index=True),
sa_column=sa.Column(
sa.DateTime(timezone=True), nullable=False, server_default=sa.func.current_timestamp(), index=True
),
)
api_client_id: UUID = Field(nullable=False, foreign_key="api_client.id")
text: str = Field(nullable=False, max_length=2**16)
backend/oasst_backend/models/user.py
+1 -1
View File
@@ -21,7 +21,7 @@ class User(SQLModel, table=True):
auth_method: str = Field(nullable=False, max_length=128, default="local")
display_name: str = Field(nullable=False, max_length=256)
created_date: Optional[datetime] = Field(
sa_column=sa.Column(sa.DateTime(), nullable=False, server_default=sa.func.current_timestamp())
sa_column=sa.Column(sa.DateTime(timezone=True), nullable=False, server_default=sa.func.current_timestamp())
)
api_client_id: UUID = Field(foreign_key="api_client.id")
enabled: bool = Field(sa_column=sa.Column(sa.Boolean, nullable=False, server_default=sa.true()))
backend/oasst_backend/models/user_stats.py
+2 -2
View File
@@ -26,11 +26,11 @@ class UserStats(SQLModel, table=True):
user_id: Optional[UUID] = Field(
sa_column=sa.Column(pg.UUID(as_uuid=True), sa.ForeignKey("user.id"), primary_key=True)
)
base_date: Optional[datetime] = Field(sa_column=sa.Column(sa.DateTime(), nullable=True))
base_date: Optional[datetime] = Field(sa_column=sa.Column(sa.DateTime(timezone=True), nullable=True))
leader_score: int = 0
modified_date: Optional[datetime] = Field(
sa_column=sa.Column(sa.DateTime(), nullable=False, server_default=sa.func.current_timestamp())
sa_column=sa.Column(sa.DateTime(timezone=True), nullable=False, server_default=sa.func.current_timestamp())
)
rank: int = Field(nullable=True)
backend/oasst_backend/prompt_repository.py
+43 -18
View File
@@ -24,11 +24,11 @@ from oasst_backend.models import (
from oasst_backend.models.payload_column_type import PayloadContainer
from oasst_backend.task_repository import TaskRepository, validate_frontend_message_id
from oasst_backend.user_repository import UserRepository
from oasst_backend.utils.database_utils import CommitMode, managed_tx_method
from oasst_shared.exceptions import OasstError, OasstErrorCode
from oasst_shared.schemas import protocol as protocol_schema
from oasst_shared.schemas.protocol import SystemStats
from sqlalchemy import update
from sqlmodel import Session, func
from sqlmodel import Session, func, not_, text, update
from starlette.status import HTTP_403_FORBIDDEN, HTTP_404_NOT_FOUND
@@ -52,6 +52,13 @@ class PromptRepository:
)
self.journal = JournalWriter(db, api_client, self.user)
def ensure_user_is_enabled(self):
if self.user is None or self.user_id is None:
raise OasstError("User required", OasstErrorCode.USER_NOT_SPECIFIED)
if self.user.deleted or not self.user.enabled:
raise OasstError("User account disabled", OasstErrorCode.USER_DISABLED)
def fetch_message_by_frontend_message_id(self, frontend_message_id: str, fail_if_missing: bool = True) -> Message:
validate_frontend_message_id(frontend_message_id)
message: Message = (
@@ -67,6 +74,7 @@ class PromptRepository:
)
return message
@managed_tx_method(CommitMode.FLUSH)
def insert_message(
self,
*,
@@ -104,8 +112,8 @@ class PromptRepository:
review_result=review_result,
)
self.db.add(message)
self.db.commit()
self.db.refresh(message)
# self.db.refresh(message)
return message
def _validate_task(
@@ -134,6 +142,7 @@ class PromptRepository:
def fetch_tree_state(self, message_tree_id: UUID) -> MessageTreeState:
return self.db.query(MessageTreeState).filter(MessageTreeState.message_tree_id == message_tree_id).one()
@managed_tx_method(CommitMode.FLUSH)
def store_text_reply(
self,
text: str,
@@ -143,6 +152,8 @@ class PromptRepository:
review_result: bool = False,
check_tree_state: bool = True,
) -> Message:
self.ensure_user_is_enabled()
validate_frontend_message_id(frontend_message_id)
validate_frontend_message_id(user_frontend_message_id)
@@ -205,10 +216,10 @@ class PromptRepository:
if not task.collective:
task.done = True
self.db.add(task)
self.db.commit()
self.journal.log_text_reply(task=task, message_id=new_message_id, role=role, length=len(text))
return user_message
@managed_tx_method(CommitMode.FLUSH)
def store_rating(self, rating: protocol_schema.MessageRating) -> MessageReaction:
message = self.fetch_message_by_frontend_message_id(rating.message_id, fail_if_missing=True)
@@ -238,6 +249,7 @@ class PromptRepository:
logger.info(f"Ranking {rating.rating} stored for task {task.id}.")
return reaction
@managed_tx_method(CommitMode.COMMIT)
def store_ranking(self, ranking: protocol_schema.MessageRanking) -> Tuple[MessageReaction, Task]:
# fetch task
task = self.task_repository.fetch_task_by_frontend_message_id(ranking.message_id)
@@ -310,6 +322,7 @@ class PromptRepository:
return reaction, task
@managed_tx_method(CommitMode.FLUSH)
def insert_toxicity(self, message_id: UUID, model: str, score: float, label: str) -> MessageToxicity:
"""Save the toxicity score of a new message in the database.
Args:
@@ -325,10 +338,9 @@ class PromptRepository:
message_toxicity = MessageToxicity(message_id=message_id, model=model, score=score, label=label)
self.db.add(message_toxicity)
self.db.commit()
self.db.refresh(message_toxicity)
return message_toxicity
@managed_tx_method(CommitMode.FLUSH)
def insert_message_embedding(self, message_id: UUID, model: str, embedding: List[float]) -> MessageEmbedding:
"""Insert the embedding of a new message in the database.
@@ -346,13 +358,11 @@ class PromptRepository:
message_embedding = MessageEmbedding(message_id=message_id, model=model, embedding=embedding)
self.db.add(message_embedding)
self.db.commit()
self.db.refresh(message_embedding)
return message_embedding
@managed_tx_method(CommitMode.FLUSH)
def insert_reaction(self, task_id: UUID, payload: db_payload.ReactionPayload) -> MessageReaction:
if self.user_id is None:
raise OasstError("User required", OasstErrorCode.USER_NOT_SPECIFIED)
self.ensure_user_is_enabled()
container = PayloadContainer(payload=payload)
reaction = MessageReaction(
@@ -363,10 +373,9 @@ class PromptRepository:
payload_type=type(payload).__name__,
)
self.db.add(reaction)
self.db.commit()
self.db.refresh(reaction)
return reaction
@managed_tx_method(CommitMode.FLUSH)
def store_text_labels(self, text_labels: protocol_schema.TextLabels) -> Tuple[TextLabels, Task, Message]:
valid_labels: Optional[list[str]] = None
@@ -436,8 +445,6 @@ class PromptRepository:
self.db.add(message)
self.db.add(model)
self.db.commit()
self.db.refresh(model)
return model, task, message
def fetch_random_message_tree(self, require_role: str = None, reviewed: bool = True) -> list[Message]:
@@ -499,10 +506,14 @@ class PromptRepository:
messages = self.db.query(Message).filter(Message.parent_id.is_(None)).order_by(func.random()).limit(size).all()
return messages
def fetch_message_tree(self, message_tree_id: UUID, reviewed: bool = True):
def fetch_message_tree(
self, message_tree_id: UUID, reviewed: bool = True, include_deleted: bool = False
) -> list[Message]:
qry = self.db.query(Message).filter(Message.message_tree_id == message_tree_id)
if reviewed:
qry = qry.filter(Message.review_result)
if not include_deleted:
qry = qry.filter(not_(Message.deleted))
return qry.all()
def fetch_multiple_random_replies(self, max_size: int = 5, message_role: str = None):
@@ -702,6 +713,22 @@ class PromptRepository:
return messages.all()
def update_children_counts(self, message_tree_id: UUID):
sql_update_children_count = """
UPDATE message SET children_count = cc.children_count
FROM (
SELECT m.id, count(c.id) - COALESCE(SUM(c.deleted::int), 0) AS children_count
FROM message m
LEFT JOIN message c ON m.id = c.parent_id
WHERE m.message_tree_id = :message_tree_id
GROUP BY m.id
) AS cc
WHERE message.id = cc.id;
"""
r = self.db.execute(text(sql_update_children_count), {"message_tree_id": message_tree_id})
logger.debug(f"update_children_count({message_tree_id=}): {r.rowcount} rows.")
@managed_tx_method(CommitMode.COMMIT)
def mark_messages_deleted(self, messages: Message | UUID | list[Message | UUID], recursive: bool = True):
"""
Marks deleted messages and all their descendants.
@@ -730,8 +757,6 @@ class PromptRepository:
parent_ids = self.db.execute(query).scalars().all()
self.db.commit()
def get_stats(self) -> SystemStats:
"""
Get data stats such as number of all messages in the system,
backend/oasst_backend/task_repository.py
+5 -6
View File
@@ -6,6 +6,7 @@ from loguru import logger
from oasst_backend.models import ApiClient, Task
from oasst_backend.models.payload_column_type import PayloadContainer
from oasst_backend.user_repository import UserRepository
from oasst_backend.utils.database_utils import CommitMode, managed_tx_method
from oasst_shared.exceptions.oasst_api_error import OasstError, OasstErrorCode
from oasst_shared.schemas import protocol as protocol_schema
from sqlmodel import Session
@@ -128,6 +129,7 @@ class TaskRepository:
assert task_model.id == task.id
return task_model
@managed_tx_method(CommitMode.COMMIT)
def bind_frontend_message_id(self, task_id: UUID, frontend_message_id: str):
validate_frontend_message_id(frontend_message_id)
@@ -142,10 +144,9 @@ class TaskRepository:
task.frontend_message_id = frontend_message_id
task.ack = True
# ToDo: check race-condition, transaction
self.db.add(task)
self.db.commit()
@managed_tx_method(CommitMode.COMMIT)
def close_task(self, frontend_message_id: str, allow_personal_tasks: bool = False):
"""
Mark task as done. No further messages will be accepted for this task.
@@ -166,8 +167,8 @@ class TaskRepository:
task.done = True
self.db.add(task)
self.db.commit()
@managed_tx_method(CommitMode.COMMIT)
def acknowledge_task_failure(self, task_id):
# find task
task: Task = self.db.query(Task).filter(Task.id == task_id, Task.api_client_id == self.api_client.id).first()
@@ -181,8 +182,8 @@ class TaskRepository:
task.ack = False
# ToDo: check race-condition, transaction
self.db.add(task)
self.db.commit()
@managed_tx_method(CommitMode.COMMIT)
def insert_task(
self,
payload: db_payload.TaskPayload,
@@ -204,8 +205,6 @@ class TaskRepository:
)
logger.debug(f"inserting {task=}")
self.db.add(task)
self.db.commit()
self.db.refresh(task)
return task
def fetch_task_by_frontend_message_id(self, message_id: str) -> Task:
backend/oasst_backend/tree_manager.py
+625 -281
View File
File diff suppressed because it is too large Load Diff
backend/oasst_backend/user_repository.py
+4 -5
View File
@@ -2,6 +2,7 @@ from typing import Optional
from uuid import UUID
from oasst_backend.models import ApiClient, User
from oasst_backend.utils.database_utils import CommitMode, managed_tx_method
from oasst_shared.exceptions import OasstError, OasstErrorCode
from oasst_shared.schemas import protocol as protocol_schema
from sqlmodel import Session
@@ -62,6 +63,7 @@ class UserRepository:
return user
@managed_tx_method(CommitMode.COMMIT)
def update_user(self, id: UUID, enabled: Optional[bool] = None, notes: Optional[str] = None) -> None:
"""
Update a user by global user ID to disable or set admin notes. Only trusted clients may update users.
@@ -83,8 +85,8 @@ class UserRepository:
user.notes = notes
self.db.add(user)
self.db.commit()
@managed_tx_method(CommitMode.COMMIT)
def mark_user_deleted(self, id: UUID) -> None:
"""
Update a user by global user ID to set deleted flag. Only trusted clients may delete users.
@@ -103,8 +105,8 @@ class UserRepository:
user.deleted = True
self.db.add(user)
self.db.commit()
@managed_tx_method(CommitMode.COMMIT)
def lookup_client_user(self, client_user: protocol_schema.User, create_missing: bool = True) -> Optional[User]:
if not client_user:
return None
@@ -127,13 +129,10 @@ class UserRepository:
auth_method=client_user.auth_method,
)
self.db.add(user)
self.db.commit()
self.db.refresh(user)
elif client_user.display_name and client_user.display_name != user.display_name:
# we found the user but the display name changed
user.display_name = client_user.display_name
self.db.add(user)
self.db.commit()
return user
def query_users(
backend/oasst_backend/user_stats_repository.py
+13 -10
View File
@@ -4,6 +4,7 @@ from uuid import UUID
import sqlalchemy as sa
from loguru import logger
from oasst_backend.config import settings
from oasst_backend.models import Message, MessageReaction, Task, User, UserStats, UserStatsTimeFrame
from oasst_backend.models.db_payload import (
LabelAssistantReplyPayload,
@@ -39,12 +40,16 @@ class UserStatsRepository:
self.session.query(User.id.label("user_id"), User.username, User.auth_method, User.display_name, UserStats)
.join(UserStats, User.id == UserStats.user_id)
.filter(UserStats.time_frame == time_frame.value)
.order_by(UserStats.leader_score.desc())
.order_by(UserStats.rank)
.limit(limit)
)
leaderboard = [_create_user_score(r) for r in self.session.exec(qry)]
return LeaderboardStats(time_frame=time_frame.value, leaderboard=leaderboard)
if len(leaderboard) > 0:
last_update = max(x.modified_date for x in leaderboard)
else:
last_update = utcnow()
return LeaderboardStats(time_frame=time_frame.value, leaderboard=leaderboard, last_updated=last_update)
def get_user_stats_all_time_frames(self, user_id: UUID) -> dict[str, UserScore | None]:
qry = (
@@ -291,13 +296,11 @@ WHERE
if __name__ == "__main__":
from oasst_backend.api.deps import get_dummy_api_client
from oasst_backend.api.deps import api_auth
from oasst_backend.database import engine
with Session(engine) as session:
api_client = get_dummy_api_client(session)
usr = UserStatsRepository(session)
# usr.update_all_time_frames()
# session.commit()
# usr.get_leader_board(UserStatsTimeFrame.total)
usr.get_user_stats_all_time_frames(UUID("0d6ff62a-0bea-4c56-ade8-b3e0520a10ce"))
with Session(engine) as db:
api_client = api_auth(settings.OFFICIAL_WEB_API_KEY, db=db)
usr = UserStatsRepository(db)
usr.update_all_time_frames()
db.commit()
backend/oasst_backend/utils/database_utils.py
+143
View File
@@ -0,0 +1,143 @@
from enum import IntEnum
from functools import wraps
from http import HTTPStatus
from typing import Callable
from loguru import logger
from oasst_backend.config import settings
from oasst_backend.database import engine
from oasst_shared.exceptions import OasstError, OasstErrorCode
from sqlalchemy.exc import OperationalError
from sqlmodel import Session, SQLModel
class CommitMode(IntEnum):
"""
Commit modes for the managed tx methods
"""
NONE = 0
FLUSH = 1
COMMIT = 2
ROLLBACK = 3
"""
* managed_tx_method and async_managed_tx_method methods are decorators functions
* to be used on class functions. It expects the Class to have a 'db' Session object
* initialised
* TODO: tx method decorator for non class methods
"""
def managed_tx_method(auto_commit: CommitMode = CommitMode.COMMIT, num_retries=settings.DATABASE_MAX_TX_RETRY_COUNT):
def decorator(f):
@wraps(f)
def wrapped_f(self, *args, **kwargs):
try:
for i in range(num_retries):
try:
result = f(self, *args, **kwargs)
if auto_commit == CommitMode.COMMIT:
self.db.commit()
elif auto_commit == CommitMode.FLUSH:
self.db.flush()
elif auto_commit == CommitMode.ROLLBACK:
self.db.rollback()
if isinstance(result, SQLModel):
self.db.refresh(result)
return result
except OperationalError:
logger.info(f"Retry {i+1}/{num_retries} after possible DB concurrent update conflict.")
self.db.rollback()
raise OasstError(
"DATABASE_MAX_RETIRES_EXHAUSTED",
error_code=OasstErrorCode.DATABASE_MAX_RETRIES_EXHAUSTED,
http_status_code=HTTPStatus.SERVICE_UNAVAILABLE,
)
except Exception as e:
logger.error("DB Rollback Failure")
raise e
return wrapped_f
return decorator
def async_managed_tx_method(
auto_commit: CommitMode = CommitMode.COMMIT, num_retries=settings.DATABASE_MAX_TX_RETRY_COUNT
):
def decorator(f):
@wraps(f)
async def wrapped_f(self, *args, **kwargs):
try:
for i in range(num_retries):
try:
result = await f(self, *args, **kwargs)
if auto_commit == CommitMode.COMMIT:
self.db.commit()
elif auto_commit == CommitMode.FLUSH:
self.db.flush()
elif auto_commit == CommitMode.ROLLBACK:
self.db.rollback()
if isinstance(result, SQLModel):
self.db.refresh(result)
return result
except OperationalError:
logger.info(f"Retry {i+1}/{num_retries} after possible DB concurrent update conflict.")
self.db.rollback()
raise OasstError(
"DATABASE_MAX_RETIRES_EXHAUSTED",
error_code=OasstErrorCode.DATABASE_MAX_RETRIES_EXHAUSTED,
http_status_code=HTTPStatus.SERVICE_UNAVAILABLE,
)
except Exception as e:
logger.exception("DB Rollback Failure")
raise e
return wrapped_f
return decorator
def default_session_factor() -> Session:
return Session(engine)
def managed_tx_function(
auto_commit: CommitMode = CommitMode.COMMIT,
num_retries=settings.DATABASE_MAX_TX_RETRY_COUNT,
session_factory: Callable[..., Session] = default_session_factor,
):
"""Passes Session object as first argument to wrapped function."""
def decorator(f):
@wraps(f)
def wrapped_f(*args, **kwargs):
try:
for i in range(num_retries):
with session_factory() as session:
try:
result = f(session, *args, **kwargs)
if auto_commit == CommitMode.COMMIT:
session.commit()
elif auto_commit == CommitMode.FLUSH:
session.flush()
elif auto_commit == CommitMode.ROLLBACK:
session.rollback()
return result
except OperationalError:
logger.info(f"Retry {i+1}/{num_retries} after possible DB concurrent update conflict.")
session.rollback()
raise OasstError(
"DATABASE_MAX_RETIRES_EXHAUSTED",
error_code=OasstErrorCode.DATABASE_MAX_RETRIES_EXHAUSTED,
http_status_code=HTTPStatus.SERVICE_UNAVAILABLE,
)
except Exception as e:
logger.error("DB Rollback Failure")
raise e
return wrapped_f
return decorator
backend/oasst_backend/utils/ranking.py
+140
View File
@@ -0,0 +1,140 @@
from typing import List
import numpy as np
def head_to_head_votes(ranks: List[List[int]]):
tallies = np.zeros((len(ranks[0]), len(ranks[0])))
names = sorted(ranks[0])
ranks = np.array(ranks)
# we want the sorted indices
ranks = np.argsort(ranks, axis=1)
for i in range(ranks.shape[1]):
for j in range(i + 1, ranks.shape[1]):
# now count the cases someone voted for i over j
over_j = np.sum(ranks[:, i] < ranks[:, j])
over_i = np.sum(ranks[:, j] < ranks[:, i])
tallies[i, j] = over_j
# tallies[i,j] = over_i
tallies[j, i] = over_i
# tallies[j,i] = over_j
return tallies, names
def cycle_detect(pairs):
"""Recursively detect cylces by removing condorcet losers until either only one pair is left or condorcet loosers no longer exist
This method upholds the invariant that in a ranking for all a,b either a>b or b>a for all a,b.
Returns
-------
out : False if the pairs do not contain a cycle, True if the pairs contain a cycle
"""
# get all condorcet losers (pairs that loose to all other pairs)
# idea: filter all losers that are never winners
# print("pairs", pairs)
if len(pairs) <= 1:
return False
losers = [c_lose for c_lose in np.unique(pairs[:, 1]) if c_lose not in pairs[:, 0]]
if len(losers) == 0:
# if we recursively removed pairs, and at some point we did not have
# a condorcet loser, that means everything is both a winner and loser,
# yielding at least one (winner,loser), (loser,winner) pair
return True
new = []
for p in pairs:
if p[1] not in losers:
new.append(p)
return cycle_detect(np.array(new))
def get_winner(pairs):
"""
This returns _one_ concordant winner.
It could be that there are multiple concordant winners, but in our case
since we are interested in a ranking, we have to choose one at random.
"""
losers = np.unique(pairs[:, 1]).astype(int)
winners = np.unique(pairs[:, 0]).astype(int)
for w in winners:
if w not in losers:
return w
def get_ranking(pairs):
"""
Abuses concordance property to get a (not necessarily unqiue) ranking.
The lack of uniqueness is due to the potential existence of multiple
equally ranked winners. We have to pick one, which is where
the non-uniqueness comes from
"""
if len(pairs) == 1:
return list(pairs[0])
w = get_winner(pairs)
# now remove the winner from the list of pairs
p_new = np.array([(a, b) for a, b in pairs if a != w])
return [w] + get_ranking(p_new)
def ranked_pairs(ranks: List[List[int]]):
"""
Expects a list of rankings for an item like:
[("w","x","z","y") for _ in range(3)]
+ [("w","y","x","z") for _ in range(2)]
+ [("x","y","z","w") for _ in range(4)]
+ [("x","z","w","y") for _ in range(5)]
+ [("y","w","x","z") for _ in range(1)]
This code is quite brain melting, but the idea is the following:
1. create a head-to-head matrix that tallies up all win-lose combinations of preferences
2. take all combinations that win more than they loose and sort those by how often they win
3. use that to create an (implicit) directed graph
4. recursively extract nodes from the graph that do not have incoming edges
5. said recursive list is the ranking
"""
tallies, names = head_to_head_votes(ranks)
tallies = tallies - tallies.T
# print(tallies)
# note: the resulting tally matrix should be skew-symmetric
# order by strength of victory (using tideman's original method, don't think it would make a difference for us)
sorted_majorities = []
for i in range(len(ranks[0])):
for j in range(len(ranks[0])):
if tallies[i, j] > 0:
sorted_majorities.append((i, j, tallies[i, j]))
# we don't explicitly deal with tied majorities here
sorted_majorities = np.array(sorted(sorted_majorities, key=lambda x: x[2], reverse=True))
# now do lock ins
lock_ins = []
for (x, y, _) in sorted_majorities:
# invariant: lock_ins has no cycles here
lock_ins.append((x, y))
# print("lock ins are now",np.array(lock_ins))
if cycle_detect(np.array(lock_ins)):
# print("backup: cycle detected")
# if there's a cycle, delete the new addition and continue
lock_ins = lock_ins[:-1]
# now simply return all winners in order, and attach the losers
# to the back. This is because the overall loser might not be unique
# and (by concordance property) may never exist in any winning set to begin with.
# (otherwise he would either not be the loser, or cycles exist!)
# Since there could be multiple overall losers, we just return them in any order
# as we are unable to find a closer ranking
numerical_ranks = np.array(get_ranking(np.array(lock_ins))).astype(int)
conversion = [names[n] for n in numerical_ranks]
return conversion
if __name__ == "__main__":
ranks = (
[("w", "x", "z", "y") for _ in range(1)]
+ [("w", "y", "x", "z") for _ in range(2)]
# + [("x","y","z","w") for _ in range(4)]
+ [("x", "z", "w", "y") for _ in range(5)]
+ [("y", "w", "x", "z") for _ in range(1)]
# [("y","z","w","x") for _ in range(1000)]
)
rp = ranked_pairs(ranks)
print(rp)
deploy/prod-node/nginx/nginx.conf
+16 -1
View File
@@ -16,6 +16,19 @@ http {
}
}
server {
listen 443 ssl http2;
server_name www.open-assistant.io;
ssl_certificate /etc/nginx/ssl/live/www.open-assistant.io/fullchain.pem;
ssl_certificate_key /etc/nginx/ssl/live/www.open-assistant.io/privkey.pem;
location / {
return 301 https://open-assistant.io$request_uri;
}
}
server {
listen 443 ssl http2;
@@ -25,7 +38,9 @@ http {
ssl_certificate_key /etc/nginx/ssl/live/open-assistant.io/privkey.pem;
location / {
return 301 https://web.prod.open-assistant.io$request_uri;
proxy_set_header Host $host;
proxy_set_header X-Real-IP $remote_addr;
proxy_pass http://127.0.0.1:3200;
}
}
docker-compose.yaml
+2 -1
View File
@@ -18,7 +18,8 @@ services:
# This DB is for the FastAPI Backend.
db:
image: postgres
image: ghcr.io/laion-ai/open-assistant/oasst-postgres
pull_policy: always
restart: always
ports:
- 5432:5432
docker/oasst-postgres/Dockerfile
+11
View File
@@ -0,0 +1,11 @@
FROM postgres:15
# install unzip
RUN apt-get update && apt-get install -y unzip curl && rm -rf /var/lib/apt/lists/*
# download aws cli
RUN curl "https://awscli.amazonaws.com/awscli-exe-linux-x86_64.zip" -o "awscliv2.zip"
RUN unzip -q awscliv2.zip
RUN ./aws/install
COPY ./backup_pg_to_s3.sh .
docker/oasst-postgres/backup_pg_to_s3.sh
Executable
+15
View File
@@ -0,0 +1,15 @@
#!/bin/bash
set -e
set -x
# filename with timestamp
filename="postgres-$(date +%Y-%m-%d_%H-%M-%S).sql"
# perform pg_dump
pg_dump -U postgres postgres > /tmp/$filename
# upload to s3
aws s3 cp /tmp/$filename s3://$S3_BUCKET_NAME/$filename
rm /tmp/$filename
docs/docs/faq/README.md
+3
View File
@@ -0,0 +1,3 @@
# Frequently Asked Questions
In this page, there are some of the most frequently asked questions.
docs/docs/faq/faq.md
+65
View File
@@ -0,0 +1,65 @@
### Docker-Compose instead of Docker Compose
If you are using `docker-compose` instead of `docker compose` (note the " "
instead of the "-"), you should update your docker cli to the latest version.
`docker compose` is the most recent version and should be used instead of
`docker-compose`
For more details and information check out
[this SO thread](https://stackoverflow.com/questions/66514436/difference-between-docker-compose-and-docker-compose)
that explains it all in detail.
### Pre-commit
We are using pre-commit to ensure the quality of the code as well as the same
code standard.
The steps that you need to follow to be able to use it are:
```bash
# install pre-commit in your python environment
pip3 install pre-commit
# install pre-commit in your github configuration
pre-commit install
```
So from now on, in your next commits it will run the `pre-commit` on the files
that have been staged. If there has been any error, you will need to solve that,
and then stage+commit again the changes.
## Docker Cannot Start Container: Permission Denied
Instead of running docker with the root command always, you could create a
`docker` group with granted permissions (root):
```bash
# Create new linux user
sudo groupadd docker
# Add the actual user to the group
sudo usermod -aG docker $USER
# Log in the group (apply the group changes to actual terminal session)
newgrp docker
```
After that, you should be able to run docker: `docker run .`. In the case you
still are not able, can try to reboot terminal:
```bash
reboot
```
### Docker Cannot Stop Container
If you try to shut down the services (`docker-compose down`), and you are
getting permission denied (using root user), you can try the following:
```bash
# Restart docker daemon
sudo systemctl restart docker.socket docker.service
# And remove the container
docker rm -f <container id>
```
docs/docs/research/general.md
+124 -8
View File
@@ -8,15 +8,29 @@ This page lists research papers that are relevant to the project.
- Generating Text From Language Models
- Automatically Generating Instruction Data for Training
- Uncertainty Estimation of Language Model Outputs
- Evidence-Guided Text Generation
- Reward Model Optimization
- Dialogue-Oriented RLHF
- Reduce Harms in Language Models
## Reinforcement Learning from Human Feedback <a name="reinforcement-learning-from-human-feedback"></a>
## Reinforcement Learning from Human Feedback
Reinforcement Learning from Human Feedback (RLHF) is a method for fine-tuning a
generative language models based on a reward model that is learned from human
preference data. This method facilitates the learning of instruction-tuned
models, among other things.
### Learning to summarize from human feedback [[ArXiv](https://arxiv.org/pdf/2009.01325.pdf)], [[Github](https://github.com/openai/summarize-from-feedback)]
### Fine-Tuning Language Models from Human Preferences [[ArXiv](https://arxiv.org/abs/1909.08593)], [[GitHub](https://github.com/openai/lm-human-preferences)]
> In this paper, we build on advances in generative pretraining of language
> models to apply reward learning to four natural language tasks: continuing
> text with positive sentiment or physically descriptive language, and
> summarization tasks on the TL;DR and CNN/Daily Mail datasets. For stylistic
> continuation we achieve good results with only 5,000 comparisons evaluated by
> humans. For summarization, models trained with 60,000 comparisons copy whole
> sentences from the input but skip irrelevant preamble.
### Learning to summarize from human feedback [[ArXiv](https://arxiv.org/abs/2009.01325)], [[GitHub](https://github.com/openai/summarize-from-feedback)]
> In this work, we show that it is possible to significantly improve summary
> quality by training a model to optimize for human preferences. We collect a
@@ -24,7 +38,18 @@ models, among other things.
> model to predict the human-preferred summary, and use that model as a reward
> function to fine-tune a summarization policy using reinforcement learning.
### Training language models to follow instructions with human feedback [[ArXiv](https://arxiv.org/pdf/2203.02155.pdf)]
### Recursively Summarizing Books with Human Feedback [[ArXiv](https://arxiv.org/abs/2109.10862)]
> Our method combines learning from human feedback with recursive task
> decomposition: we use models trained on smaller parts of the task to assist
> humans in giving feedback on the broader task. We collect a large volume of
> demonstrations and comparisons from human labelers. Our resulting model
> generates sensible summaries of entire books, even matching the quality of
> human-written summaries in a few cases (∼5% of books). We achieve
> state-of-the-art results on the recent BookSum dataset for book-length
> summarization. We release datasets of samples from our model.
### Training language models to follow instructions with human feedback [[ArXiv](https://arxiv.org/abs/2203.02155)]
> Starting with a set of labeler-written prompts and prompts submitted through
> the OpenAI API, we collect a dataset of labeler demonstrations of the desired
@@ -33,7 +58,7 @@ models, among other things.
> fine-tune this supervised model using reinforcement learning from human
> feedback.
### Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback [[ArXiv](https://arxiv.org/pdf/2204.05862.pdf)]
### Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback [[ArXiv](https://arxiv.org/abs/2204.05862)]
> We apply preference modeling and reinforcement learning from human feedback
> (RLHF) to finetune language models to act as helpful and harmless assistants.
@@ -41,6 +66,31 @@ models, among other things.
> evaluations, and is fully compatible with training for specialized skills such
> as python coding and summarization.
### Self-critiquing models for assisting human evaluators [[ArXiv](https://arxiv.org/abs/2206.05802)]
> We fine-tune large language models to write natural language critiques
> (natural language critical comments) using behavioral cloning. On a
> topic-based summarization task, critiques written by our models help humans
> find flaws in summaries that they would have otherwise missed. We study
> scaling properties of critiquing with both topic-based summarization and
> synthetic tasks. Finally, we motivate and introduce a framework for comparing
> critiquing ability to generation and discrimination ability. These results are
> a proof of concept for using AI-assisted human feedback to scale the
> supervision of machine learning systems to tasks that are difficult for humans
> to evaluate directly. We release our training datasets.
### Is Reinforcement Learning (Not) for Natural Language Processing?: Benchmarks, Baselines, and Building Blocks for Natural Language Policy Optimization [[ArXiv](https://arxiv.org/abs/2210.01241)]
> We tackle the problem of aligning pre-trained large language models (LMs) with
> human preferences. We present the GRUE (General Reinforced-language
> Understanding Evaluation) benchmark, a set of 6 language generation tasks
> which are supervised by reward functions which capture automated measures of
> human preference. Finally, we introduce an easy-to-use, performant RL
> algorithm, NLPO (Natural Language Policy Optimization) that learns to
> effectively reduce the combinatorial action space in language generation. We
> show that RL techniques are generally better than supervised methods at
> aligning LMs to human preferences.
## Generating Text From Language Models
A language model generates output text token by token, autoregressively. The
@@ -48,7 +98,7 @@ large search space of this task requires some method of narrowing down the set
of tokens to be considered in each step. This method, in turn, has a big impact
on the quality of the resulting text.
### RANKGEN: Improving Text Generation with Large Ranking Models [[ArXiv](https://arxiv.org/pdf/2205.09726.pdf)], [[Github](https://github.com/martiansideofthemoon/rankgen)]
### RANKGEN: Improving Text Generation with Large Ranking Models [[ArXiv](https://arxiv.org/abs/2205.09726)], [[GitHub](https://github.com/martiansideofthemoon/rankgen)]
> Given an input sequence (or prefix), modern language models often assign high
> probabilities to output sequences that are repetitive, incoherent, or
@@ -65,7 +115,7 @@ annotated data for the purpose of training
[instruction-aligned](https://openai.com/blog/instruction-following/) language
models.
### SELF-INSTRUCT: Aligning Language Model with Self Generated Instructions [[ArXiv](https://arxiv.org/pdf/2212.10560.pdf)], [[Github](https://github.com/yizhongw/self-instruct)].
### SELF-INSTRUCT: Aligning Language Model with Self Generated Instructions [[ArXiv](https://arxiv.org/abs/2212.10560)], [[GitHub](https://github.com/yizhongw/self-instruct)].
> We introduce SELF-INSTRUCT, a framework for improving the
> instruction-following capabilities of pretrained language models by
@@ -76,7 +126,7 @@ models.
> SuperNaturalInstructions, on par with the performance of InstructGPT-0011,
> which is trained with private user data and human annotations.
### Tuning Language Models with (Almost) No Human Labor. [[ArXiv](https://arxiv.org/pdf/2212.09689.pdf)], [[Github](https://github.com/orhonovich/unnatural-instructions)].
### Tuning Language Models with (Almost) No Human Labor. [[ArXiv](https://arxiv.org/abs/2212.09689)], [[GitHub](https://github.com/orhonovich/unnatural-instructions)].
> In this work, we introduce Unnatural Instructions: a large dataset of creative
> and diverse instructions, collected with virtually no human labor. We collect
@@ -91,7 +141,7 @@ models.
## Uncertainty Estimation of Language Model Outputs
### Teaching models to express their uncertainty in words [[Arxiv](https://arxiv.org/pdf/2205.14334.pdf)]
### Teaching models to express their uncertainty in words [[ArXiv](https://arxiv.org/abs/2205.14334)]
> We show that a GPT-3 model can learn to express uncertainty about its own
> answers in natural language -- without use of model logits. When given a
@@ -100,3 +150,69 @@ models.
> are well calibrated. The model also remains moderately calibrated under
> distribution shift, and is sensitive to uncertainty in its own answers, rather
> than imitating human examples.
## Evidence-Guided Text Generation
### WebGPT: Browser-assisted question-answering with human feedback [[ArXiv](https://arxiv.org/abs/2112.09332)]
> We fine-tune GPT-3 to answer long-form questions using a text-based
> web-browsing environment, which allows the model to search and navigate the
> web. We are able to train models on the task using imitation learning, and
> then optimize answer quality with human feedback. Models must collect
> references while browsing in support of their answers. Our best model is
> obtained by fine-tuning GPT-3 using behavior cloning, and then performing
> rejection sampling against a reward model.
### Teaching language models to support answers with verified quotes [[ArXiv](https://arxiv.org/abs/2203.11147)]
> In this work we use RLHF to train "open-book" QA models that generate answers
> whilst also citing specific evidence for their claims, which aids in the
> appraisal of correctness. Supporting evidence is drawn from multiple documents
> found via a search engine, or from a single user-provided document. However,
> analysis on the adversarial TruthfulQA dataset shows why citation is only one
> part of an overall strategy for safety and trustworthiness: not all claims
> supported by evidence are true.
## Reward Model Optimization
### Scaling Laws for Reward Model Overoptimization [[ArXiv](https://arxiv.org/abs/2210.10760)], [[Preceding Blogpost](https://openai.com/blog/measuring-goodharts-law/)]
> In this work, we use a synthetic setup in which a fixed "gold-standard" reward
> model plays the role of humans, providing labels used to train a proxy reward
> model. We study how the gold reward model score changes as we optimize against
> the proxy reward model using either reinforcement learning or best-of-n
> sampling. We study the effect on this relationship of the size of the reward
> model dataset. We explore the implications of these empirical results for
> theoretical considerations in AI alignment.
## Dialogue-Oriented RLHF
### Dynamic Planning in Open-Ended Dialogue using Reinforcement Learning [[ArXiv](https://arxiv.org/abs/2208.02294)]
> Building automated agents that can carry on rich open-ended conversations with
> humans "in the wild" remains a formidable challenge. In this work we develop a
> real-time, open-ended dialogue system that uses reinforcement learning (RL) to
> power a bot's conversational skill at scale. Trained using crowd-sourced data,
> our novel system is able to substantially exceeds several metrics of interest
> in a live experiment with real users of the Google Assistant.
### Improving alignment of dialogue agents via targeted human judgements [[ArXiv](https://arxiv.org/abs/2209.14375)]
> We present Sparrow, an information-seeking dialogue agent trained to be more
> helpful, correct, and harmless compared to prompted language model baselines
> First, to make our agent more helpful and harmless, we break down the
> requirements for good dialogue into natural language rules the agent should
> followy. Second, our agent provides evidence from sources supporting factual
> claims when collecting preference judgements over model statements.Finally, we
> conduct extensive analyses showing that though our model learns to follow our
> rules it can exhibit distributional biases.
## Reduce Harms in Language Models
### Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned [[ArXiv](https://arxiv.org/abs/2209.07858)]
> We investigate scaling behaviors for red teaming. We find that the RLHF models
> are increasingly difficult to red team as they scale, and we find a flat trend
> with scale for the other model types. We exhaustively describe our
> instructions, processes, statistical methodologies, and uncertainty about red
> teaming.
docs/sidebars.js
+9
View File
@@ -70,6 +70,15 @@ const sidebars = {
},
items: ["presentations/list"],
},
{
type: "category",
label: "FAQ",
link: {
type: "doc",
id: "faq/README",
},
items: ["faq/faq"],
},
],
};
model/supervised_finetuning/configs/config.yaml
+1
View File
@@ -38,6 +38,7 @@ defaults:
- instruct_tuning
- wmt2019_de-en
- samsum
- soda_dialogue
cache_dir: .cache
loss_fn: CrossEntropyLoss
eval_size:
model/supervised_finetuning/custom_datasets/__init__.py
+4 -1
View File
@@ -2,7 +2,7 @@
High level functions for model training
"""
from custom_datasets.prompt_dialogue import InstructionTuning, PromptGeneratedDataset
from custom_datasets.qa_datasets import SODA, JokeExplaination, QADataset, WebGPT
from custom_datasets.qa_datasets import SODA, JokeExplaination, QADataset, WebGPT, SODADialogue
from custom_datasets.summarization import SummarizationDataset
from custom_datasets.toxic_conversation import ProsocialDialogue, ProsocialDialogueExplaination
from custom_datasets.translation import WMT2019, DiveMT, TEDTalk
@@ -71,6 +71,9 @@ def get_one_dataset(conf, dataset_name):
elif dataset_name == "soda":
dataset = SODA(conf.cache_dir)
train, eval = train_val_dataset(dataset, val_split=0.1)
elif dataset_name == "soda_dialogue":
dataset = SODADialogue(conf.cache_dir)
train, eval = train_val_dataset(dataset, val_split=0.1)
elif dataset_name == "joke":
dataset = JokeExplaination(conf.cache_dir)
train, eval = train_val_dataset(dataset, val_split=0.2)
model/supervised_finetuning/custom_datasets/qa_datasets.py
+60 -2
View File
@@ -121,7 +121,12 @@ class SODA(Dataset):
def process_soda_convo(self, data):
pairs = []
play_as = data["speakers"][1]
prefix = "<prefix>{}. {}</prefix>".format(data["narrative"], "your name {}".format(play_as))
prefix = "{}{}. {}{}".format(
QA_SPECIAL_TOKENS["StartPrefix"],
data["narrative"],
"your name {}".format(play_as),
QA_SPECIAL_TOKENS["EndPrefix"],
)
question, answer = "", ""
prefix, postfix = "", ""
previous_chat = []
@@ -134,7 +139,9 @@ class SODA(Dataset):
answer = convo
postfix = data["speakers"][idx]
if len(question) and len(answer) and prefix != postfix and postfix == play_as:
history = "<sep>".join(["{}<bot>{}".format(*p) for p in previous_chat])
history = "<sep>".join(
["{}{}{}".format(p[0], QA_SPECIAL_TOKENS["Answer"], p[1]) for p in previous_chat]
)
if len(history):
history += "<sep>"
pairs.append((prefix + history + question, answer))
@@ -163,6 +170,57 @@ class SODA(Dataset):
return question, answer
class SODADialogue(Dataset):
url = "https://drive.google.com/uc?id=1TOGQfr419n8wpzJpYLLw4nB3tSKD8zXV"
def __init__(self, cache_dir, verbose=True):
path = os.path.join(cache_dir, "soda_dialog.jsonl")
if not os.path.exists(path):
import gzip
import shutil
import gdown
gdown.download(self.url, output=os.path.join(cache_dir, "soda_dialog.jsonl.gz"))
with gzip.open(os.path.join(cache_dir, "soda_dialog.jsonl.gz"), "rb") as f_in:
with open(path, "wb") as f_out:
shutil.copyfileobj(f_in, f_out)
self.pairs = []
faulty = 0
with open(path) as fin:
for line in fin:
conversation = json.loads(line)
question_answer_pairs = ()
question_answers = conversation["text"].split("User: ")
for question_answer in question_answers[1:]: # first element is empty
try:
question, answer = question_answer.split("\nAssistant: ")
question_answer_pairs += (
question,
answer,
)
except ValueError:
# there might be some extra 'User: ' or 'Assistant: ' tokens in the dataset that cause trouble..
faulty += 1
continue
self.pairs.append(question_answer_pairs)
if verbose:
print("For SODA dialogue dataset found {} faults within the total {} dialogs".format(faulty, len(self)))
def __len__(self):
return len(self.pairs)
def __getitem__(self, index):
return self.pairs[index]
class JokeExplaination(Dataset):
name = "joke"
model/supervised_finetuning/requirements.txt
+5 -3
View File
@@ -3,10 +3,12 @@ bitsandbytes==0.36.0.post2
datasets==2.8.0
deepspeed==0.7.7
evaluate==0.4.0
gdown
mpi4py==3.1.4
nltk==3.8.1
numpy==1.23.0
PyYAML==6.0
numpy>=1.22.4
py7zr
PyYAML>=6.0
scikit_learn==1.2.0
torch==1.13.1
torch>=1.11.0
transformers==4.25.1
notebooks/data-augmentation/codet-data/Augment_CodeT_codegen.ipynb
+198
View File
@@ -0,0 +1,198 @@
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## CodeT Code Generation Datasets\n",
"\n",
"[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/LAION-AI/Open-Assistant/blob/main/notebooks/data-augmentation/codet-data/Augment_CodeT_codegen.ipynb)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"This notebook contains code to parse CodeT code generation prompt and solution data and modify to `(prompt, solution)` pairs outputted in a `.jsonl` file.\n",
"\n",
"Requirements: `requests`"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
"import json\n",
"from pathlib import Path\n",
"import requests\n",
"from typing import Dict, List, Tuple"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"DATA_FILES: List[str] = [\n",
" \"HumanEval_for_code_generation.jsonl\",\n",
" \"mbpp_sanitized_for_code_generation.jsonl\",\n",
"]\n",
"\n",
"OUT_FILES: List[str] = [\n",
" \"HumanEval_codegen.jsonl\",\n",
" \"mbpp_codegen.jsonl\",\n",
"]\n",
"\n",
"FILE_PATHS: List[Path] = [Path(f\"data/{data_file}\") for data_file in DATA_FILES]\n",
"\n",
"OUT_PATHS: List[Path] = [Path(f\"data/augmented/{out_file}\") for out_file in OUT_FILES]"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [],
"source": [
"def download_file(filename: str):\n",
" url = f\"https://raw.githubusercontent.com/microsoft/CodeT/main/CodeT/data/dataset/{filename}\"\n",
" response = requests.get(url)\n",
" with open(f\"data/{filename}\", \"wb\") as f:\n",
" f.write(response.content)\n",
"\n",
"\n",
"for filename in DATA_FILES:\n",
" download_file(filename)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"We can find the docstring, use its contents as the instruction (prefixed with \"Write a function corresponding to the docstring:\") and then use the content prior to the docstring and the canonical solution as the response."
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [],
"source": [
"def get_docstring_indices(prompt_lines: List[str]) -> Tuple[int, int]:\n",
" docstring_start, docstring_end = None, None\n",
"\n",
" for i, line in enumerate(prompt_lines):\n",
" if not (line.strip().startswith('\"\"\"') or line.strip().startswith(\"'''\")):\n",
" continue\n",
" if docstring_start:\n",
" docstring_end = i\n",
" break\n",
" docstring_start = i\n",
"\n",
" if docstring_end:\n",
" return docstring_start, docstring_end\n",
" raise ValueError(f\"No complete docstring found!\\n{prompt_lines}\")\n",
"\n",
"\n",
"def get_before(prompt_lines: List[str], before: int) -> List[str]:\n",
" before_lines = prompt_lines[:before]\n",
" return before_lines\n",
"\n",
"\n",
"def get_between(prompt_lines: List[str], start: int, end: int) -> List[str]:\n",
" between_lines = prompt_lines[start:end]\n",
" return between_lines"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [],
"source": [
"def get_request_and_solution(sample: dict) -> Tuple[List[str], List[str]]:\n",
" prompt = sample[\"prompt\"]\n",
" prompt_lines = prompt.splitlines()\n",
"\n",
" docstring_start, docstring_end = get_docstring_indices(prompt_lines)\n",
"\n",
" # Extract prompt\n",
" in_docstring = get_between(prompt_lines, docstring_start, docstring_end)\n",
" if '\"\"\"' in in_docstring[0] or \"'''\" in in_docstring[0]:\n",
" in_docstring[0] = in_docstring[0].replace('\"\"\"', \"\").replace(\"...\", \"\").strip()\n",
" request = \"Write a Python function corresponding to the docstring: \" + \" \".join([p.strip() for p in in_docstring])\n",
"\n",
" # Extract solution\n",
" before_docstring = get_before(prompt_lines, docstring_start)\n",
" after_docstring = sample[\"canonical_solution\"].splitlines()\n",
" solution = before_docstring + after_docstring\n",
" # Gets rid of consecutive empty lines\n",
" solution = [v for i, v in enumerate(solution) if v != \"\" or v != solution[i - 1]]\n",
" solution = \"\\n\".join(solution)\n",
"\n",
" return request, solution"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [],
"source": [
"def process_file(file_path: Path, out_path: Path):\n",
" lines = file_path.read_text().splitlines()\n",
" samples = list(map(json.loads, lines))\n",
"\n",
" output = []\n",
" for sample in samples:\n",
" prompt, solution = get_request_and_solution(sample)\n",
" output.append({\"prompt\": prompt, \"solution\": solution})\n",
"\n",
" with open(out_path, \"w\") as f:\n",
" for sample in output:\n",
" f.write(json.dumps(sample))\n",
" f.write(\"\\n\")"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [],
"source": [
"for file_path, out_path in zip(FILE_PATHS, OUT_PATHS):\n",
" process_file(file_path, out_path)"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3.10.5 ('venv': venv)",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.10.5"
},
"orig_nbformat": 4,
"vscode": {
"interpreter": {
"hash": "1f9a0efd3e4a33b8f30a65df6ca5a95cc3f93ce2f11519ee8c13fe711de61465"
}
}
},
"nbformat": 4,
"nbformat_minor": 2
}
notebooks/data-augmentation/codet-data/Augment_CodeT_testgen.ipynb
+191
View File
@@ -0,0 +1,191 @@
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## CodeT Test Generation Datasets\n",
"\n",
"[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/LAION-AI/Open-Assistant/blob/main/notebooks/data-augmentation/codet-data/Augment_CodeT_testgen.ipynb)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"This notebook contains code to parse CodeT test case generation prompt and solution data and modify to `(prompt, solution)` pairs outputted in a `.jsonl` file.\n",
"\n",
"Requirements: `requests`"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
"import json\n",
"from pathlib import Path\n",
"import requests\n",
"from typing import List, Tuple"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"DATA_FILES: List[str] = [\n",
" \"HumanEval_for_test_case_generation.jsonl\",\n",
" \"mbpp_sanitized_for_test_case_generation.jsonl\",\n",
"]\n",
"\n",
"OUT_FILES: List[str] = [\n",
" \"HumanEval_testgen.jsonl\",\n",
" \"mbpp_testgen.jsonl\",\n",
"]\n",
"\n",
"FILE_PATHS: List[Path] = [Path(f\"data/{data_file}\") for data_file in DATA_FILES]\n",
"\n",
"OUT_PATHS: List[Path] = [Path(f\"data/augmented/{out_file}\") for out_file in OUT_FILES]"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [],
"source": [
"def download_file(filename: str):\n",
" url = f\"https://raw.githubusercontent.com/microsoft/CodeT/main/CodeT/data/dataset/{filename}\"\n",
" response = requests.get(url)\n",
" with open(f\"data/{filename}\", \"wb\") as f:\n",
" f.write(response.content)\n",
"\n",
"\n",
"for filename in DATA_FILES:\n",
" download_file(filename)"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [],
"source": [
"def get_docstring_indices(prompt_lines: List[str]) -> Tuple[int, int]:\n",
" docstring_start, docstring_end = None, None\n",
"\n",
" for i, line in enumerate(prompt_lines):\n",
" if not (line.strip().startswith('\"\"\"') or line.strip().startswith(\"'''\")):\n",
" continue\n",
" if docstring_start:\n",
" docstring_end = i\n",
" break\n",
" docstring_start = i\n",
"\n",
" if docstring_end:\n",
" return docstring_start, docstring_end\n",
" raise ValueError(f\"No complete docstring found!\\n{prompt_lines}\")\n",
"\n",
"\n",
"def get_between(prompt_lines: List[str], start: int, end: int) -> List[str]:\n",
" between_lines = prompt_lines[start:end]\n",
" return between_lines"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {},
"outputs": [],
"source": [
"def get_request(sample: dict) -> List[str]:\n",
" prompt = sample[\"prompt\"]\n",
" prompt_lines = prompt.splitlines()\n",
"\n",
" docstring_start, docstring_end = get_docstring_indices(prompt_lines)\n",
"\n",
" # Extract prompt\n",
" in_docstring = get_between(prompt_lines, docstring_start, docstring_end)\n",
" if '\"\"\"' in in_docstring[0] or \"'''\" in in_docstring[0]:\n",
" in_docstring[0] = in_docstring[0].replace('\"\"\"', \"\").replace(\"...\", \"\").strip()\n",
" request = \"Write a test for a Python function with the following docstring: \" + \" \".join(\n",
" [p.strip() for p in in_docstring]\n",
" )\n",
"\n",
" return request\n",
"\n",
"\n",
"def get_test_code(sample: dict) -> List[str]:\n",
" test = sample[\"test\"]\n",
" test_lines = test.splitlines()\n",
" start = 0\n",
" for i, line in enumerate(test_lines):\n",
" if \"def check(\" in line:\n",
" start = i\n",
" return \"\\n\".join(test_lines[start:])"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {},
"outputs": [],
"source": [
"def process_file(file_path: Path, out_path: Path):\n",
" lines = file_path.read_text().splitlines()\n",
" samples = list(map(json.loads, lines))\n",
"\n",
" output = []\n",
" for sample in samples:\n",
" prompt = get_request(sample)\n",
" test = get_test_code(sample)\n",
" output.append({\"prompt\": prompt, \"solution\": test})\n",
"\n",
" with open(out_path, \"w\") as f:\n",
" for sample in output:\n",
" f.write(json.dumps(sample))\n",
" f.write(\"\\n\")"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {},
"outputs": [],
"source": [
"for file_path, out_path in zip(FILE_PATHS, OUT_PATHS):\n",
" process_file(file_path, out_path)"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3.10.5 ('venv': venv)",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.10.5"
},
"orig_nbformat": 4,
"vscode": {
"interpreter": {
"hash": "1f9a0efd3e4a33b8f30a65df6ca5a95cc3f93ce2f11519ee8c13fe711de61465"
}
}
},
"nbformat": 4,
"nbformat_minor": 2
}
notebooks/data-augmentation/codet-data/README.md
+15
View File
@@ -0,0 +1,15 @@
# CodeT Datasets
This folder contains two notebooks.
One will download the data used for Microsoft CodeT for tuning a model for
Python code generation from function docstrings, augment the data into prompt
and solution pairs and write them to `.jsonl` files.
The other will download the data used for Microsoft CodeT for tuning a model for
Python test generation from corresponding function docstrings, augment the data
into prompt and solution pairs and write them to `.jsonl` files.
## Requirements
Both notebooks require the library `requests`.
notebooks/diverse/README.md
+11
View File
@@ -0,0 +1,11 @@
# DIVERSE Downloader
Diverse is a notebook that downloads the DIVERSE dataset and converts it into
OpenAssistant Data Scheme formats.
---
## Contributing
Feel free to contribute to this notebook. It's not perfect and additional
functionality is planned.
notebooks/diverse/diverse.ipynb
+477
View File
@@ -0,0 +1,477 @@
{
"cells": [
{
"cell_type": "markdown",
"id": "00b2848c",
"metadata": {},
"source": [
"# Diverse Dataset"
]
},
{
"cell_type": "markdown",
"source": [
"[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1CmXjXVrmPtpAVBaogBSuDclM0O6Zzewf?usp=sharing)"
],
"metadata": {
"collapsed": false
}
},
{
"cell_type": "markdown",
"id": "d81932b9",
"metadata": {},
"source": [
"The purpose of this notebook is to download the DIVERSE dataset and convert it into a format that can be used for training the OpenAssistant.\n",
"\n",
"The DIVERSE repo can be found here: https://github.com/microsoft/CodeT/tree/main/DIVERSE\n",
"\n",
"If you extend or use this work, please cite the relevant papers:\n",
"```\n",
"@article{li2022advance,\n",
" title={On the Advance of Making Language Models Better Reasoners},\n",
" author={Li, Yifei and Lin, Zeqi and Zhang, Shizhuo and Fu, Qiang and Chen, Bei and Lou, Jian-Guang and Chen, Weizhu},\n",
" journal={arXiv preprint arXiv:2206.02336},\n",
" year={2022}\n",
"}\n",
"\n",
"```"
]
},
{
"cell_type": "markdown",
"id": "a8c98078",
"metadata": {},
"source": [
"# OpenAssistant Data Scheme"
]
},
{
"cell_type": "markdown",
"id": "2731f88f",
"metadata": {},
"source": [
"We will use the data scheme that can be found in the docs for Open-Assistant. This code is taken from the StackExchange notebook."
]
},
{
"cell_type": "code",
"execution_count": 13,
"id": "d35ab066",
"metadata": {},
"outputs": [],
"source": [
"from typing import TypeVar, List, Dict, Any, Literal\n",
"from json import JSONEncoder\n",
"\n",
"T = TypeVar(\"T\", bound=\"ConversationTreeNode\")\n",
"\n",
"\n",
"class ConversationTreeNode:\n",
" text: str # The text of the node\n",
" role: Literal[\"prompter\", \"assistant\"] # Whether the node is a user prompt/follow-up or an assistant response\n",
" children: List[T] # The children of the node (if you have a linear conversation, this will be of length 0 or 1)\n",
" metadata: Dict[str, Any] # Node metadata (see below)\n",
"\n",
" def __init__(\n",
" self, text: str, role: Literal[\"prompter\", \"assistant\"], children: List[T], metadata: Dict[str, Any]\n",
" ) -> None:\n",
" self.text = text\n",
" self.role = role\n",
" self.children = children\n",
" self.metadata = metadata\n",
"\n",
"\n",
"class ConversationTree:\n",
" root: ConversationTreeNode # The node containing the initial prompt\n",
" metadata: Dict[str, Any] # Tree metadata, different from root node metadata.\n",
"\n",
" def __init__(self, root: ConversationTreeNode, metadata: Dict[str, Any]) -> None:\n",
" self.root = root\n",
" self.metadata = metadata\n",
"\n",
"\n",
"# subclass JSONEncoder\n",
"class TreeEncoder(JSONEncoder):\n",
" def default(self, o):\n",
" return o.__dict__"
]
},
{
"cell_type": "markdown",
"id": "e7457bae",
"metadata": {},
"source": [
"# Download and convert"
]
},
{
"cell_type": "markdown",
"id": "54b0fd63",
"metadata": {},
"source": [
"We firstly import pandas and any other libraries that we'll need."
]
},
{
"cell_type": "code",
"execution_count": 14,
"id": "9317d4b4",
"metadata": {},
"outputs": [],
"source": [
"import pandas as pd\n",
"import json"
]
},
{
"cell_type": "markdown",
"id": "62dc4e18",
"metadata": {},
"source": [
"The following is a simple function to take the data (which has two columns) and convert it to a tree with a root note (question) and one child (answer)."
]
},
{
"cell_type": "code",
"execution_count": 19,
"id": "963e0d92",
"metadata": {},
"outputs": [],
"source": [
"import re\n",
"\n",
"\n",
"def convert_diverse(dataset_json_path):\n",
" # read files using pandas\n",
" ds = pd.read_json(dataset_json_path, lines=True)\n",
"\n",
" # create dataset name from path\n",
" ds_name = \"diverse\" + file.split(\"data\")[-1].replace(\"/\", \"_\").split(\".\")[0]\n",
" print(\"*****\", ds_name, \"****\")\n",
" print(\"Example of raw dataset\")\n",
" print(ds.head(2))\n",
"\n",
" # create conversation forest\n",
" # Print first sample so the user of this notebook has an idea of what he's looking at\n",
" first_sample = True\n",
" print(\"\\nExamples from converted dataset\")\n",
" conversation_forest = []\n",
" for item in ds[\"context\"]:\n",
" # build nodes and tree\n",
" # Find all answers:\n",
"\n",
" answers = re.findall(r\"Answer:?(.*?)#\", item.replace(\"\\n\", \" \"))\n",
" questions = re.findall(r\"Question:?(.*?) Answer:\", item.replace(\"\\n\", \" \"))\n",
"\n",
" # The last question does not contain an aswer so we drop it every time.\n",
" if len(answers) < len(questions):\n",
" questions.pop(-1)\n",
"\n",
" for (answer, question) in zip(answers, questions):\n",
" if first_sample:\n",
" print(f\"Q: {question}\")\n",
" print(f\"A: {answer}\")\n",
" root = ConversationTreeNode(text=question, role=\"prompter\", children=[], metadata=None)\n",
" child = ConversationTreeNode(text=answer, role=\"assistant\", children=[], metadata=None)\n",
" root.children.append(child)\n",
" conversation_tree = ConversationTree(root=root, metadata={\"dataset\": ds_name})\n",
" conversation_forest.append(conversation_tree)\n",
"\n",
" first_sample = False\n",
"\n",
" conversation_forest_json = [\n",
" json.loads(TreeEncoder().encode(conversation_tree)) for conversation_tree in conversation_forest\n",
" ]\n",
"\n",
" print(json.dumps(conversation_forest_json, indent=4), file=open(f\"./{ds_name}.json\", \"w+\"))\n",
" print(\"\\n\")"
]
},
{
"cell_type": "markdown",
"id": "e4448c9a",
"metadata": {},
"source": [
"We now clone the repository containing the dataset"
]
},
{
"cell_type": "code",
"execution_count": 16,
"id": "06e7719e",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Cloning into 'CodeT'...\r\n",
"remote: Enumerating objects: 144, done.\u001B[K\r\n",
"remote: Counting objects: 100% (16/16), done.\u001B[K\r\n",
"remote: Compressing objects: 100% (16/16), done.\u001B[K\r\n",
"remote: Total 144 (delta 1), reused 0 (delta 0), pack-reused 128\u001B[Kving objects: 8% (12/144), 3.70 MiB | 1.35 MiB/s Receiving objects: 12% (18/144), 5.13 MiB | 1.58 MiB/s Receiving objects: 13% (19/144), 11.36 MiB | 2.39 MiB/s Receiving objects: 13% (20/144), 19.19 MiB | 3.97 MiB/s Receiving objects: 15% (22/144), 19.19 MiB | 3.97 MiB/s Receiving objects: 23% (34/144), 22.15 MiB | 4.37 MiB/s Receiving objects: 27% (39/144), 22.15 MiB | 4.37 MiB/s Receiving objects: 29% (42/144), 25.30 MiB | 4.77 MiB/s Receiving objects: 32% (47/144), 28.71 MiB | 5.22 MiB/s Receiving objects: 41% (60/144), 32.41 MiB | 5.62 MiB/s Receiving objects: 54% (78/144), 32.41 MiB | 5.62 MiB/s Receiving objects: 60% (87/144), 39.34 MiB | 6.20 MiB/s Receiving objects: 61% (88/144), 47.14 MiB | 6.79 MiB/s Receiving objects: 64% (93/144), 51.36 MiB | 7.12 MiB/s Receiving objects: 66% (96/144), 55.58 MiB | 7.40 MiB/s \r\n",
"Receiving objects: 100% (144/144), 56.76 MiB | 4.97 MiB/s, done.\r\n",
"Resolving deltas: 100% (33/33), done.\r\n",
"Checking out files: 100% (64/64), done.\r\n"
]
}
],
"source": [
"!git clone https://github.com/microsoft/CodeT.git"
]
},
{
"cell_type": "code",
"execution_count": 17,
"id": "89a166c9",
"metadata": {},
"outputs": [],
"source": [
"diverse_files = [\n",
" \"CodeT/DIVERSE/data/sqa/split1/test.jsonl\",\n",
" \"CodeT/DIVERSE/data/sqa/split1/train.jsonl\",\n",
" \"CodeT/DIVERSE/data/sqa/split2/test.jsonl\",\n",
" \"CodeT/DIVERSE/data/sqa/split2/train.jsonl\",\n",
" \"CodeT/DIVERSE/data/gsm8k/test.jsonl\",\n",
" \"CodeT/DIVERSE/data/gsm8k/train.jsonl\",\n",
"]"
]
},
{
"cell_type": "code",
"execution_count": 20,
"id": "2da75b14",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"***** diverse_sqa_split1_test ****\n",
"Example of raw dataset\n",
" context \\\n",
"0 Question:\\nIs clerk of Supreme Court of Canada... \n",
"1 Question:\\nIs Saturn named after king of gods ... \n",
"\n",
" samples \\\n",
"0 [Snoopy is a dog.\\nChance is a dog.\\nDogs look... \n",
"1 [Snoopy is a cartoon dog.\\nChance is a cartoon... \n",
"\n",
" metadata \n",
"0 {'type': 'solution', 'question': 'Does Snoopy ... \n",
"1 {'type': 'solution', 'question': 'Does Snoopy ... \n",
"\n",
"Examples from converted dataset\n",
"Q: Is clerk of Supreme Court of Canada safe profession for someone with seismophobia?\n",
"A: The Supreme Court of Canada is in Ottawa, Canada. Ottawa is in Ontario, Canada. Ontario is in the Canadian Shield. The Canadian Shield is a stable tectonic plate. Thus, Ottawa is not prone to earthquakes. Thus, the clerk of the Supreme Court of Canada is a safe profession for someone with seismophobia. So the answer is yes.\n",
"Q: During the Cuban revolution, did the US experience a population boom?\n",
"A: The Cuban revolution was in 1959. The US population in 1959 was about 180 million. The US population in 2010 was about 310 million. Thus, the US population increased by about 130 million. So the answer is yes.\n",
"Q: Can the largest crustacean stretch out completely on a king-sized mattress?\n",
"A: The largest crustacean is the Japanese spider crab. The Japanese spider crab has a leg span of 3.8 meters. A king-sized mattress is 1.9 meters wide. Thus, the Japanese spider crab could not stretch out completely on a king-sized mattress. So the answer is no.\n",
"Q: Could morphine cure HIV?\n",
"A: Morphine is a painkiller. HIV is a virus. Painkillers do not cure viruses. Thus, morphine could not cure HIV. So the answer is no.\n",
"Q: Is Christopher Nolan indebted to Bob Kane?\n",
"A: Bob Kane created Batman. Christopher Nolan directed the Dark Knight trilogy. The Dark Knight trilogy is about Batman. Thus, Christopher Nolan is indebted to Bob Kane. So the answer is yes.\n",
"Q: In baseball, is a \"Homer\" named after the poet Homer who wrote the Odyssey?\n",
"A: A \"Homer\" is a home run. The term \"home run\" was coined by Harry Wright in 1858. Homer lived from about 800 BC to about 700 BC. Thus, the term \"home run\" was not named after the poet Homer. So the answer is no.\n",
"\n",
"\n",
"***** diverse_sqa_split1_train ****\n",
"Example of raw dataset\n",
" context \\\n",
"0 Question:\\nIs eating a Dicopomorpha echmeptery... \n",
"1 Question:\\nWas Alexander the Great baptized?\\n... \n",
"\n",
" samples \\\n",
"0 [New York Public Library is a library.\\nSix Fl... \n",
"1 [The New York Public Library is in New York Ci... \n",
"\n",
" metadata \n",
"0 {'type': 'solution', 'question': 'Could you go... \n",
"1 {'type': 'solution', 'question': 'Could you go... \n",
"\n",
"Examples from converted dataset\n",
"Q: Is eating a Dicopomorpha echmepterygis size Uranium pellet fatal?\n",
"A: The Dicopomorpha echmepterygis is a beetle. The mass of a beetle is about 0.1 g. The mass of a Uranium pellet is about 10 g. Thus, eating a Dicopomorpha echmepterygis is not the same as eating a Uranium pellet. So the answer is no.\n",
"Q: Does Jerry Seinfeld hang out at the Budweiser Party Deck?\n",
"A: Jerry Seinfeld is a comedian. Comedians perform at comedy clubs. The Budweiser Party Deck is not a comedy club. Thus, Jerry Seinfeld does not hang out at the Budweiser Party Deck. So the answer is no.\n",
"Q: Does The Jungle Book contain racist subtext?\n",
"A: The Jungle Book is a book about a boy raised by wolves. The boy is white. The wolves are black. Thus, The Jungle Book contains racist subtext. So the answer is yes.\n",
"Q: Does Dean Cain have less days to birthday than Will Ferrell every 4th of July?\n",
"A: Dean Cain was born on July 31. Will Ferrell was born on July 16. July 4 is the middle of July. Thus, Dean Cain has more days to his birthday than Will Ferrell every 4th of July. So the answer is no.\n",
"Q: Do mail carriers need multiple uniforms?\n",
"A: Mail carriers need to wear uniforms. Uniforms are clothes. People need multiple clothes. Thus, mail carriers need multiple uniforms. So the answer is yes.\n",
"Q: Can Curiosity take samples of rocks from Lacus Temporis?\n",
"A: Curiosity is a rover on Mars. Lacus Temporis is a lake on Mars. Curiosity is not in Lacus Temporis. Thus, Curiosity cannot take samples of rocks from Lacus Temporis. So the answer is no.\n",
"\n",
"\n",
"***** diverse_sqa_split2_test ****\n",
"Example of raw dataset\n",
" context \\\n",
"0 Question:\\nDoes the density of helium cause vo... \n",
"1 Question:\\nIs Mark Cuban able to visit Norther... \n",
"\n",
" samples \\\n",
"0 [The New York Public Library is in New York Ci... \n",
"1 [The distance from New York Public Library to ... \n",
"\n",
" metadata \n",
"0 {'type': 'solution', 'question': 'Could you go... \n",
"1 {'type': 'solution', 'question': 'Could you go... \n",
"\n",
"Examples from converted dataset\n",
"Q: Does the density of helium cause voices to sound deeper?\n",
"A: The density of helium is 0.1785 g/L, which is less than air. The density of air is 1.225 g/L. The density of helium is less than air, so helium causes voices to sound higher. Thus, helium does not cause voices to sound deeper. So the answer is no.\n",
"Q: Would Janet Jackson avoid a dish with ham?\n",
"A: Janet Jackson is a vegetarian. Vegetarians do not eat meat. Ham is a type of meat. Thus, Janet Jackson would avoid a dish with ham. So the answer is yes.\n",
"Q: Do people watching Coen brothers films in Guinea Bissau need subtitles?\n",
"A: The Coen brothers are American. Americans speak English. Guinea Bissau is in Africa. Africans speak Portuguese. Thus, people in Guinea Bissau would need subtitles to watch Coen brothers films. So the answer is yes.\n",
"Q: Could a Gladiator's weapon crush a diamond?\n",
"A: A gladiator's weapon was a sword. The hardness of a diamond is 10. The hardness of a sword is 5. Thus, a gladiator's weapon could not crush a diamond. So the answer is no.\n",
"Q: Can Spartina Patens thrive in the Sahara Desert?\n",
"A: Spartina Patens is a salt marsh grass. The Sahara Desert is a desert. Deserts are dry. Thus, Spartina Patens would not thrive in the Sahara Desert. So the answer is no.\n",
"Q: Were all the materials to make a cannon known during the bronze age?\n",
"A: The bronze age was about 3000 BC to 1200 BC. The bronze age was before the iron age. Thus, iron was not known during the bronze age. Iron is needed to make a cannon. Thus, all the materials to make a cannon were not known during the bronze age. So the answer is no.\n",
"\n",
"\n",
"***** diverse_sqa_split2_train ****\n",
"Example of raw dataset\n",
" context \\\n",
"0 Question:\\nWas the Peak of the Andes hidden fr... \n",
"1 Question:\\nIs Mark Cuban able to visit Norther... \n",
"\n",
" samples \\\n",
"0 [Snoopy is a dog.\\nChance is a dog.\\nChance is... \n",
"1 [Snoopy is a cartoon dog.\\nChance is a cartoon... \n",
"\n",
" metadata \n",
"0 {'type': 'solution', 'question': 'Does Snoopy ... \n",
"1 {'type': 'solution', 'question': 'Does Snoopy ... \n",
"\n",
"Examples from converted dataset\n",
"Q: Was the Peak of the Andes hidden from the view of the Colossus of Rhodes?\n",
"A: The Colossus of Rhodes was a statue of the Greek god Helios. The Colossus of Rhodes was located on the island of Rhodes. The Peak of the Andes is in South America. Thus, the Peak of the Andes was not visible from the Colossus of Rhodes. So the answer is yes.\n",
"Q: Can you swim to Miami from New York?\n",
"A: The distance from New York to Miami is about 1,500 miles. The fastest swimmer can swim about 2 miles per hour. Thus, it would take about 750 hours to swim from New York to Miami. Thus, you could not swim to Miami from New York. So the answer is no.\n",
"Q: Is Freya a combination of Athena and Aphrodite?\n",
"A: Freya is the Norse goddess of love and beauty. Athena is the Greek goddess of wisdom and war. Aphrodite is the Greek goddess of love and beauty. Thus, Freya is a combination of Athena and Aphrodite. So the answer is yes.\n",
"Q: Were the Great Pyramids built by a theocratic government?\n",
"A: The Great Pyramids were built by the Egyptians. The Egyptians were ruled by a pharaoh. The pharaoh was considered a god. Thus, the Great Pyramids were built by a theocratic government. So the answer is yes.\n",
"Q: Was P. G. Wodehouse's favorite book The Hunger Games?\n",
"A: P. G. Wodehouse's favorite book was The Pickwick Papers. The Pickwick Papers was written by Charles Dickens. The Hunger Games was written by Suzanne Collins. Thus, P. G. Wodehouse's favorite book was not The Hunger Games. So the answer is no.\n",
"Q: Can Burundi's communicate with citizens of New Brunswick?\n",
"A: Burundi's speak French. New Brunswick is in Canada. Canada is a bilingual country. Thus, Burundi's can communicate with citizens of New Brunswick. So the answer is yes.\n",
"\n",
"\n",
"***** diverse_gsm8k_test ****\n",
"Example of raw dataset\n",
" context \\\n",
"0 Question:\\nA community is building a metal fen... \n",
"1 Question:\\nThe white rabbit can hop 15 meters ... \n",
"\n",
" samples \\\n",
"0 [She eats 3 and uses 4, so that is 7 eggs.\\n16... \n",
"1 [She eats 3 eggs for breakfast and uses 4 in m... \n",
"\n",
" metadata \n",
"0 {'question': 'Janet’s ducks lay 16 eggs per da... \n",
"1 {'question': 'Janet’s ducks lay 16 eggs per da... \n",
"\n",
"Examples from converted dataset\n",
"Q: A community is building a metal fence. Each fence panel is made of 3 metal sheets, and 2 metal beams. The fence is made of 10 fence panels. If each sheet is made of 10 metal rods and each metal beam is made of 4 metal rods, how many metal rods does the community need for the fence?\n",
"A: In each panel, the metal sheets use 3 metal sheets * 10 metal rods = <<3*10=30>>30 metal rods. In each panel, the metal beams use 2 metal beams * 4 metal rods = <<2*4=8>>8 metal rods. So each panel uses 30 + 8 = <<30+8=38>>38 metal rods. The entire fence therefore needs 38 metal rods * 10 fence panels = <<38*10=380>>380 metal rods. \n",
"Q: John buys 3 dress shirts. They sell for $20 each. He also has to pay 10% tax on everything. How much did he pay in total?\n",
"A: The shirts cost 3*$20=$<<3*20=60>>60 before tax The tax cost $60*.1=$<<60*.1=6>>6 So in total they paid $60+$6=$<<60+6=66>>66 \n",
"Q: Bob gets rent assistance because he's low-income. If he gets a raise of $0.50/hour and works 40 hours a week, how much more will he actually earn a week if his housing benefit is reduced by $60/month?\n",
"A: First find the total increase in Bob's earnings: $0.50/hour * 40 hours/week = $<<0.50*40=20>>20/week Then find the weekly decrease in Bob's housing assistance: $60/month / 4 weeks/month = $<<60/4=15>>15/week Then subtract the lost assistance from the increased wages to find Bob's net increase in money: $20/week - $15/week = $<<20-15=5>>5/week \n",
"Q: Annie plants 3 pots of basil, 9 pots of rosemary, and 6 pots of thyme. Each basil plant has 4 leaves, each rosemary plant has 18 leaves, and each thyme plant has 30 leaves. How many leaves are there total?\n",
"A: First find the total number of basil leaves: 3 pots * 4 leaves/pot = <<3*4=12>>12 leaves Then find the total number of rosemary leaves: 9 pots * 18 leaves/pot = <<9*18=162>>162 leaves Then find the total number of thyme leaves: 6 pots * 30 leaves/pot = <<6*30=180>>180 leaves Then add the number of each type of leaf to find the total number of leaves: 12 leaves + 162 leaves + 180 leaves = <<12+162+180=354>>354 leaves \n",
"Q: There are 7 mL of solution in each of 6 test tubes. Dr. Igor takes all of the solution and then evenly distributes it into 3 beakers. How many mL of solution are in each beaker?\n",
"A: 7 * 6 = <<7*6=42>>42 mL 42/3 = <<42/3=14>>14 mL Each beaker holds 14 mL of solution. \n",
"Q: Janet pays $40/hour for 3 hours per week of clarinet lessons and $28/hour for 5 hours a week of piano lessons. How much more does she spend on piano lessons than clarinet lessons in a year?\n",
"A: First find the total Janet spends on clarinet lessons per week: $40/hour * 3 hours/week = $<<40*3=120>>120/week Then find the total Janet spends on piano lessons per week: $28/hour * 5 hours/week = $<<28*5=140>>140/week Then subtract her weekly clarinet spending from her weekly piano spending to find the weekly difference: $140/week - $120/week = $<<140-120=20>>20/week Then multiply the weekly difference by the number of weeks in a year to find the annual difference: $20/week * 52 weeks/year = $<<20*52=1040>>1040/year \n",
"Q: A normal lemon tree produces 60 lemons per year. Jim has specially engineered lemon trees that produce 50% more lemons per year. He has a grove that is 50 trees by 30 trees. How many lemons does he produce in 5 years?\n",
"A: Each tree produces 60*.5=<<60*.5=30>>30 more lemons than normal So they each produce 60+30=<<60+30=90>>90 lemons He has 50*30=<<50*30=1500>>1500 trees So every year he produces 1500*90=<<1500*90=135000>>135000 lemons That means he produces 135000*5=<<135000*5=675000>>675,000 \n",
"Q: Billy weighs 9 pounds more than Brad. Brad weighs 5 pounds more than Carl. If Carl weighs 145 pounds, how much does Billy weigh, in pounds?\n",
"A: Brad weighs 145+5=<<145+5=150>>150 pounds. Billy weighs 150+9=<<150+9=159>>159 pounds. \n",
"\n",
"\n",
"***** diverse_gsm8k_train ****\n",
"Example of raw dataset\n",
" context \\\n",
"0 Question:\\nA magician has a top hat with 20 re... \n",
"1 Question:\\nA community is building a metal fen... \n",
"\n",
" samples \\\n",
"0 [There are 125 cars in total\\n64% of them are ... \n",
"1 [64% of 125 is <<64/100*125=80>>80.\\n8% of 125... \n",
"\n",
" metadata \n",
"0 {'question': 'Pauline has 125 matchbox cars. T... \n",
"1 {'question': 'Pauline has 125 matchbox cars. T... \n",
"\n",
"Examples from converted dataset\n",
"Q: A magician has a top hat with 20 red marbles and a top hat with 30 blue marbles. If he takes away 3 red marbles and four times as many blue marbles as red marbles (without looking), how many marbles in total does he have left?\n",
"A: He had 20 red marbles and took away 3 leaving 20-3 = <<20-3=17>>17 red marbles He took 4 times as many blue marbles as red marbles which is 4*3 = <<4*3=12>>12 blue marbles He took 12 blue marbles from 30 leaving 30-12 = 18 blue marbles He now has 17+18 = <<17+18=35>>35 marbles left \n",
"Q: Lucas wants to get a dog but his parents think he already has too many pets and won't have enough space. He already has 12 pet beds in his room but manages to fit another 8 pet beds. His parents argue that each pet is going to need 2 beds each to feel comfortable. According to his parent's argument, how many pets does Lucas have enough room for?\n",
"A: Lucas has a total of 12 existing pet beds + 8 new pet beds = <<12+8=20>>20 pet beds. So according to his parents, Lucas has enough room for 20 pet beds / 2 pet beds per pet = <<20/2=10>>10 pets. \n",
"Q: Super Clean Car Wash Company cleans 80 cars per day. They make $5 per car washed. How much money will they make in 5 days?\n",
"A: Each day they will make 80 × $5 = $<<80*5=400>>400. They will make $400 × 5 = $<<400*5=2000>>2000 in 5 days. \n",
"Q: Eighteen hours ago, Beth and I took 100 photographs of our project. Today, Beth and I will take 20% fewer photographs of the same project. If we were to take 300 photographs of the project, how many photographs would we take to reach the target?\n",
"A: If you took 100 photographs of the project 18 hours ago, and today 20% few photographs have been taken, then 20/100*100 = 20 fewer photographs of the project have been taken today. The total number of photographs of the project that have been taken today is 100-20 = <<100-20=80>>80 So far, you've taken 80+100 = <<80+100=180>>180 photographs of the project. Since the target number of photographs is 300, the number of photographs that you need to take to reach the target is 300-180 = <<300-180=120>>120 \n",
"Q: Ruby was going to order pizza for dinner. Her son would only eat pepperoni pizza. Her daughter would only eat sausage. Ruby and her husband wanted black olive and mushroom pizza. To make life easy, Ruby decided to order an entire pizza for each of her children and she would split one with her husband. The pizza restaurant charged $10.00 per pizza and $1.00 per topping. She also needed to add a $5.00 tip. Including tip, how much would the pizza order cost?\n",
"A: Ruby was going to order 1 for her son, 1 for her daughter and 1 to share with her husband. So she needed to order 1+1+1 = <<1+1+1=3>>3 pizzas Each pizza cost $10 and she was ordering 3 so that comes to 10*3 = $<<10*3=30.00>>30.00 She needed to order pepperoni, sausage, black olive and mushroom, which came to 4 toppings at $1.00 each so 4*1 = $<<4*1=4.00>>4.00 extra for toppings The pizzas cost $30 and $4 for the toppings so the total costs of the pizzas came to 30+4 = $<<30+4=34.00>>34.00 She also had to add a $5.00 tip to her current order total of $34.00 so 5+34.00 = $<<5+34=39.00>>39.00 for the total order \n",
"Q: Yves and his siblings ordered pizza and asked to have it cut into 16 slices. During dinner time, they only ate one-fourth of it. The next day, Yves ate one-fourth of the remaining pizza. Then his two siblings ate 2 slices each. How many slices of pizza were left?\n",
"A: During dinner time, Yves and his siblings ate 16/4 = <<16/4=4>>4 slices. So the next day, there were still 16 - 4 = <<16-4=12>>12 slices left. The next day, Yves ate 12/4 = <<12/4=3>>3 slices of pizza. Thus, there were 12 - 3 = <<12-3=9>>9 slices left. Then, his two siblings ate 2 x 2 = <<2*2=4>>4 slices of pizza. Therefore, there were still 9 - 4 = <<9-4=5>>5 slices of pizza left. \n",
"Q: Pria bought a new car that advertised an estimated gas mileage of 35 miles per gallon. The car has a 12-gallon tank. She filled her car full of gas and was able to drive a total of 372 miles. What was the difference, in miles per gallon, between Pria's mileage and the advertised mileage?\n",
"A: Pria's car achieved a rate of 372 miles / 12 gallons = <<372/12=31>>31 miles per gallon. Therefore, it was a difference of 35 - 31 = <<35-31=4>>4 miles per gallon. \n",
"Q: Janet has 24 dresses. Half of them have pockets. Of those, a third have 2 pockets and the rest have 3 pockets. How many total pockets do her dresses have?\n",
"A: She has 24/2=<<24/2=12>>12 dresses with pockets Of those 12/3=4 have 2 pockets So 12-4=<<12-4=8>>8 have three pockets So the dresses with 2 pockets have 2*4=<<2*4=8>>8 pockets The other dresses contribute 8*3=<<8*3=24>>24 pockets So she has a total of 8+24=<<8+24=32>>32 pockets \n",
"\n",
"\n"
]
}
],
"source": [
"for file in diverse_files:\n",
" convert_diverse(file)"
]
},
{
"cell_type": "code",
"execution_count": null,
"outputs": [],
"source": [],
"metadata": {
"collapsed": false
}
}
],
"metadata": {
"kernelspec": {
"name": "conda-env-open-assistant-py",
"language": "python",
"display_name": "Python [conda env:open-assistant]"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.10.0"
},
"vscode": {
"interpreter": {
"hash": "25d5c2324055587ceaeef27650c79ce8358ea61d7689f2e0b8ada5d53f85bce4"
}
}
},
"nbformat": 4,
"nbformat_minor": 5
}
oasst-shared/oasst_shared/exceptions/oasst_api_error.py
+8 -2
View File
@@ -18,6 +18,7 @@ class OasstErrorCode(IntEnum):
DATABASE_URI_NOT_SET = 1
API_CLIENT_NOT_AUTHORIZED = 2
ROOT_TOKEN_NOT_AUTHORIZED = 3
DATABASE_MAX_RETRIES_EXHAUSTED = 4
TOO_MANY_REQUESTS = 429
SERVER_ERROR0 = 500
@@ -31,6 +32,7 @@ class OasstErrorCode(IntEnum):
TASK_INTERACTION_REQUEST_FAILED = 1004
TASK_GENERATION_FAILED = 1005
TASK_REQUESTED_TYPE_NOT_AVAILABLE = 1006
TASK_AVAILABILITY_QUERY_FAILED = 1007
# 2000-3000: prompt_repository
INVALID_FRONTEND_MESSAGE_ID = 2000
@@ -38,7 +40,7 @@ class OasstErrorCode(IntEnum):
RATING_OUT_OF_RANGE = 2002
INVALID_RANKING_VALUE = 2003
INVALID_TASK_TYPE = 2004
USER_NOT_SPECIFIED = 2005
NO_MESSAGE_TREE_FOUND = 2006
NO_REPLIES_FOUND = 2007
INVALID_MESSAGE = 2008
@@ -60,11 +62,15 @@ class OasstErrorCode(IntEnum):
TASK_NOT_COLLECTIVE = 2106
TASK_NOT_ASSIGNED_TO_USER = 2106
TASK_UNEXPECTED_PAYLOAD_TYPE_ = 2107
USER_NOT_FOUND = 2200
# 3000-4000: external resources
HUGGINGFACE_API_ERROR = 3001
# 4000-5000: user
USER_NOT_SPECIFIED = 4000
USER_DISABLED = 4001
USER_NOT_FOUND = 4002
class OasstError(Exception):
"""Base class for Open-Assistant exceptions."""
oasst-shared/oasst_shared/schemas/protocol.py
+1
View File
@@ -392,6 +392,7 @@ class UserScore(BaseModel):
class LeaderboardStats(BaseModel):
time_frame: str
last_updated: datetime
leaderboard: List[UserScore]
oasst-shared/oasst_shared/utils.py
+36 -3
View File
@@ -10,14 +10,47 @@ def utcnow() -> datetime:
return datetime.now(timezone.utc)
def unaware_to_utc(d: datetime | None) -> datetime:
"""Set timezeno to UTC if datetime is unaware (tzinfo == None)."""
if d and d.tzinfo is None:
return d.replace(tzinfo=timezone.utc)
return d
class TimerError(Exception):
"""A custom exception used to report errors in use of Timer class"""
class ScopeTimer:
def __init__(self):
self.start()
def start(self) -> None:
"""Measure new start time"""
self.start_time = time.perf_counter()
def stop(self) -> float:
"""Store and return the elapsed time"""
self.elapsed = time.perf_counter() - self.start_time
return self.elapsed
def __enter__(self):
"""Start a new timer as a context manager"""
self.start()
return self
def __exit__(self, *exc_info):
"""Stop the context manager timer"""
self.stop()
def log_timing(func=None, *, log_kwargs: bool = False, level: int | str = "DEBUG") -> None:
def decorator(func):
@wraps(func)
def wrapped(*args, **kwargs):
start = time.time()
timer = ScopeTimer()
result = func(*args, **kwargs)
end = time.time()
elapsed = end - start
elapsed = timer.stop()
if log_kwargs:
kwargs = ", ".join([f"{k}={v}" for k, v in kwargs.items()])
logger.log(level, f"Function '{func.__name__}({kwargs})' executed in {elapsed:f} s")
scripts/postprocessing/rankings.py
+3 -3
View File
@@ -101,7 +101,7 @@ def ranked_pairs(ranks: List[List[int]]):
# order by strength of victory (using tideman's original method, don't think it would make a difference for us)
sorted_majorities = []
for i in range(len(ranks[0])):
for j in range(len(ranks[i])):
for j in range(len(ranks[0])):
if tallies[i, j] > 0:
sorted_majorities.append((i, j, tallies[i, j]))
# we don't explicitly deal with tied majorities here
@@ -132,8 +132,8 @@ if __name__ == "__main__":
[("w", "x", "z", "y") for _ in range(1)]
+ [("w", "y", "x", "z") for _ in range(2)]
# + [("x","y","z","w") for _ in range(4)]
+ [("x", "z", "w", "y") for _ in range(5)]
+ [("y", "w", "x", "z") for _ in range(1)]
# + [("x", "z", "w", "y") for _ in range(5)]
# + [("y", "w", "x", "z") for _ in range(1)]
# [("y","z","w","x") for _ in range(1000)]
)
rp = ranked_pairs(ranks)
text-frontend/auto_main.py
+250
View File
@@ -0,0 +1,250 @@
"""Simple REPL frontend."""
import http
import random
import requests
import typer
app = typer.Typer()
# debug constants
USER = {"id": "1234", "display_name": "John Doe", "auth_method": "local"}
def _random_message_id():
return str(random.randint(1000, 9999))
def _render_message(message: dict) -> str:
"""Render a message to the user."""
if message["is_assistant"]:
return f"Assistant: {message['text']}"
return f"Prompter: {message['text']}"
@app.command()
def main(backend_url: str = "http://127.0.0.1:8080", api_key: str = "1234"):
"""automates tasks"""
def _post(path: str, json: dict) -> dict:
response = requests.post(f"{backend_url}{path}", json=json, headers={"X-API-Key": api_key})
response.raise_for_status()
if response.status_code == http.HTTPStatus.NO_CONTENT:
return None
return response.json()
def gen_random_text():
return " ".join([random.choice(["hello", "world", "foo", "bar"]) for _ in range(10)])
def gen_random_ranking(messages):
"""rank messages randomly and return list of indexes in order of rank randomly"""
print("Ranking")
print(messages)
print(len(messages))
ranks = [i for i in range(len(messages))]
shuffled = random.shuffle(ranks)
print(ranks)
print(shuffled)
return ranks
tasks = [_post("/api/v1/tasks/", {"type": "random", "user": USER})]
q = 0
while tasks:
task = tasks.pop(0)
print(task)
match (task["type"]):
case "initial_prompt":
typer.echo("Please provide an initial prompt to the assistant.")
if task["hint"]:
typer.echo(f"Hint: {task['hint']}")
# acknowledge task
message_id = _random_message_id()
_post(f"/api/v1/tasks/{task['id']}/ack", {"message_id": message_id})
prompt = gen_random_text()
user_message_id = _random_message_id()
# send interaction
new_task = _post(
"/api/v1/tasks/interaction",
{
"type": "text_reply_to_message",
"message_id": message_id,
"task_id": task["id"],
"user_message_id": user_message_id,
"text": prompt,
"user": USER,
},
)
tasks.append(new_task)
case "label_initial_prompt":
typer.echo("Label the following prompt:")
typer.echo(task["prompt"])
# acknowledge task
message_id = _random_message_id()
_post(f"/api/v1/tasks/{task['id']}/ack", {"message_id": message_id})
valid_labels = task["valid_labels"]
labels_dict = None
if task["mode"] == "simple" and len(valid_labels) == 1:
answer = random.choice([True, False])
labels_dict = {valid_labels[0]: 1 if answer else 0}
else:
while labels_dict is None:
labels = random.sample(valid_labels, random.randint(1, len(valid_labels)))
if all([label in valid_labels for label in labels]):
labels_dict = {label: "1" if label in labels else "0" for label in valid_labels}
else:
invalid_labels = [label for label in labels if label not in valid_labels]
typer.echo(f"Invalid labels: {', '.join(invalid_labels)}. Valid: {', '.join(valid_labels)}")
# send labels
new_task = _post(
"/api/v1/tasks/interaction",
{
"type": "text_labels",
"message_id": task["message_id"],
"task_id": task["id"],
"text": task["prompt"],
"labels": labels_dict,
"user": USER,
},
)
tasks.append(new_task)
case "prompter_reply":
# acknowledge task
message_id = _random_message_id()
user_message_id = _random_message_id()
_post(f"/api/v1/tasks/{task['id']}/ack", {"message_id": message_id})
# send interaction
new_task = _post(
"/api/v1/tasks/interaction",
{
"type": "text_reply_to_message",
"message_id": message_id,
"task_id": task["id"],
"user_message_id": user_message_id,
"text": gen_random_text(),
"user": USER,
},
)
tasks.append(new_task)
case "assistant_reply":
# acknowledge task
message_id = _random_message_id()
user_message_id = _random_message_id()
_post(f"/api/v1/tasks/{task['id']}/ack", {"message_id": message_id})
# send interaction
new_task = _post(
"/api/v1/tasks/interaction",
{
"type": "text_reply_to_message",
"message_id": message_id,
"task_id": task["id"],
"user_message_id": user_message_id,
"text": gen_random_text(),
"user": USER,
},
)
tasks.append(new_task)
case "rank_prompter_replies" | "rank_assistant_replies":
# acknowledge task
message_id = _random_message_id()
user_message_id = _random_message_id()
_post(f"/api/v1/tasks/{task['id']}/ack", {"message_id": message_id})
# send interaction
ranking = gen_random_ranking(task["replies"])
print(ranking)
new_task = _post(
"/api/v1/tasks/interaction",
{
"type": "message_ranking",
"message_id": message_id,
"task_id": task["id"],
"ranking": ranking,
"user": USER,
},
)
tasks.append(new_task)
case "rank_initial_prompts":
# acknowledge task
message_id = _random_message_id()
user_message_id = _random_message_id()
_post(f"/api/v1/tasks/{task['id']}/ack", {"message_id": message_id})
# send interaction
ranking = gen_random_ranking(task["prompots"])
new_task = _post(
"/api/v1/tasks/interaction",
{
"type": "message_ranking",
"message_id": message_id,
"ranking": ranking,
"user": USER,
},
)
tasks.append(new_task)
case "label_prompter_reply" | "label_assistant_reply":
# acknowledge task
typer.echo("Here is the conversation so far:")
for message in task["conversation"]["messages"]:
typer.echo(_render_message(message))
typer.echo("Label the following reply:")
typer.echo(task["reply"])
message_id = _random_message_id()
user_message_id = _random_message_id()
_post(f"/api/v1/tasks/{task['id']}/ack", {"message_id": message_id})
valid_labels = task["valid_labels"]
labels_dict = None
if task["mode"] == "simple" and len(valid_labels) == 1:
answer = random.choice([True, False])
labels_dict = {valid_labels[0]: 1 if answer else 0}
else:
while labels_dict is None:
labels = random.sample(valid_labels, random.randint(1, len(valid_labels)))
if all([label in valid_labels for label in labels]):
labels_dict = {label: "1" if label in labels else "0" for label in valid_labels}
else:
invalid_labels = [label for label in labels if label not in valid_labels]
typer.echo(f"Invalid labels: {', '.join(invalid_labels)}. Valid: {', '.join(valid_labels)}")
# send interaction
new_task = _post(
"/api/v1/tasks/interaction",
{
"type": "text_labels",
"message_id": task["message_id"],
"task_id": task["id"],
"text": task["reply"],
"labels": labels_dict,
"user": USER,
},
)
tasks.append(new_task)
case "task_done":
typer.echo("Task done!")
# rerun with new task slected from above cases
# add a new task
q += 1
if q == 10:
typer.echo("Task done!")
break
tasks = [_post("/api/v1/tasks/", {"type": "random", "user": USER})]
#
case _:
typer.echo(f"Unknown task type {task['type']}")
# rerun with new task slected from above cases
if __name__ == "__main__":
app()
website/cypress/e2e/auth/signin.cy.ts
-7
View File
@@ -27,13 +27,6 @@ describe("signin flow", () => {
});
});
});
it("shows the logged in users email address if logged in with email", () => {
const emailAddress = "user@example.com";
cy.signInWithEmail(emailAddress);
// The user will only see the email address if the window is wide enough, not technically required as even when hidden this will find it in the page.
cy.viewport(1920, 1000);
cy.contains('[data-cy="username"]', emailAddress);
});
});
export {};
website/next-i18next.config.js
+6
View File
@@ -0,0 +1,6 @@
module.exports = {
i18n: {
defaultLocale: "en",
locales: ["en"],
},
};
website/next.config.js
+3
View File
@@ -1,4 +1,6 @@
/** @type {import('next').NextConfig} */
const { i18n } = require("./next-i18next.config");
const nextConfig = {
output: "standalone",
reactStrictMode: true,
@@ -16,6 +18,7 @@ const nextConfig = {
*/
// scrollRestoration: true,
},
i18n,
};
module.exports = nextConfig;
website/package-lock.json
Generated
+214 -108
View File
@@ -29,22 +29,25 @@
"eslint-config-next": "13.0.6",
"eslint-plugin-simple-import-sort": "^8.0.0",
"focus-visible": "^5.2.0",
"formik": "^2.2.9",
"framer-motion": "^6.5.1",
"install": "^0.13.0",
"next": "13.0.6",
"next-auth": "^4.18.6",
"next-i18next": "^13.0.3",
"nodemailer": "^6.8.0",
"npm": "^9.2.0",
"postcss-focus-visible": "^7.1.0",
"react": "18.2.0",
"react-dom": "18.2.0",
"react-feature-flags": "^1.0.0",
"react-hook-form": "^7.42.1",
"react-i18next": "^12.1.4",
"react-icons": "^4.7.1",
"react-table": "^7.8.0",
"sharp": "^0.31.3",
"swr": "^2.0.0",
"tailwindcss": "^3.2.4",
"unique-username-generator": "^1.1.3",
"use-debounce": "^9.0.2"
},
"devDependencies": {
@@ -12653,6 +12656,15 @@
"@types/unist": "*"
}
},
"node_modules/@types/hoist-non-react-statics": {
"version": "3.3.1",
"resolved": "https://registry.npmjs.org/@types/hoist-non-react-statics/-/hoist-non-react-statics-3.3.1.tgz",
"integrity": "sha512-iMIqiko6ooLrTh1joXodJK5X9xeEALT1kM5G3ZLhD3hszxBdIEd5C75U834D9mLcINgD4OyZf5uQXjkuYydWvA==",
"dependencies": {
"@types/react": "*",
"hoist-non-react-statics": "^3.3.0"
}
},
"node_modules/@types/html-minifier-terser": {
"version": "6.1.0",
"resolved": "https://registry.npmjs.org/@types/html-minifier-terser/-/html-minifier-terser-6.1.0.tgz",
@@ -12859,8 +12871,7 @@
"node_modules/@types/prop-types": {
"version": "15.7.5",
"resolved": "https://registry.npmjs.org/@types/prop-types/-/prop-types-15.7.5.tgz",
"integrity": "sha512-JCB8C6SnDoQf0cNycqd/35A7MjcnK+ZTqE7judS6o7utxUCg6imJg3QK2qzHKszlTjcj2cn+NwMB2i96ubpj7w==",
"devOptional": true
"integrity": "sha512-JCB8C6SnDoQf0cNycqd/35A7MjcnK+ZTqE7judS6o7utxUCg6imJg3QK2qzHKszlTjcj2cn+NwMB2i96ubpj7w=="
},
"node_modules/@types/qs": {
"version": "6.9.7",
@@ -12872,7 +12883,6 @@
"version": "18.0.26",
"resolved": "https://registry.npmjs.org/@types/react/-/react-18.0.26.tgz",
"integrity": "sha512-hCR3PJQsAIXyxhTNSiDFY//LhnMZWpNNr5etoCqx/iUfGc5gXWtQR2Phl908jVR6uPXacojQWTg4qRpkxTuGug==",
"devOptional": true,
"dependencies": {
"@types/prop-types": "*",
"@types/scheduler": "*",
@@ -12891,8 +12901,7 @@
"node_modules/@types/scheduler": {
"version": "0.16.2",
"resolved": "https://registry.npmjs.org/@types/scheduler/-/scheduler-0.16.2.tgz",
"integrity": "sha512-hppQEBDmlwhFAXKJX2KnWLYu5yMfi91yazPb2l+lbJiwW+wdo1gNeRA+3RgNSO39WYX2euey41KEwnqesU2Jew==",
"devOptional": true
"integrity": "sha512-hppQEBDmlwhFAXKJX2KnWLYu5yMfi91yazPb2l+lbJiwW+wdo1gNeRA+3RgNSO39WYX2euey41KEwnqesU2Jew=="
},
"node_modules/@types/semver": {
"version": "7.3.13",
@@ -16550,7 +16559,6 @@
"version": "3.27.1",
"resolved": "https://registry.npmjs.org/core-js/-/core-js-3.27.1.tgz",
"integrity": "sha512-GutwJLBChfGCpwwhbYoqfv03LAfmiz7e7D/BNxzeMxwQf10GRSzqiOjx7AmtEk+heiD/JWmBuyBPgFtx0Sg1ww==",
"dev": true,
"hasInstallScript": true,
"funding": {
"type": "opencollective",
@@ -20304,47 +20312,6 @@
"node": ">= 6"
}
},
"node_modules/formik": {
"version": "2.2.9",
"resolved": "https://registry.npmjs.org/formik/-/formik-2.2.9.tgz",
"integrity": "sha512-LQLcISMmf1r5at4/gyJigGn0gOwFbeEAlji+N9InZF6LIMXnFNkO42sCI8Jt84YZggpD4cPWObAZaxpEFtSzNA==",
"funding": [
{
"type": "individual",
"url": "https://opencollective.com/formik"
}
],
"dependencies": {
"deepmerge": "^2.1.1",
"hoist-non-react-statics": "^3.3.0",
"lodash": "^4.17.21",
"lodash-es": "^4.17.21",
"react-fast-compare": "^2.0.1",
"tiny-warning": "^1.0.2",
"tslib": "^1.10.0"
},
"peerDependencies": {
"react": ">=16.8.0"
}
},
"node_modules/formik/node_modules/deepmerge": {
"version": "2.2.1",
"resolved": "https://registry.npmjs.org/deepmerge/-/deepmerge-2.2.1.tgz",
"integrity": "sha512-R9hc1Xa/NOBi9WRVUWg19rl1UB7Tt4kuPd+thNJgFZoxXsTz7ncaPaeIm+40oSGuP33DfMb4sZt1QIGiJzC4EA==",
"engines": {
"node": ">=0.10.0"
}
},
"node_modules/formik/node_modules/react-fast-compare": {
"version": "2.0.4",
"resolved": "https://registry.npmjs.org/react-fast-compare/-/react-fast-compare-2.0.4.tgz",
"integrity": "sha512-suNP+J1VU1MWFKcyt7RtjiSWUjvidmQSlqu+eHslq+342xCbGTYmC0mEhPCOHxlW0CywylOC1u2DFAT+bv4dBw=="
},
"node_modules/formik/node_modules/tslib": {
"version": "1.14.1",
"resolved": "https://registry.npmjs.org/tslib/-/tslib-1.14.1.tgz",
"integrity": "sha512-Xni35NKzjgMrwevysHTCArtLDpPvye8zV/0E4EyYn43P7/7qvQwPh9BGkHewbMulVntbigmcT7rdX3BNo9wRJg=="
},
"node_modules/forwarded": {
"version": "0.2.0",
"resolved": "https://registry.npmjs.org/forwarded/-/forwarded-0.2.0.tgz",
@@ -21345,6 +21312,14 @@
"node": ">= 12"
}
},
"node_modules/html-parse-stringify": {
"version": "3.0.1",
"resolved": "https://registry.npmjs.org/html-parse-stringify/-/html-parse-stringify-3.0.1.tgz",
"integrity": "sha512-KknJ50kTInJ7qIScF3jeaFRpMpE8/lfiTdzf/twXyPBLAGrLRTmkz3AdTnKeh40X8k9L2fdYwEp/42WGXIRGcg==",
"dependencies": {
"void-elements": "3.1.0"
}
},
"node_modules/html-tags": {
"version": "3.2.0",
"resolved": "https://registry.npmjs.org/html-tags/-/html-tags-3.2.0.tgz",
@@ -21487,6 +21462,34 @@
"integrity": "sha512-iimHkHPfIAQ8zCDQLgn08pRqSVioyWvnGfaQ8gond2wf7Jq2jJ+24ykmnRyiz3fIldcn4oUuQXpjqKLhSVR7lw==",
"dev": true
},
"node_modules/i18next": {
"version": "22.4.9",
"resolved": "https://registry.npmjs.org/i18next/-/i18next-22.4.9.tgz",
"integrity": "sha512-8gWMmUz460KJDQp/ob3MNUX84cVuDRY9PLFPnV8d+Qezz/6dkjxwOaH70xjrCNDO+JrUL25iXfAIN9wUkInNZw==",
"funding": [
{
"type": "individual",
"url": "https://locize.com"
},
{
"type": "individual",
"url": "https://locize.com/i18next.html"
},
{
"type": "individual",
"url": "https://www.i18next.com/how-to/faq#i18next-is-awesome.-how-can-i-support-the-project"
}
],
"peer": true,
"dependencies": {
"@babel/runtime": "^7.20.6"
}
},
"node_modules/i18next-fs-backend": {
"version": "2.1.1",
"resolved": "https://registry.npmjs.org/i18next-fs-backend/-/i18next-fs-backend-2.1.1.tgz",
"integrity": "sha512-FTnj+UmNgT3YRml5ruRv0jMZDG7odOL/OP5PF5mOqvXud2vHrPOOs68Zdk6iqzL47cnnM0ZVkK2BAvpFeDJToA=="
},
"node_modules/iconv-lite": {
"version": "0.4.24",
"resolved": "https://registry.npmjs.org/iconv-lite/-/iconv-lite-0.4.24.tgz",
@@ -26443,12 +26446,8 @@
"node_modules/lodash": {
"version": "4.17.21",
"resolved": "https://registry.npmjs.org/lodash/-/lodash-4.17.21.tgz",
"integrity": "sha512-v2kDEe57lecTulaDIuNTPy3Ry4gLGJ6Z1O3vE1krgXZNrsQ+LFTGHVxVjcXPs17LhbZVGedAJv8XZ1tvj5FvSg=="
},
"node_modules/lodash-es": {
"version": "4.17.21",
"resolved": "https://registry.npmjs.org/lodash-es/-/lodash-es-4.17.21.tgz",
"integrity": "sha512-mKnC+QJ9pWVzv+C4/U3rRsHapFfHvQFoFB92e52xeyGMcX6/OlIl78je1u8vePzYZSkkogMPJ2yjxxsb89cxyw=="
"integrity": "sha512-v2kDEe57lecTulaDIuNTPy3Ry4gLGJ6Z1O3vE1krgXZNrsQ+LFTGHVxVjcXPs17LhbZVGedAJv8XZ1tvj5FvSg==",
"dev": true
},
"node_modules/lodash.debounce": {
"version": "4.0.8",
@@ -27568,6 +27567,45 @@
}
}
},
"node_modules/next-i18next": {
"version": "13.0.3",
"resolved": "https://registry.npmjs.org/next-i18next/-/next-i18next-13.0.3.tgz",
"integrity": "sha512-7AA8J6WbkxRBtSf1+97LSAE7btxWZHsBIJEJ3FuTSBgYtpRiO5NGjcb8XbNAlz6yGU0TtS+yZE+/Wu83KhIT1Q==",
"funding": [
{
"type": "individual",
"url": "https://locize.com/i18next.html"
},
{
"type": "individual",
"url": "https://www.i18next.com/how-to/faq#i18next-is-awesome.-how-can-i-support-the-project"
},
{
"type": "individual",
"url": "https://locize.com"
},
{
"type": "individual",
"url": "https://github.com/belgattitude"
}
],
"dependencies": {
"@babel/runtime": "^7.20.6",
"@types/hoist-non-react-statics": "^3.3.1",
"core-js": "^3",
"hoist-non-react-statics": "^3.3.2",
"i18next-fs-backend": "^2.1.0"
},
"engines": {
"node": ">=14"
},
"peerDependencies": {
"i18next": "^22.0.6",
"next": ">= 12.0.0",
"react": ">= 17.0.2",
"react-i18next": "^12.1.1"
}
},
"node_modules/next/node_modules/postcss": {
"version": "8.4.14",
"resolved": "https://registry.npmjs.org/postcss/-/postcss-8.4.14.tgz",
@@ -32527,6 +32565,42 @@
"node": ">=10"
}
},
"node_modules/react-hook-form": {
"version": "7.42.1",
"resolved": "https://registry.npmjs.org/react-hook-form/-/react-hook-form-7.42.1.tgz",
"integrity": "sha512-2UIGqwMZksd5HS55crTT1ATLTr0rAI4jS7yVuqTaoRVDhY2Qc4IyjskCmpnmdYqUNOYFy04vW253tb2JRVh+IQ==",
"engines": {
"node": ">=12.22.0"
},
"funding": {
"type": "opencollective",
"url": "https://opencollective.com/react-hook-form"
},
"peerDependencies": {
"react": "^16.8.0 || ^17 || ^18"
}
},
"node_modules/react-i18next": {
"version": "12.1.4",
"resolved": "https://registry.npmjs.org/react-i18next/-/react-i18next-12.1.4.tgz",
"integrity": "sha512-XQND7jYtgM7ht5PH3yIZljCRpAMTlH/zmngM9ZjToqa+0BR6xuu8c7QF0WIIOEjcMTB2S3iOfpN/xG/ZrAnO6g==",
"dependencies": {
"@babel/runtime": "^7.20.6",
"html-parse-stringify": "^3.0.1"
},
"peerDependencies": {
"i18next": ">= 19.0.0",
"react": ">= 16.8.0"
},
"peerDependenciesMeta": {
"react-dom": {
"optional": true
},
"react-native": {
"optional": true
}
}
},
"node_modules/react-icons": {
"version": "4.7.1",
"resolved": "https://registry.npmjs.org/react-icons/-/react-icons-4.7.1.tgz",
@@ -35486,11 +35560,6 @@
"resolved": "https://registry.npmjs.org/tiny-invariant/-/tiny-invariant-1.3.1.tgz",
"integrity": "sha512-AD5ih2NlSssTCwsMznbvwMZpJ1cbhkGd2uueNxzv2jDlEeZdU04JQfRnggJQ8DrcVBGjAsCKwFBbDlVNtEMlzw=="
},
"node_modules/tiny-warning": {
"version": "1.0.3",
"resolved": "https://registry.npmjs.org/tiny-warning/-/tiny-warning-1.0.3.tgz",
"integrity": "sha512-lBN9zLN/oAf68o3zNXYrdCt1kP8WsiGW8Oo2ka41b2IM5JL/S1CTyX1rW0mb/zSuJun0ZUrDxx4sqvYS2FWzPA=="
},
"node_modules/tmp": {
"version": "0.2.1",
"resolved": "https://registry.npmjs.org/tmp/-/tmp-0.2.1.tgz",
@@ -36022,6 +36091,11 @@
"imurmurhash": "^0.1.4"
}
},
"node_modules/unique-username-generator": {
"version": "1.1.3",
"resolved": "https://registry.npmjs.org/unique-username-generator/-/unique-username-generator-1.1.3.tgz",
"integrity": "sha512-TB6YdqPMKMpTSgxAzjZkKWtmpZPHvARoWreCKBpc1UrLFz/0C6Q96/qdjpLr9OXPCHk16sD1LHjTr3JDj7q2JA=="
},
"node_modules/unist-builder": {
"version": "2.0.3",
"resolved": "https://registry.npmjs.org/unist-builder/-/unist-builder-2.0.3.tgz",
@@ -36542,6 +36616,14 @@
"integrity": "sha512-2ham8XPWTONajOR0ohOKOHXkm3+gaBmGut3SRuu75xLd/RRaY6vqgh8NBYYk7+RW3u5AtzPQZG8F10LHkl0lAQ==",
"dev": true
},
"node_modules/void-elements": {
"version": "3.1.0",
"resolved": "https://registry.npmjs.org/void-elements/-/void-elements-3.1.0.tgz",
"integrity": "sha512-Dhxzh5HZuiHQhbvTW9AMetFfBHDMYpo23Uo9btPXgdYP+3T5S+p+jgNy7spra+veYhBP2dCSgxR/i2Y02h5/6w==",
"engines": {
"node": ">=0.10.0"
}
},
"node_modules/w3c-xmlserializer": {
"version": "4.0.0",
"resolved": "https://registry.npmjs.org/w3c-xmlserializer/-/w3c-xmlserializer-4.0.0.tgz",
@@ -46896,6 +46978,15 @@
"@types/unist": "*"
}
},
"@types/hoist-non-react-statics": {
"version": "3.3.1",
"resolved": "https://registry.npmjs.org/@types/hoist-non-react-statics/-/hoist-non-react-statics-3.3.1.tgz",
"integrity": "sha512-iMIqiko6ooLrTh1joXodJK5X9xeEALT1kM5G3ZLhD3hszxBdIEd5C75U834D9mLcINgD4OyZf5uQXjkuYydWvA==",
"requires": {
"@types/react": "*",
"hoist-non-react-statics": "^3.3.0"
}
},
"@types/html-minifier-terser": {
"version": "6.1.0",
"resolved": "https://registry.npmjs.org/@types/html-minifier-terser/-/html-minifier-terser-6.1.0.tgz",
@@ -47088,8 +47179,7 @@
"@types/prop-types": {
"version": "15.7.5",
"resolved": "https://registry.npmjs.org/@types/prop-types/-/prop-types-15.7.5.tgz",
"integrity": "sha512-JCB8C6SnDoQf0cNycqd/35A7MjcnK+ZTqE7judS6o7utxUCg6imJg3QK2qzHKszlTjcj2cn+NwMB2i96ubpj7w==",
"devOptional": true
"integrity": "sha512-JCB8C6SnDoQf0cNycqd/35A7MjcnK+ZTqE7judS6o7utxUCg6imJg3QK2qzHKszlTjcj2cn+NwMB2i96ubpj7w=="
},
"@types/qs": {
"version": "6.9.7",
@@ -47101,7 +47191,6 @@
"version": "18.0.26",
"resolved": "https://registry.npmjs.org/@types/react/-/react-18.0.26.tgz",
"integrity": "sha512-hCR3PJQsAIXyxhTNSiDFY//LhnMZWpNNr5etoCqx/iUfGc5gXWtQR2Phl908jVR6uPXacojQWTg4qRpkxTuGug==",
"devOptional": true,
"requires": {
"@types/prop-types": "*",
"@types/scheduler": "*",
@@ -47120,8 +47209,7 @@
"@types/scheduler": {
"version": "0.16.2",
"resolved": "https://registry.npmjs.org/@types/scheduler/-/scheduler-0.16.2.tgz",
"integrity": "sha512-hppQEBDmlwhFAXKJX2KnWLYu5yMfi91yazPb2l+lbJiwW+wdo1gNeRA+3RgNSO39WYX2euey41KEwnqesU2Jew==",
"devOptional": true
"integrity": "sha512-hppQEBDmlwhFAXKJX2KnWLYu5yMfi91yazPb2l+lbJiwW+wdo1gNeRA+3RgNSO39WYX2euey41KEwnqesU2Jew=="
},
"@types/semver": {
"version": "7.3.13",
@@ -49994,8 +50082,7 @@
"core-js": {
"version": "3.27.1",
"resolved": "https://registry.npmjs.org/core-js/-/core-js-3.27.1.tgz",
"integrity": "sha512-GutwJLBChfGCpwwhbYoqfv03LAfmiz7e7D/BNxzeMxwQf10GRSzqiOjx7AmtEk+heiD/JWmBuyBPgFtx0Sg1ww==",
"dev": true
"integrity": "sha512-GutwJLBChfGCpwwhbYoqfv03LAfmiz7e7D/BNxzeMxwQf10GRSzqiOjx7AmtEk+heiD/JWmBuyBPgFtx0Sg1ww=="
},
"core-js-compat": {
"version": "3.27.1",
@@ -52929,37 +53016,6 @@
"mime-types": "^2.1.12"
}
},
"formik": {
"version": "2.2.9",
"resolved": "https://registry.npmjs.org/formik/-/formik-2.2.9.tgz",
"integrity": "sha512-LQLcISMmf1r5at4/gyJigGn0gOwFbeEAlji+N9InZF6LIMXnFNkO42sCI8Jt84YZggpD4cPWObAZaxpEFtSzNA==",
"requires": {
"deepmerge": "^2.1.1",
"hoist-non-react-statics": "^3.3.0",
"lodash": "^4.17.21",
"lodash-es": "^4.17.21",
"react-fast-compare": "^2.0.1",
"tiny-warning": "^1.0.2",
"tslib": "^1.10.0"
},
"dependencies": {
"deepmerge": {
"version": "2.2.1",
"resolved": "https://registry.npmjs.org/deepmerge/-/deepmerge-2.2.1.tgz",
"integrity": "sha512-R9hc1Xa/NOBi9WRVUWg19rl1UB7Tt4kuPd+thNJgFZoxXsTz7ncaPaeIm+40oSGuP33DfMb4sZt1QIGiJzC4EA=="
},
"react-fast-compare": {
"version": "2.0.4",
"resolved": "https://registry.npmjs.org/react-fast-compare/-/react-fast-compare-2.0.4.tgz",
"integrity": "sha512-suNP+J1VU1MWFKcyt7RtjiSWUjvidmQSlqu+eHslq+342xCbGTYmC0mEhPCOHxlW0CywylOC1u2DFAT+bv4dBw=="
},
"tslib": {
"version": "1.14.1",
"resolved": "https://registry.npmjs.org/tslib/-/tslib-1.14.1.tgz",
"integrity": "sha512-Xni35NKzjgMrwevysHTCArtLDpPvye8zV/0E4EyYn43P7/7qvQwPh9BGkHewbMulVntbigmcT7rdX3BNo9wRJg=="
}
}
},
"forwarded": {
"version": "0.2.0",
"resolved": "https://registry.npmjs.org/forwarded/-/forwarded-0.2.0.tgz",
@@ -53746,6 +53802,14 @@
}
}
},
"html-parse-stringify": {
"version": "3.0.1",
"resolved": "https://registry.npmjs.org/html-parse-stringify/-/html-parse-stringify-3.0.1.tgz",
"integrity": "sha512-KknJ50kTInJ7qIScF3jeaFRpMpE8/lfiTdzf/twXyPBLAGrLRTmkz3AdTnKeh40X8k9L2fdYwEp/42WGXIRGcg==",
"requires": {
"void-elements": "3.1.0"
}
},
"html-tags": {
"version": "3.2.0",
"resolved": "https://registry.npmjs.org/html-tags/-/html-tags-3.2.0.tgz",
@@ -53846,6 +53910,20 @@
"integrity": "sha512-iimHkHPfIAQ8zCDQLgn08pRqSVioyWvnGfaQ8gond2wf7Jq2jJ+24ykmnRyiz3fIldcn4oUuQXpjqKLhSVR7lw==",
"dev": true
},
"i18next": {
"version": "22.4.9",
"resolved": "https://registry.npmjs.org/i18next/-/i18next-22.4.9.tgz",
"integrity": "sha512-8gWMmUz460KJDQp/ob3MNUX84cVuDRY9PLFPnV8d+Qezz/6dkjxwOaH70xjrCNDO+JrUL25iXfAIN9wUkInNZw==",
"peer": true,
"requires": {
"@babel/runtime": "^7.20.6"
}
},
"i18next-fs-backend": {
"version": "2.1.1",
"resolved": "https://registry.npmjs.org/i18next-fs-backend/-/i18next-fs-backend-2.1.1.tgz",
"integrity": "sha512-FTnj+UmNgT3YRml5ruRv0jMZDG7odOL/OP5PF5mOqvXud2vHrPOOs68Zdk6iqzL47cnnM0ZVkK2BAvpFeDJToA=="
},
"iconv-lite": {
"version": "0.4.24",
"resolved": "https://registry.npmjs.org/iconv-lite/-/iconv-lite-0.4.24.tgz",
@@ -57574,12 +57652,8 @@
"lodash": {
"version": "4.17.21",
"resolved": "https://registry.npmjs.org/lodash/-/lodash-4.17.21.tgz",
"integrity": "sha512-v2kDEe57lecTulaDIuNTPy3Ry4gLGJ6Z1O3vE1krgXZNrsQ+LFTGHVxVjcXPs17LhbZVGedAJv8XZ1tvj5FvSg=="
},
"lodash-es": {
"version": "4.17.21",
"resolved": "https://registry.npmjs.org/lodash-es/-/lodash-es-4.17.21.tgz",
"integrity": "sha512-mKnC+QJ9pWVzv+C4/U3rRsHapFfHvQFoFB92e52xeyGMcX6/OlIl78je1u8vePzYZSkkogMPJ2yjxxsb89cxyw=="
"integrity": "sha512-v2kDEe57lecTulaDIuNTPy3Ry4gLGJ6Z1O3vE1krgXZNrsQ+LFTGHVxVjcXPs17LhbZVGedAJv8XZ1tvj5FvSg==",
"dev": true
},
"lodash.debounce": {
"version": "4.0.8",
@@ -58467,6 +58541,18 @@
"uuid": "^8.3.2"
}
},
"next-i18next": {
"version": "13.0.3",
"resolved": "https://registry.npmjs.org/next-i18next/-/next-i18next-13.0.3.tgz",
"integrity": "sha512-7AA8J6WbkxRBtSf1+97LSAE7btxWZHsBIJEJ3FuTSBgYtpRiO5NGjcb8XbNAlz6yGU0TtS+yZE+/Wu83KhIT1Q==",
"requires": {
"@babel/runtime": "^7.20.6",
"@types/hoist-non-react-statics": "^3.3.1",
"core-js": "^3",
"hoist-non-react-statics": "^3.3.2",
"i18next-fs-backend": "^2.1.0"
}
},
"nice-try": {
"version": "1.0.5",
"resolved": "https://registry.npmjs.org/nice-try/-/nice-try-1.0.5.tgz",
@@ -61952,6 +62038,21 @@
}
}
},
"react-hook-form": {
"version": "7.42.1",
"resolved": "https://registry.npmjs.org/react-hook-form/-/react-hook-form-7.42.1.tgz",
"integrity": "sha512-2UIGqwMZksd5HS55crTT1ATLTr0rAI4jS7yVuqTaoRVDhY2Qc4IyjskCmpnmdYqUNOYFy04vW253tb2JRVh+IQ==",
"requires": {}
},
"react-i18next": {
"version": "12.1.4",
"resolved": "https://registry.npmjs.org/react-i18next/-/react-i18next-12.1.4.tgz",
"integrity": "sha512-XQND7jYtgM7ht5PH3yIZljCRpAMTlH/zmngM9ZjToqa+0BR6xuu8c7QF0WIIOEjcMTB2S3iOfpN/xG/ZrAnO6g==",
"requires": {
"@babel/runtime": "^7.20.6",
"html-parse-stringify": "^3.0.1"
}
},
"react-icons": {
"version": "4.7.1",
"resolved": "https://registry.npmjs.org/react-icons/-/react-icons-4.7.1.tgz",
@@ -64265,11 +64366,6 @@
"resolved": "https://registry.npmjs.org/tiny-invariant/-/tiny-invariant-1.3.1.tgz",
"integrity": "sha512-AD5ih2NlSssTCwsMznbvwMZpJ1cbhkGd2uueNxzv2jDlEeZdU04JQfRnggJQ8DrcVBGjAsCKwFBbDlVNtEMlzw=="
},
"tiny-warning": {
"version": "1.0.3",
"resolved": "https://registry.npmjs.org/tiny-warning/-/tiny-warning-1.0.3.tgz",
"integrity": "sha512-lBN9zLN/oAf68o3zNXYrdCt1kP8WsiGW8Oo2ka41b2IM5JL/S1CTyX1rW0mb/zSuJun0ZUrDxx4sqvYS2FWzPA=="
},
"tmp": {
"version": "0.2.1",
"resolved": "https://registry.npmjs.org/tmp/-/tmp-0.2.1.tgz",
@@ -64665,6 +64761,11 @@
"imurmurhash": "^0.1.4"
}
},
"unique-username-generator": {
"version": "1.1.3",
"resolved": "https://registry.npmjs.org/unique-username-generator/-/unique-username-generator-1.1.3.tgz",
"integrity": "sha512-TB6YdqPMKMpTSgxAzjZkKWtmpZPHvARoWreCKBpc1UrLFz/0C6Q96/qdjpLr9OXPCHk16sD1LHjTr3JDj7q2JA=="
},
"unist-builder": {
"version": "2.0.3",
"resolved": "https://registry.npmjs.org/unist-builder/-/unist-builder-2.0.3.tgz",
@@ -65038,6 +65139,11 @@
"integrity": "sha512-2ham8XPWTONajOR0ohOKOHXkm3+gaBmGut3SRuu75xLd/RRaY6vqgh8NBYYk7+RW3u5AtzPQZG8F10LHkl0lAQ==",
"dev": true
},
"void-elements": {
"version": "3.1.0",
"resolved": "https://registry.npmjs.org/void-elements/-/void-elements-3.1.0.tgz",
"integrity": "sha512-Dhxzh5HZuiHQhbvTW9AMetFfBHDMYpo23Uo9btPXgdYP+3T5S+p+jgNy7spra+veYhBP2dCSgxR/i2Y02h5/6w=="
},
"w3c-xmlserializer": {
"version": "4.0.0",
"resolved": "https://registry.npmjs.org/w3c-xmlserializer/-/w3c-xmlserializer-4.0.0.tgz",
website/package.json
+4 -1
View File
@@ -46,22 +46,25 @@
"eslint-config-next": "13.0.6",
"eslint-plugin-simple-import-sort": "^8.0.0",
"focus-visible": "^5.2.0",
"formik": "^2.2.9",
"framer-motion": "^6.5.1",
"install": "^0.13.0",
"next": "13.0.6",
"next-auth": "^4.18.6",
"next-i18next": "^13.0.3",
"nodemailer": "^6.8.0",
"npm": "^9.2.0",
"postcss-focus-visible": "^7.1.0",
"react": "18.2.0",
"react-dom": "18.2.0",
"react-feature-flags": "^1.0.0",
"react-hook-form": "^7.42.1",
"react-i18next": "^12.1.4",
"react-icons": "^4.7.1",
"react-table": "^7.8.0",
"sharp": "^0.31.3",
"swr": "^2.0.0",
"tailwindcss": "^3.2.4",
"unique-username-generator": "^1.1.3",
"use-debounce": "^9.0.2"
},
"devDependencies": {
website/public/locales/en/common.json
+4
View File
@@ -0,0 +1,4 @@
{
"discord": "Discord",
"github": "GitHub"
}
website/public/locales/en/index.json
+16
View File
@@ -0,0 +1,16 @@
{
"title": "Open Assistant",
"subtitle": "Conversational AI for everyone.",
"description": "Conversational AI for everyone. An open source project to create a chat enabled GPT LLM run by LAION and contributors around the world.",
"blurb": "We believe we can create a revolution.",
"blurb1": "In the same way that Stable Diffusion helped the world make art and images in new ways, we want to improve the world by providing amazing conversational AI.",
"join_us_title": "Join us",
"join_us_description": "All open source projects begin with people like you. Open source is the belief that if we collaborate we can together gift our knowledge and technology to the world for the benefit of humanity. Are you in? Find us here:",
"faq_title": "Frequently Asked Questions",
"faq_items": {
"q0": "How far along is this project?",
"a0": "We are in the early stages of development, working from established research in applying RLHF to large language models.",
"q1": "Who is behind Open Assistant?",
"a1": "Open Assistant is a project organized by LAION and individuals around the world interested in bringing this technology to everyone."
}
}
website/src/components/AnimatedCircles/AnimatedCircles.tsx
+52
View File
@@ -0,0 +1,52 @@
import { Box, useColorMode } from "@chakra-ui/react";
import React, { useId } from "react";
export const AnimatedCircles = () => {
const id = useId();
const { colorMode } = useColorMode();
const baseRingColor = colorMode === "light" ? "#d4d4d4" : "#005a69";
const gradStopColor = colorMode === "light" ? "#06b6d4" : "#00f2ff";
return (
<Box className="absolute left-1/2 top-4 h-[1026px] w-[1026px] -translate-x-1/3 stroke-gray-300/70 [mask-image:linear-gradient(to_bottom,white_20%,transparent_75%)] sm:top-16 sm:-translate-x-1/2 lg:-top-16 lg:ml-12 xl:-top-14 xl:ml-0">
<svg
viewBox="0 0 1026 1026"
fill="none"
aria-hidden="true"
className="absolute inset-0 h-full w-full animate-spin-slow"
>
<path
d="M1025 513c0 282.77-229.23 512-512 512S1 795.77 1 513 230.23 1 513 1s512 229.23 512 512Z"
stroke={baseRingColor}
strokeOpacity="0.7"
/>
<path d="M513 1025C230.23 1025 1 795.77 1 513" stroke={`url(#${id}-gradient-1)`} strokeLinecap="round" />
<defs>
<linearGradient id={`${id}-gradient-1`} x1="1" y1="513" x2="1" y2="1025" gradientUnits="userSpaceOnUse">
<stop stopColor={gradStopColor} />
<stop offset="1" stopColor={gradStopColor} stopOpacity="0" />
</linearGradient>
</defs>
</svg>
<svg
viewBox="0 0 1026 1026"
fill="none"
aria-hidden="true"
className="absolute inset-0 h-full w-full animate-spin-reverse-slower"
>
<path
d="M913 513c0 220.914-179.086 400-400 400S113 733.914 113 513s179.086-400 400-400 400 179.086 400 400Z"
stroke={baseRingColor}
strokeOpacity="0.7"
/>
<path d="M913 513c0 220.914-179.086 400-400 400" stroke={`url(#${id}-gradient-2)`} strokeLinecap="round" />
<defs>
<linearGradient id={`gradient-2`} x1="913" y1="513" x2="913" y2="913" gradientUnits="userSpaceOnUse">
<stop stopColor={gradStopColor} />
<stop offset="1" stopColor={gradStopColor} stopOpacity="0" />
</linearGradient>
</defs>
</svg>
</Box>
);
};
website/src/components/AnimatedCircles/index.tsx
+1
View File
@@ -0,0 +1 @@
export { AnimatedCircles } from "./AnimatedCircles";
website/src/components/CallToAction.tsx
+47 -47
View File
@@ -1,9 +1,14 @@
import { useColorMode } from "@chakra-ui/react";
import { Box, Link, Text, useColorMode } from "@chakra-ui/react";
import { useTranslation } from "next-i18next";
import { useId } from "react";
import { FaDiscord, FaGithub } from "react-icons/fa";
import { Container } from "./Container";
function CircleBackground({ width = 558, height = 558, ...props }) {
const CIRCLE_HEIGHT = 558;
const CIRCLE_WIDTH = 558;
function CircleBackground() {
const id = useId();
const { colorMode } = useColorMode();
@@ -11,7 +16,14 @@ function CircleBackground({ width = 558, height = 558, ...props }) {
const gradStopColor = colorMode === "light" ? "#fff" : "#000";
return (
<svg viewBox="0 0 558 558" width={width} height={height} fill="none" aria-hidden="true" {...props}>
<svg
viewBox={`0 0 ${CIRCLE_HEIGHT} ${CIRCLE_WIDTH}`}
width={CIRCLE_WIDTH}
height={CIRCLE_HEIGHT}
fill="none"
aria-hidden="true"
className="animate-spin-slower"
>
<defs>
<linearGradient id={id} x1="79" y1="16" x2="105" y2="237" gradientUnits="userSpaceOnUse">
<stop stopColor={gradStopColor} />
@@ -30,66 +42,54 @@ function CircleBackground({ width = 558, height = 558, ...props }) {
export function CallToAction() {
const { colorMode } = useColorMode();
const { t } = useTranslation();
const bgColorClass = colorMode === "light" ? "bg-gray-900" : "bg-gray-50";
const headingColorClass = colorMode === "light" ? "text-white" : "text-black";
const textColorClass = colorMode === "light" ? "text-gray-300" : "text-black";
return (
<section id="join-us" className={`relative overflow-hidden py-20 sm:py-28 ${bgColorClass} ${textColorClass}`}>
<div className="absolute top-1/2 left-20 -translate-y-1/2 sm:left-1/2 sm:-translate-x-1/2">
<CircleBackground className="animate-spin-slower" />
</div>
<Box
as="section"
id="join-us"
className={`relative overflow-hidden py-20 sm:py-28 ${bgColorClass} ${textColorClass}`}
>
<Box className="absolute top-1/2 left-20 -translate-y-1/2 sm:left-1/2 sm:-translate-x-1/2">
<CircleBackground />
</Box>
<Container className="relative">
<div className="mx-auto max-w-md sm:text-center">
<h2 className={`text-3xl font-medium tracking-tight sm:text-4xl ${headingColorClass}`}>Join Us</h2>
<p className="mt-4 text-lg">
All open source projects begin with people like you. Open source is the belief that if we collaborate we can
together gift our knowledge and technology to the world for the benefit of humanity. Are you in? Find us
here:
</p>
<div className="mt-8 flex justify-center">
<a href="https://ykilcher.com/open-assistant-discord" rel="noreferrer" target="_blank">
<Box className="mx-auto max-w-md sm:text-center">
<Text as="h2" className={`text-3xl font-medium tracking-tight sm:text-4xl ${headingColorClass}`}>
{t("index:join_us_title")}
</Text>
<Text as="p" className="mt-4 text-lg">
{t("index:join_us_description")}
</Text>
<Box className="mt-8 flex justify-center">
<Link href="https://ykilcher.com/open-assistant-discord" rel="noreferrer" target="_blank">
<button
type="button"
className="mb-2 ml-6 flex items-center rounded-md border border-transparent bg-blue-600 px-6 py-3 text-base font-medium text-white shadow-sm hover:bg-blue-700 focus:outline-none focus:ring-2 focus:ring-blue-500 focus:ring-offset-2"
>
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 640 512" className="w-6 h-6">
<path
fill="currentColor"
d="M524.531,69.836a1.5,1.5,0,0,0-.764-.7A485.065,485.065,0,0,0,404.081,32.03a1.816,1.816,0,0,0-1.923.91,337.461,337.461,0,0,0-14.9,30.6,447.848,447.848,0,0,0-134.426,0,309.541,309.541,0,0,0-15.135-30.6,1.89,1.89,0,0,0-1.924-.91A483.689,483.689,0,0,0,116.085,69.137a1.712,1.712,0,0,0-.788.676C39.068,183.651,18.186,294.69,28.43,404.354a2.016,2.016,0,0,0,.765,1.375A487.666,487.666,0,0,0,176.02,479.918a1.9,1.9,0,0,0,2.063-.676A348.2,348.2,0,0,0,208.12,430.4a1.86,1.86,0,0,0-1.019-2.588,321.173,321.173,0,0,1-45.868-21.853,1.885,1.885,0,0,1-.185-3.126c3.082-2.309,6.166-4.711,9.109-7.137a1.819,1.819,0,0,1,1.9-.256c96.229,43.917,200.41,43.917,295.5,0a1.812,1.812,0,0,1,1.924.233c2.944,2.426,6.027,4.851,9.132,7.16a1.884,1.884,0,0,1-.162,3.126,301.407,301.407,0,0,1-45.89,21.83,1.875,1.875,0,0,0-1,2.611,391.055,391.055,0,0,0,30.014,48.815,1.864,1.864,0,0,0,2.063.7A486.048,486.048,0,0,0,610.7,405.729a1.882,1.882,0,0,0,.765-1.352C623.729,277.594,590.933,167.465,524.531,69.836ZM222.491,337.58c-28.972,0-52.844-26.587-52.844-59.239S193.056,219.1,222.491,219.1c29.665,0,53.306,26.82,52.843,59.239C275.334,310.993,251.924,337.58,222.491,337.58Zm195.38,0c-28.971,0-52.843-26.587-52.843-59.239S388.437,219.1,417.871,219.1c29.667,0,53.307,26.82,52.844,59.239C470.715,310.993,447.538,337.58,417.871,337.58Z"
/>
</svg>
<span className="text-lg ml-3">Discord</span>
<FaDiscord size={25} />
<Text as="span" className="text-lg ml-3">
{t("discord")}
</Text>
</button>
</a>
<a href="https://github.com/LAION-AI/Open-Assistant" rel="noreferrer" target="_blank">
</Link>
<Link href="https://github.com/LAION-AI/Open-Assistant" rel="noreferrer" target="_blank">
<button
type="button"
className="mb-2 ml-6 flex items-center rounded-md border border-transparent bg-blue-600 px-6 py-3 text-base font-medium text-white shadow-sm hover:bg-blue-700 focus:outline-none focus:ring-2 focus:ring-blue-500 focus:ring-offset-2"
>
<svg
className="mr-2 -ml-1 w-6 h-6"
aria-hidden="true"
focusable="false"
data-prefix="fab"
data-icon="github"
role="img"
xmlns="http://www.w3.org/2000/svg"
viewBox="0 0 496 512"
>
<path
fill="currentColor"
d="M165.9 397.4c0 2-2.3 3.6-5.2 3.6-3.3 .3-5.6-1.3-5.6-3.6 0-2 2.3-3.6 5.2-3.6 3-.3 5.6 1.3 5.6 3.6zm-31.1-4.5c-.7 2 1.3 4.3 4.3 4.9 2.6 1 5.6 0 6.2-2s-1.3-4.3-4.3-5.2c-2.6-.7-5.5 .3-6.2 2.3zm44.2-1.7c-2.9 .7-4.9 2.6-4.6 4.9 .3 2 2.9 3.3 5.9 2.6 2.9-.7 4.9-2.6 4.6-4.6-.3-1.9-3-3.2-5.9-2.9zM244.8 8C106.1 8 0 113.3 0 252c0 110.9 69.8 205.8 169.5 239.2 12.8 2.3 17.3-5.6 17.3-12.1 0-6.2-.3-40.4-.3-61.4 0 0-70 15-84.7-29.8 0 0-11.4-29.1-27.8-36.6 0 0-22.9-15.7 1.6-15.4 0 0 24.9 2 38.6 25.8 21.9 38.6 58.6 27.5 72.9 20.9 2.3-16 8.8-27.1 16-33.7-55.9-6.2-112.3-14.3-112.3-110.5 0-27.5 7.6-41.3 23.6-58.9-2.6-6.5-11.1-33.3 2.6-67.9 20.9-6.5 69 27 69 27 20-5.6 41.5-8.5 62.8-8.5s42.8 2.9 62.8 8.5c0 0 48.1-33.6 69-27 13.7 34.7 5.2 61.4 2.6 67.9 16 17.7 25.8 31.5 25.8 58.9 0 96.5-58.9 104.2-114.8 110.5 9.2 7.9 17 22.9 17 46.4 0 33.7-.3 75.4-.3 83.6 0 6.5 4.6 14.4 17.3 12.1C428.2 457.8 496 362.9 496 252 496 113.3 383.5 8 244.8 8zM97.2 352.9c-1.3 1-1 3.3 .7 5.2 1.6 1.6 3.9 2.3 5.2 1 1.3-1 1-3.3-.7-5.2-1.6-1.6-3.9-2.3-5.2-1zm-10.8-8.1c-.7 1.3 .3 2.9 2.3 3.9 1.6 1 3.6 .7 4.3-.7 .7-1.3-.3-2.9-2.3-3.9-2-.6-3.6-.3-4.3 .7zm32.4 35.6c-1.6 1.3-1 4.3 1.3 6.2 2.3 2.3 5.2 2.6 6.5 1 1.3-1.3 .7-4.3-1.3-6.2-2.2-2.3-5.2-2.6-6.5-1zm-11.4-14.7c-1.6 1-1.6 3.6 0 5.9 1.6 2.3 4.3 3.3 5.6 2.3 1.6-1.3 1.6-3.9 0-6.2-1.4-2.3-4-3.3-5.6-2z"
></path>
</svg>
<span className="text-lg ml-1">Github</span>
<FaGithub size={25} />
<Text as="span" className="text-lg ml-3">
{t("github")}
</Text>
</button>
</a>
</div>
</div>
</Link>
</Box>
</Box>
</Container>
</section>
</Box>
);
}
website/src/components/CollapsableText.tsx
+5 -2
View File
@@ -11,6 +11,8 @@ import {
} from "@chakra-ui/react";
import React, { ReactNode } from "react";
const killEvent = (e) => e.stopPropagation();
export const CollapsableText = ({
text,
maxLength = 220,
@@ -44,8 +46,9 @@ export const CollapsableText = ({
</Button>
</span>
<Modal isOpen={isOpen} onClose={onClose} size="xl" scrollBehavior={"inside"}>
<ModalOverlay style={{ width: "100%", height: "100%" }}>
<ModalContent maxH="400">
{/* we kill the event here to disable drag and drop, since it is in the same container */}
<ModalOverlay onMouseDown={killEvent}>
<ModalContent alignItems="center">
<ModalHeader>Full Text</ModalHeader>
<ModalCloseButton />
<ModalBody>{text}</ModalBody>
website/src/components/Dashboard/LeaderboardTable.tsx
+5 -29
View File
@@ -1,17 +1,16 @@
import { Box, Link, Stack, StackDivider, Text, useColorModeValue } from "@chakra-ui/react";
import { Box, Link, Text, useColorModeValue } from "@chakra-ui/react";
import NextLink from "next/link";
import { get } from "src/lib/api";
import useSWR from "swr";
import { LeaderboardGridCell } from "src/components/LeaderboardGridCell";
import { LeaderboardTimeFrame } from "src/types/Leaderboard";
export function LeaderboardTable() {
const backgroundColor = useColorModeValue("white", "gray.700");
const accentColor = useColorModeValue("gray.200", "gray.900");
const { data: leaderboardEntries } = useSWR("/api/leaderboard", get);
return (
<main className="h-fit col-span-3">
<div className="flex flex-col gap-4">
<div className="flex items-end justify-between">
<Text className="text-2xl font-bold">Top 5 Contributors</Text>
<Text className="text-2xl font-bold">Top 5 Contributors Today</Text>
<Link as={NextLink} href="/leaderboard" _hover={{ textDecoration: "none" }}>
<Text color="blue.400" className="text-sm font-bold">
View All -&gt;
@@ -25,30 +24,7 @@ export function LeaderboardTable() {
borderRadius="xl"
className="p-6 shadow-sm"
>
<Stack divider={<StackDivider />} spacing="4">
<div className="grid grid-cols-4 items-center font-bold">
<p>Name</p>
<div className="col-start-4 flex justify-center">
<p>Score</p>
</div>
</div>
{leaderboardEntries?.map(({ display_name, score }, idx) => (
<div key={idx} className="grid grid-cols-4 items-center">
<div className="flex items-center gap-3">
{/*
<Image alt="Profile Picture" src={item.image} boxSize="7" borderRadius="full"></Image>
*/}
<p>{display_name}</p>
{/*
<Badge colorScheme="purple">{item.streakCount}</Badge>
*/}
</div>
<Box bg={backgroundColor} className="col-start-4 flex justify-center">
<p>{score}</p>
</Box>
</div>
))}
</Stack>
<LeaderboardGridCell timeFrame={LeaderboardTimeFrame.day} />
</Box>
</div>
</main>
website/src/components/Dashboard/SlimFooter.tsx
+46
View File
@@ -0,0 +1,46 @@
import { Box, Divider } from "@chakra-ui/react";
import Image from "next/image";
import Link from "next/link";
import { useMemo } from "react";
export function SlimFooter() {
return (
<footer>
<Box>
<Divider />
<Box display="flex" gap="4" flexDir="column" alignItems="center" my="8">
<Box display="flex" alignItems="center">
<Link href="/" aria-label="Home" className="flex items-center gap-1">
<Image src="/images/logos/logo.svg" className="mx-auto object-fill" width="48" height="48" alt="logo" />
</Link>
</Box>
<nav>
<Box display="flex" gap="5" fontSize="xs" color="blue.500">
<FooterLink href="/privacy-policy" label="Privacy Policy" />
<FooterLink href="/terms-of-service" label="Terms of Service" />
<FooterLink href="https://github.com/LAION-AI/Open-Assistant" label="Github" />
<FooterLink href="https://ykilcher.com/open-assistant-discord" label="Discord" />
<FooterLink href="https://projects.laion.ai/Open-Assistant/" label="Docs" />
</Box>
</nav>
</Box>
</Box>
</footer>
);
}
const FooterLink = ({ href, label }: { href: string; label: string }) =>
useMemo(
() => (
<Link
href={href}
rel="noopener noreferrer nofollow"
target="_blank"
aria-label={label}
className="hover:underline underline-offset-2"
>
{label}
</Link>
),
[href, label]
);
website/src/components/Dashboard/WelcomeCard.tsx
+43
View File
@@ -0,0 +1,43 @@
import { Box, Divider, Text, useColorMode } from "@chakra-ui/react";
import { useSession } from "next-auth/react";
export function WelcomeCard() {
const { colorMode } = useColorMode();
const backgroundColor = colorMode === "light" ? "white" : "gray.700";
const titleColor = colorMode === "light" ? "blue.500" : "blue.300";
const { data: session } = useSession();
if (!session) {
return <></>;
}
if (session && session.user && session.user.isNew)
return (
<>
<Box
bgGradient="linear(to-r, blue.300, purple.500)"
borderRadius="xl"
p="1px"
shadow="base"
position="relative"
>
<Box bg={backgroundColor} borderRadius="xl" p="6" pt="4" pr="12">
<Box pb="2">
<Text as="h1" fontWeight="extrabold" fontSize="3xl" color={titleColor}>
Welcome, {session.user.name || "Contributor"}!
</Text>
</Box>
<Box>
<Text>
Open Assistant is an open-source AI assistant that uses and trains advanced language models to
understand and respond to humans.
</Text>
<Divider my="4" />
<Text>Complete tasks to help train the model and earn points.</Text>
</Box>
</Box>
</Box>
</>
);
}
website/src/components/Dashboard/index.ts
+1
View File
@@ -1,2 +1,3 @@
export { LeaderboardTable } from "./LeaderboardTable";
export { TaskOption } from "./TaskOption";
export { WelcomeCard } from "./WelcomeCard";
website/src/components/EmptyState.tsx
-4
View File
@@ -28,7 +28,3 @@ export const EmptyState = (props: EmptyStateProps) => {
export const TaskEmptyState = () => {
return <EmptyState text="Looks like no tasks were found." icon={FiAlertTriangle} />;
};
export const PageEmptyState = () => {
return <EmptyState text="Sorry, the page you are looking for does not exist." icon={FiAlertTriangle} />;
};
website/src/components/Faq.tsx
+25 -56
View File
@@ -1,73 +1,42 @@
import { useColorMode } from "@chakra-ui/react";
import { Box, List, ListItem, Text, useColorMode } from "@chakra-ui/react";
import { useTranslation } from "next-i18next";
import { Container } from "./Container";
const faqs = [
[
{
question: "How far along is this project?",
answer:
"We are in the early stages of development, working from established research in applying RLHF to large language models.",
},
],
[
{
question: "Who is behind Open Assistant?",
answer:
"Open Assistant is a project organized by LAION and individuals around the world interested in bringing this technology to everyone.",
},
],
[
// {
// question: 'Where can I learn more?',
// answer:
// 'Please feel free to reach out to us on Discord. We are happy to answer any questions you may have.',
// },
],
];
const FAQS = Array.from({ length: 2 });
export function Faq() {
const { colorMode } = useColorMode();
const { t } = useTranslation("index");
const headingColorClass = colorMode === "light" ? "text-gray-900" : "text-white";
const textColorClass = colorMode === "light" ? "text-gray-700" : "text-gray-100";
return (
<section id="faq" aria-labelledby="faqs-title" className="border-t border-gray-200 py-20 sm:py-32">
<Box as="section" id="faq" aria-labelledby="faqs-title" className="border-t border-gray-200 py-20 sm:py-32">
<Container className="">
<div className="mx-auto max-w-2xl lg:mx-0">
<h2 id="faqs-title" className={`text-3xl font-medium tracking-tight ${headingColorClass}`}>
Frequently Asked Questions
</h2>
{/* <p className="mt-2 text-lg text-gray-600">
If you have anything else you want to ask,{' '}
<Link
href="mailto:info@open-assistant.tech"
className="text-gray-900 underline"
>
reach out to us
</Link>
.
</p> */}
</div>
<ul
<Box className="mx-auto max-w-2xl lg:mx-0">
<Text as="h2" id="faqs-title" className={`text-3xl font-medium tracking-tight ${headingColorClass}`}>
{t("faq_title")}
</Text>
</Box>
<List
role="list"
className="mx-auto mt-16 grid max-w-2xl grid-cols-1 gap-8 sm:mt-20 lg:max-w-none lg:grid-cols-3"
>
{faqs.map((column, columnIndex) => (
<li key={columnIndex}>
<ul role="list" className="space-y-10">
{column.map((faq, faqIndex) => (
<li key={faqIndex}>
<h3 className={`text-lg font-semibold leading-6 ${headingColorClass}`}>{faq.question}</h3>
<p className={`mt-4 text-sm ${textColorClass}`}>{faq.answer}</p>
</li>
))}
</ul>
</li>
))}
</ul>
{FAQS.map((_, index) => {
return (
<ListItem className="space-y-10" key={`question_${index}`}>
<Text as="h3" className={`text-lg font-semibold leading-6 ${headingColorClass}`}>
{t(`faq_items.q${index}`)}
</Text>
<Text as="p" className={`mt-4 text-sm ${textColorClass}`}>
{t(`faq_items.a${index}`)}
</Text>
</ListItem>
);
})}
</List>
</Container>
</section>
</Box>
);
}
website/src/components/FlaggableElement.tsx
+5 -1
View File
@@ -26,7 +26,7 @@ import { useEffect, useReducer } from "react";
import { FiAlertCircle } from "react-icons/fi";
import { get, post } from "src/lib/api";
import { Message } from "src/types/Conversation";
import { colors } from "styles/Theme/colors";
import { colors } from "src/styles/Theme/colors";
import useSWR from "swr";
import useSWRMutation from "swr/mutation";
@@ -105,6 +105,10 @@ export const FlaggableElement = (props: FlaggableElementProps) => {
if (isLoading) {
return;
}
if (!data) {
updateReport({ type: "load_labels", labels: [] });
return;
}
const { valid_labels } = data;
updateReport({ type: "load_labels", labels: valid_labels });
}, [data, isLoading]);
website/src/components/Footer.tsx
+63 -35
View File
@@ -1,40 +1,72 @@
import { useColorMode } from "@chakra-ui/react";
import { Box, Divider, Flex, Text, useColorMode } from "@chakra-ui/react";
import Image from "next/image";
import Link from "next/link";
import { useMemo } from "react";
export function Footer() {
const { colorMode } = useColorMode();
const bgColorClass = colorMode === "light" ? "bg-transparent" : "bg-gray-800";
const borderClass = colorMode === "light" ? "border-slate-200" : "border-transparent";
const backgroundColor = colorMode === "light" ? "white" : "gray.800";
const textColor = colorMode === "light" ? "black" : "gray.300";
return (
<footer className={bgColorClass}>
<div className={`flex mx-auto max-w-7xl justify-between border-t p-10 ${borderClass}`}>
<div className="flex items-center pr-8">
<Link href="/" aria-label="Home" className="flex items-center">
<Image src="/images/logos/logo.svg" className="mx-auto object-fill" width="52" height="52" alt="logo" />
</Link>
<footer>
<Box bg={backgroundColor}>
<Divider />
<Box
display="flex"
flexDirection={["column", "row"]}
justifyContent="space-between"
alignItems="center"
gap="6"
p="8"
pb={["14", "8"]}
w="full"
mx="auto"
maxWidth="7xl"
>
<Flex alignItems="center">
<Box pr="2">
<Link href="/" aria-label="Home">
<Image src="/images/logos/logo.svg" width="52" height="52" alt="logo" />
</Link>
</Box>
<div className="ml-2">
<p className="text-base font-bold">Open Assistant</p>
<p className="text-sm">Conversational AI for everyone.</p>
</div>
</div>
<Box>
<Text fontSize="md" fontWeight="bold">
Open Assistant
</Text>
<Text fontSize="sm" color="gray.500">
Conversational AI for everyone.
</Text>
</Box>
</Flex>
<nav className="grid grid-cols-2 gap-20 leading-5 text-sm">
<div className="flex flex-col">
<b className="pb-1">Legal</b>
<FooterLink href="/privacy-policy" label="Privacy Policy" />
<FooterLink href="/terms-of-service" label="Terms of Service" />
</div>
<div className="flex flex-col">
<b className="pb-1">Connect</b>
<FooterLink href="https://github.com/LAION-AI/Open-Assistant" label="Github" />
<FooterLink href="https://ykilcher.com/open-assistant-discord" label="Discord" />
</div>
</nav>
</div>
<nav>
<Box display="flex" flexDirection={["column", "row"]} gap={["6", "14"]} fontSize="sm">
<Flex direction="column" alignItems={["center", "start"]}>
<Text fontWeight="bold" color={textColor}>
Legal
</Text>
<FooterLink href="/privacy-policy" label="Privacy Policy" />
<FooterLink href="/terms-of-service" label="Terms of Service" />
</Flex>
<Flex direction="column" alignItems={["center", "start"]}>
<Text fontWeight="bold" color={textColor}>
Connect
</Text>
<FooterLink href="https://github.com/LAION-AI/Open-Assistant" label="Github" />
<FooterLink href="https://ykilcher.com/open-assistant-discord" label="Discord" />
</Flex>
<Flex direction="column" alignItems={["center", "start"]}>
<Text fontWeight="bold" color={textColor}>
About
</Text>
<FooterLink href="https://projects.laion.ai/Open-Assistant" label="Docs" />
</Flex>
</Box>
</nav>
</Box>
</Box>
</footer>
);
}
@@ -42,14 +74,10 @@ export function Footer() {
const FooterLink = ({ href, label }: { href: string; label: string }) =>
useMemo(
() => (
<Link
href={href}
rel="noopener noreferrer nofollow"
target="_blank"
aria-label={label}
className="hover:underline underline-offset-2"
>
{label}
<Link href={href} rel="noopener noreferrer nofollow" target="_blank" aria-label={label}>
<Text color="blue.500" textUnderlineOffset={2} _hover={{ textDecoration: "underline" }}>
{label}
</Text>
</Link>
),
[href, label]
website/src/components/Header/Header.tsx
+18 -21
View File
@@ -1,4 +1,4 @@
import { Box, Button, Text, useColorMode } from "@chakra-ui/react";
import { Box, Button, Text, Flex } from "@chakra-ui/react";
import Image from "next/image";
import Link from "next/link";
import { useSession } from "next-auth/react";
@@ -13,10 +13,12 @@ function AccountButton() {
return;
}
return (
<Link href="/auth/signin" aria-label="Home" className="flex items-center">
<Button variant="outline" leftIcon={<FaUser />}>
Sign in
</Button>
<Link href="/auth/signin" aria-label="Home">
<Flex alignItems="center">
<Button variant="outline" leftIcon={<FaUser />}>
Sign in
</Button>
</Flex>
</Link>
);
}
@@ -25,30 +27,25 @@ export function Header(props) {
const { data: session } = useSession();
const homeURL = session ? "/dashboard" : "/";
const { colorMode } = useColorMode();
const borderClass = props.transparent
? ""
: colorMode === "light"
? "border-b border-gray-400"
: "border-b border-zinc-800";
return (
<nav className={`oa-basic-theme ${borderClass}`}>
<Box className="relative z-10 flex justify-between px-4 py-4">
<div className="relative z-10 flex items-center gap-10">
<Link href={homeURL} aria-label="Home" className="flex items-center">
<nav className="oa-basic-theme">
<Box display="flex" justifyContent="space-between" p="4">
<Link href={homeURL} aria-label="Home">
<Flex alignItems="center">
<Image src="/images/logos/logo.svg" className="mx-auto object-fill" width="50" height="50" alt="logo" />
<Text fontFamily="inter" fontSize="2xl" fontWeight="bold" className="ml-3">
<Text fontFamily="inter" fontSize="2xl" fontWeight="bold" ml="3">
Open Assistant
</Text>
</Link>
</div>
<div className="flex items-center gap-4">
</Flex>
</Link>
<Flex alignItems="center" gap="4">
<Flags authorizedFlags={["flagTest"]}>
<div>FlagTest</div>
<Text>FlagTest</Text>
</Flags>
<AccountButton />
<UserMenu />
</div>
</Flex>
</Box>
</nav>
);
website/src/components/Header/UserMenu.tsx
+1 -1
View File
@@ -74,7 +74,7 @@ export function UserMenu() {
<Box display="flex" alignItems="center" gap="3" p="1" paddingRight={[1, 1, 1, 6, 6]}>
<Avatar size="sm" bgImage={session.user.image}></Avatar>
<Text data-cy="username" className="hidden lg:flex">
{session.user.name || session.user.email}
{session.user.name || "New User"}
</Text>
</Box>
</MenuButton>
website/src/components/Hero.tsx
+25 -76
View File
@@ -1,87 +1,36 @@
import { useColorMode } from "@chakra-ui/react";
import { Box, Text, useColorMode } from "@chakra-ui/react";
import Image from "next/image";
import { useId } from "react";
import { useTranslation } from "next-i18next";
import { Container } from "./Container";
function BackgroundIllustration(props) {
const id = useId();
const { colorMode } = useColorMode();
const baseRingColor = colorMode === "light" ? "#d4d4d4" : "#005a69";
const gradStopColor = colorMode === "light" ? "#06b6d4" : "#00f2ff";
return (
<div {...props}>
<svg
viewBox="0 0 1026 1026"
fill="none"
aria-hidden="true"
className="absolute inset-0 h-full w-full animate-spin-slow"
>
<path
d="M1025 513c0 282.77-229.23 512-512 512S1 795.77 1 513 230.23 1 513 1s512 229.23 512 512Z"
stroke={baseRingColor}
strokeOpacity="0.7"
/>
<path d="M513 1025C230.23 1025 1 795.77 1 513" stroke={`url(#${id}-gradient-1)`} strokeLinecap="round" />
<defs>
<linearGradient id={`${id}-gradient-1`} x1="1" y1="513" x2="1" y2="1025" gradientUnits="userSpaceOnUse">
<stop stopColor={gradStopColor} />
<stop offset="1" stopColor={gradStopColor} stopOpacity="0" />
</linearGradient>
</defs>
</svg>
<svg
viewBox="0 0 1026 1026"
fill="none"
aria-hidden="true"
className="absolute inset-0 h-full w-full animate-spin-reverse-slower"
>
<path
d="M913 513c0 220.914-179.086 400-400 400S113 733.914 113 513s179.086-400 400-400 400 179.086 400 400Z"
stroke={baseRingColor}
strokeOpacity="0.7"
/>
<path d="M913 513c0 220.914-179.086 400-400 400" stroke={`url(#${id}-gradient-2)`} strokeLinecap="round" />
<defs>
<linearGradient id={`${id}-gradient-2`} x1="913" y1="513" x2="913" y2="913" gradientUnits="userSpaceOnUse">
<stop stopColor={gradStopColor} />
<stop offset="1" stopColor={gradStopColor} stopOpacity="0" />
</linearGradient>
</defs>
</svg>
</div>
);
}
import { AnimatedCircles } from "./AnimatedCircles";
export function Hero() {
const { t } = useTranslation("index");
const { colorMode } = useColorMode();
const pTextColor = colorMode === "light" ? "text-gray-600" : "text-white";
const fancyTextGradientClasses =
colorMode === "light" ? "from-blue-600 via-sky-400 to-blue-700" : "from-blue-500 via-sky-300 to-blue-400";
return (
<div className="overflow-hidden py-20 sm:py-32 lg:pb-32 xl:pb-36">
<Box className="overflow-hidden py-20 sm:py-32 lg:pb-32 xl:pb-36">
<Container className="">
<div className="lg:grid lg:grid-cols-12 lg:gap-x-8 lg:gap-y-20">
<div className="relative z-10 mx-auto max-w-2xl lg:col-span-7 lg:max-w-none lg:pt-6 xl:col-span-6">
<h1 className="text-5xl mb-6 font-bold tracking-tight">Open Assistant</h1>
<p
className={`bg-gradient-to-r ${fancyTextGradientClasses} mt-8 text-3xl inline bg-clip-text font-display tracking-tight text-transparent`}
<Box className="lg:grid lg:grid-cols-12 lg:gap-x-8 lg:gap-y-20">
<Box className="relative mx-auto max-w-2xl lg:col-span-7 lg:max-w-none lg:pt-6 xl:col-span-6">
<Text as="h1" className="text-5xl mb-6 font-bold tracking-tight">
{t("title")}
</Text>
<Text
as="h2"
className={`bg-gradient-to-r ${fancyTextGradientClasses} font-bold mt-8 text-3xl inline bg-clip-text font-display tracking-tight text-transparent`}
>
<b>Conversational AI for everyone.</b>
</p>
<p className={`mt-6 text-lg ${pTextColor}`}>We believe we can create a revolution.</p>
<p className={`mt-6 text-lg ${pTextColor}`}>
In the same way that Stable Diffusion helped the world make art and images in new ways, we want to improve
the world by providing amazing conversational AI.
</p>
</div>
<div className="relative mt-10 sm:mt-20 lg:col-span-5 lg:row-span-2 lg:mt-0 xl:col-span-6">
<BackgroundIllustration className="absolute left-1/2 top-4 h-[1026px] w-[1026px] -translate-x-1/3 stroke-gray-300/70 [mask-image:linear-gradient(to_bottom,white_20%,transparent_75%)] sm:top-16 sm:-translate-x-1/2 lg:-top-16 lg:ml-12 xl:-top-14 xl:ml-0" />
<div className="-mx-4 h-[448px] px-9 [mask-image:linear-gradient(to_bottom,white_60%,transparent)] sm:mx-0 lg:absolute lg:-inset-x-10 lg:-top-10 lg:-bottom-20 lg:h-auto lg:px-0 lg:pt-10 xl:-bottom-32">
{t("subtitle")}
</Text>
<Text className={`mt-6 text-lg ${pTextColor}`}>{t("blurb")}</Text>
<Text className={`mt-6 text-lg ${pTextColor}`}>{t("blurb1")}</Text>
</Box>
<Box className="relative mt-10 sm:mt-20 lg:col-span-5 lg:row-span-2 lg:mt-0 xl:col-span-6">
<AnimatedCircles />
<Box className="-mx-4 h-[448px] px-9 [mask-image:linear-gradient(to_bottom,white_60%,transparent)] sm:mx-0 lg:absolute lg:-inset-x-10 lg:-top-10 lg:-bottom-20 lg:h-auto lg:px-0 lg:pt-10 xl:-bottom-32">
<Image
src="/images/logos/logo.svg"
className="mx-auto mr-6 object-fill"
@@ -89,10 +38,10 @@ export function Hero() {
height="450"
alt={""}
/>
</div>
</div>
</div>
</Box>
</Box>
</Box>
</Container>
</div>
</Box>
);
}
website/src/components/Layout.tsx
+10 -4
View File
@@ -1,9 +1,11 @@
// https://nextjs.org/docs/basic-features/layouts
import { Box, Grid } from "@chakra-ui/react";
import type { NextPage } from "next";
import { FiBarChart2, FiLayout, FiMessageSquare, FiUsers } from "react-icons/fi";
import { Header } from "src/components/Header";
import { SlimFooter } from "./Dashboard/SlimFooter";
import { Footer } from "./Footer";
import { SideMenuLayout } from "./SideMenuLayout";
@@ -28,7 +30,7 @@ export const getTransparentHeaderLayout = (page: React.ReactElement) => (
);
export const getDashboardLayout = (page: React.ReactElement) => (
<div className="grid grid-rows-[min-content_1fr_min-content] h-full justify-items-stretch">
<Grid templateRows="min-content 1fr" h="full">
<Header transparent={true} />
<SideMenuLayout
menuButtonOptions={[
@@ -52,10 +54,14 @@ export const getDashboardLayout = (page: React.ReactElement) => (
},
]}
>
{page}
<Grid templateRows="1fr min-content" h="full">
<Box>{page}</Box>
<Box mt="10">
<SlimFooter />
</Box>
</Grid>
</SideMenuLayout>
<Footer />
</div>
</Grid>
);
export const getAdminLayout = (page: React.ReactElement) => (
website/src/components/LeaderboardGridCell/LeaderboardGridCell.tsx
+1 -1
View File
@@ -8,7 +8,7 @@ import useSWRImmutable from "swr/immutable";
const columns = [
{
Header: "Rank",
accessor: (item: LeaderboardEntity, rowIndex: number) => "#" + (item.user_rank + 1),
accessor: "rank",
style: { width: "90px" },
},
{
website/src/components/Messages/MessageTableEntry.tsx
+6 -1
View File
@@ -30,7 +30,12 @@ export function MessageTableEntry(props: MessageTableEntryProps) {
{props.enabled ? (
<Box width={["full", "full", "full", "fit-content"]} maxWidth={["full", "full", "full", "2xl"]}>
<Link href={`/messages/${item.id}`}>
<LinkBox bg={item.is_assistant ? backgroundColor : backgroundColor2} p="4" borderRadius="md">
<LinkBox
bg={item.is_assistant ? backgroundColor : backgroundColor2}
p="4"
borderRadius="md"
whiteSpace="pre-line"
>
{item.text}
</LinkBox>
</Link>
website/src/components/Messages/MessageWithChildren.tsx
+18 -37
View File
@@ -1,9 +1,8 @@
import { Box, CircularProgress, Stack, StackProps, Text, TextProps, useColorModeValue } from "@chakra-ui/react";
import { boolean } from "boolean";
import { useState } from "react";
import { MessageTableEntry } from "src/components/Messages/MessageTableEntry";
import { get } from "src/lib/api";
import useSWR from "swr";
import { Message } from "src/types/Conversation";
import useSWRImmutable from "swr/immutable";
const MessageHeaderProps: TextProps = {
fontSize: "xl",
@@ -21,39 +20,24 @@ const MessageStackProps: StackProps = {
interface MessageWithChildrenProps {
id: string;
depth?: number;
maxDepth?: number;
maxDepth: number;
isOnlyChild?: boolean;
}
export function MessageWithChildren(props: MessageWithChildrenProps) {
const backgroundColor = useColorModeValue("white", "gray.800");
const childBackgroundColor = useColorModeValue("gray.200", "gray.700");
const { id, depth = 0, maxDepth, isOnlyChild = true } = props;
const { id, depth, maxDepth, isOnlyChild = true } = props;
const { isLoading, data: message } = useSWRImmutable<Message>(`/api/messages/${id}`, get);
const { isLoading: isLoadingChildren, data: children = [] } = useSWRImmutable<Message[]>(
`/api/messages/${id}/children`,
get
);
const [message, setMessage] = useState(null);
const [children, setChildren] = useState(null);
const { isLoading } = useSWR(id ? `/api/messages/${id}` : null, get, {
onSuccess: (data) => {
setMessage(data);
},
onError: () => {
setMessage(null);
},
});
const { isLoading: isLoadingChildren } = useSWR(id ? `/api/messages/${id}/children` : null, get, {
onSuccess: (data) => {
setChildren(data);
},
onError: () => {
setChildren(null);
},
});
const renderRecursive = maxDepth && ((depth && depth < maxDepth) || !depth);
const isFirst = depth === 0 || !depth;
const isFirstOrOnly = isFirst || boolean(isOnlyChild);
const renderRecursive = depth < maxDepth || depth === 0;
const isFirst = depth === 0;
const isFirstOrOnly = isFirst || !!isOnlyChild;
if (isLoading || isLoadingChildren) {
return <CircularProgress isIndeterminate />;
@@ -73,15 +57,15 @@ export function MessageWithChildren(props: MessageWithChildrenProps) {
</Box>
</>
)}
{children && Array.isArray(children) && children.length > 0 ? (
renderRecursive ? (
{children.length > 0 &&
(renderRecursive ? (
<Stack {...MessageStackProps}>
<Box bg={childBackgroundColor} padding="4" borderRadius="xl">
{children.map((item, idx) => (
<Box flex="1" key={`recursiveMessageWChildren_${idx}`}>
{children.map((item) => (
<Box flex="1" key={`recursiveMessageWChildren_${item.id}`}>
<MessageWithChildren
id={item.id}
depth={depth ? depth + 1 : 1}
depth={depth + 1}
maxDepth={maxDepth}
isOnlyChild={children.length === 1 && isOnlyChild}
/>
@@ -110,10 +94,7 @@ export function MessageWithChildren(props: MessageWithChildrenProps) {
</Box>
</Stack>
</>
)
) : (
<></>
)}
))}
</>
);
}
website/src/components/Roadmap.tsx
+1 -1
View File
@@ -51,7 +51,7 @@ const Roadmap = () => {
</div>
<h4 className="font-bold text-xl text-[#858585] text-center max-w-[10rem]">Growing Up</h4>
<ul className="ml-6 md:ml-8 lg:ml-6 space-y-4 text-[#858585] list-disc">
<li>Third-Party Extentions</li>
<li>Third-Party Extensions</li>
<li>Device Control</li>
<li>Multi-Modality</li>
</ul>
website/src/components/RoleSelect.tsx
+25
View File
@@ -0,0 +1,25 @@
import { Select, SelectProps } from "@chakra-ui/react";
import { forwardRef } from "react";
import { ElementOf } from "src/types/utils";
export const roles = ["general", "admin", "banned"] as const;
export type Role = ElementOf<typeof roles>;
type RoleSelectProps = Omit<SelectProps, "defaultValue"> & {
defaultValue?: Role;
value?: Role;
};
export const RoleSelect = forwardRef<HTMLSelectElement, RoleSelectProps>((props, ref) => {
return (
<Select {...props} ref={ref}>
{roles.map((role) => (
<option value={role} key={role}>
{role}
</option>
))}
</Select>
);
});
RoleSelect.displayName = "RoleSelect";
website/src/components/SideMenuLayout.tsx
+3 -3
View File
@@ -12,11 +12,11 @@ export const SideMenuLayout = (props: SideMenuLayoutProps) => {
return (
<Box backgroundColor={colorMode === "light" ? colors.light.bg : colors.dark.bg} className="sm:overflow-hidden">
<Box className="sm:flex h-full lg:gap-6">
<Box className="p-3 lg:p-6 lg:pr-0">
<Box display="flex" flexDirection={["column", "row"]} h="full" gap={["0", "0", "0", "6"]}>
<Box p={["3", "3", "3", "6"]} pr={["3", "3", "3", "0"]}>
<SideMenu buttonOptions={props.menuButtonOptions} />
</Box>
<Box className="flex flex-col overflow-y-auto p-3 lg:p-6 lg:pl-0 gap-14 w-full">{props.children}</Box>
<Box className="overflow-y-auto p-3 lg:p-6 lg:pl-1 w-full">{props.children}</Box>
</Box>
</Box>
);
website/src/components/Survey/LabelRadioGroup.tsx
+49 -4
View File
@@ -1,4 +1,18 @@
import { Box, Button, Flex, useColorMode } from "@chakra-ui/react";
import {
Box,
Button,
Flex,
IconButton,
Popover,
PopoverArrow,
PopoverBody,
PopoverCloseButton,
PopoverContent,
PopoverTrigger,
Text,
useColorMode,
} from "@chakra-ui/react";
import { InformationCircleIcon } from "@heroicons/react/20/solid";
import { useId, useState } from "react";
import { colors } from "src/styles/Theme/colors";
@@ -8,6 +22,17 @@ interface LabelRadioGroupProps {
isEditable?: boolean;
}
const label_messages: { [label: string]: { description: string; explanation: string[] } } = {
spam: {
description: "Is the message spam?",
explanation: [
'We consider the following unwanted content as spam: trolling, intentional undermining of our purpose, illegal material, material that violates our code of conduct, and other things that are inappropriate for our dataset. We collect these under the common heading of "spam".',
"This is not an assessment of whether this message is the best possible answer. Especially for prompts or user-replies, we very much want to retain all kinds of responses in the dataset, so that the assistant can learn to reply appropriately.",
"Please mark this text as spam only if it is clearly unsuited to be part of our dataset, as outlined above, and try not to make any subjective value-judgments beyond that.",
],
},
};
export const LabelRadioGroup = (props: LabelRadioGroupProps) => {
const [labelValues, setLabelValues] = useState<number[]>(Array.from({ length: props.labelIDs.length }).map(() => 0));
const [interactionFlag, setInteractionFlag] = useState(false);
@@ -17,7 +42,7 @@ export const LabelRadioGroup = (props: LabelRadioGroupProps) => {
{props.labelIDs.map((labelId, idx) => (
<LabelRadioItem
key={idx}
labelId={labelId}
labelText={label_messages[labelId] || { description: labelId }}
labelValue={labelValues[idx]}
clickHandler={(newValue) => {
const newState = labelValues.slice();
@@ -45,7 +70,7 @@ interface ButtonState {
}
interface LabelRadioItemProps {
labelId: string;
labelText: { description: string; explanation?: string[] };
labelValue: number;
clickHandler: (newVal: number) => unknown;
states: ButtonState[];
@@ -63,7 +88,27 @@ const LabelRadioItem = (props: LabelRadioItemProps) => {
<Box data-cy="label-group-item" data-label-type="radio">
<label className="text-sm" htmlFor={id}>
{/* TODO: display real text instead of just the id */}
<span className={labelTextClass}>{props.labelId}</span>
<span className={labelTextClass}>{props.labelText.description}</span>
{props.labelText.explanation ? (
<Popover>
<PopoverTrigger>
<IconButton
aria-label="explanation"
variant="link"
icon={<InformationCircleIcon className="h-5 w-5" />}
></IconButton>
</PopoverTrigger>
<PopoverContent>
<PopoverArrow />
<PopoverCloseButton />
<PopoverBody>
{props.labelText.explanation.map((paragraph, idx) => (
<Text key={idx}>{paragraph}</Text>
))}
</PopoverBody>
</PopoverContent>
</Popover>
) : null}
</label>
<Flex direction="row" gap={6} justify="center">
{props.states.map((item, idx) => (
website/src/components/Survey/LabelSliderGroup.tsx
+2 -2
View File
@@ -1,6 +1,6 @@
import { Grid, Slider, SliderFilledTrack, SliderThumb, SliderTrack, useColorMode } from "@chakra-ui/react";
import { useId, useState } from "react";
import { colors } from "styles/Theme/colors";
import { colors } from "src/styles/Theme/colors";
// TODO: consolidate with FlaggableElement
interface LabelSliderGroupProps {
@@ -59,8 +59,8 @@ function CheckboxSliderItem(props: {
>
<SliderTrack>
<SliderFilledTrack />
<SliderThumb />
</SliderTrack>
<SliderThumb bg="gainsboro" />
</Slider>
</>
);
website/src/components/Tasks/CreateTask.tsx
+2 -9
View File
@@ -4,6 +4,7 @@ import { MessageTable } from "src/components/Messages/MessageTable";
import { TrackedTextarea } from "src/components/Survey/TrackedTextarea";
import { TwoColumnsWithCards } from "src/components/Survey/TwoColumnsWithCards";
import { TaskSurveyProps } from "src/components/Tasks/Task";
import { TaskHeader } from "src/components/Tasks/TaskHeader";
export const CreateTask = ({
task,
@@ -14,7 +15,6 @@ export const CreateTask = ({
}: TaskSurveyProps<{ text: string }>) => {
const cardColor = useColorModeValue("gray.50", "gray.800");
const titleColor = useColorModeValue("gray.800", "gray.300");
const labelColor = useColorModeValue("gray.600", "gray.400");
const [inputText, setInputText] = useState("");
const textChangeHandler = (event: React.ChangeEvent<HTMLTextAreaElement>) => {
@@ -33,14 +33,7 @@ export const CreateTask = ({
<div data-cy="task" data-task-type="create-task">
<TwoColumnsWithCards>
<>
<Stack spacing="1">
<Text fontSize="xl" fontWeight="bold" color={titleColor}>
{taskType.label}
</Text>
<Text fontSize="md" color={labelColor}>
{taskType.overview}
</Text>
</Stack>
<TaskHeader taskType={taskType} />
{task.conversation ? (
<Box mt="4" borderRadius="lg" bg={cardColor} className="p-3 sm:p-6">
<MessageTable messages={task.conversation.messages} />
website/src/components/Tasks/EvaluateTask.tsx
+4 -11
View File
@@ -1,19 +1,19 @@
import { Box, Stack, Text, useColorModeValue } from "@chakra-ui/react";
import { Box, useColorModeValue } from "@chakra-ui/react";
import { useEffect } from "react";
import { MessageTable } from "src/components/Messages/MessageTable";
import { Sortable } from "src/components/Sortable/Sortable";
import { SurveyCard } from "src/components/Survey/SurveyCard";
import { TaskSurveyProps } from "src/components/Tasks/Task";
import { TaskHeader } from "src/components/Tasks/TaskHeader";
export const EvaluateTask = ({
task,
taskType,
isEditable,
isDisabled,
onReplyChanged,
}: TaskSurveyProps<{ ranking: number[] }>) => {
const cardColor = useColorModeValue("gray.50", "gray.800");
const titleColor = useColorModeValue("gray.800", "gray.300");
const labelColor = useColorModeValue("gray.600", "gray.400");
let messages = [];
if (task.conversation) {
@@ -36,14 +36,7 @@ export const EvaluateTask = ({
<div data-cy="task" data-task-type="evaluate-task">
<Box mb="4">
<SurveyCard>
<Stack spacing="1">
<Text fontSize="xl" fontWeight="bold" color={titleColor}>
Instructions
</Text>
<Text fontSize="md" color={labelColor}>
Given the following {sortables}, sort them from best to worst, best being first, worst being last.
</Text>
</Stack>
<TaskHeader taskType={taskType} />
<Box mt="4" p="6" borderRadius="lg" bg={cardColor}>
<MessageTable messages={messages} />
</Box>
website/src/components/Tasks/LabelTask.tsx → website/src/components/Tasks/LabelTask/LabelTask.tsx
+8 -13
View File
@@ -1,5 +1,4 @@
import { Box } from "@chakra-ui/react";
import { Text, useColorModeValue } from "@chakra-ui/react";
import { Box, useColorModeValue } from "@chakra-ui/react";
import { useEffect, useState } from "react";
import { MessageView } from "src/components/Messages";
import { MessageTable } from "src/components/Messages/MessageTable";
@@ -7,6 +6,7 @@ import { LabelRadioGroup } from "src/components/Survey/LabelRadioGroup";
import { LabelSliderGroup } from "src/components/Survey/LabelSliderGroup";
import { TwoColumnsWithCards } from "src/components/Survey/TwoColumnsWithCards";
import { TaskSurveyProps } from "src/components/Tasks/Task";
import { TaskHeader } from "src/components/Tasks/TaskHeader";
import { TaskType } from "src/types/Task";
export const LabelTask = ({
@@ -19,7 +19,10 @@ export const LabelTask = ({
const [sliderValues, setSliderValues] = useState<number[]>(new Array(valid_labels.length).fill(0));
useEffect(() => {
onReplyChanged({ content: { labels: {}, text: task.reply, message_id: task.message_id }, state: "DEFAULT" });
onReplyChanged({
content: { labels: {}, text: task.reply, message_id: task.message_id },
state: "NOT_SUBMITTABLE",
});
}, [task, onReplyChanged]);
const onSliderChange = (values: number[]) => {
@@ -33,20 +36,12 @@ export const LabelTask = ({
};
const cardColor = useColorModeValue("gray.50", "gray.800");
const titleColor = useColorModeValue("gray.800", "gray.300");
const labelColor = useColorModeValue("gray.600", "gray.400");
return (
<div data-cy="task" data-task-type="label-task">
<TwoColumnsWithCards>
<>
<Text fontSize="xl" fontWeight="bold" color={titleColor}>
{taskType.label}
</Text>
<Text fontSize="md" color={labelColor}>
{taskType.overview}
</Text>
<TaskHeader taskType={taskType} />
{task.conversation ? (
<Box mt="4" p="6" borderRadius="lg" bg={cardColor}>
<MessageTable
@@ -66,7 +61,7 @@ export const LabelTask = ({
</Box>
)}
</>
{valid_labels.length === 1 ? (
{task.mode === "simple" ? (
<LabelRadioGroup labelIDs={task.valid_labels} isEditable={isEditable} onChange={onSliderChange} />
) : (
<LabelSliderGroup labelIDs={task.valid_labels} isEditable={isEditable} onChange={onSliderChange} />
website/src/components/Tasks/LabelTask/index.tsx
+1
View File
@@ -0,0 +1 @@
export * from "./LabelTask";
website/src/components/Tasks/Task/Task.stories.tsx
+32
View File
@@ -0,0 +1,32 @@
import React from "react";
import { Task } from "./Task";
export default {
title: "tasks/Task",
component: Task,
};
const Template = ({ frontendId, task, trigger, mutate }) => {
return <Task frontendId={frontendId} task={task} trigger={trigger} mutate={mutate} />;
};
export const Default = Template.bind({});
Default.args = {
frontendId: "1234",
task: {
conversation: [],
id: "1234-4321",
mandatory_labels: ["spam"],
message_id: "772f843e-f740-4aad-a44f-e3cf0260692c",
reply: "1231231231",
type: "label_prompter_reply",
valid_labels: ["spam", "fails_task"],
},
trigger: (id, update_type, content) => {
console.log(content);
},
mutate: () => {
console.log("mutate");
},
};
website/src/components/Tasks/Task.tsx → website/src/components/Tasks/Task/Task.tsx
+12 -2
View File
@@ -27,7 +27,9 @@ export const Task = ({ frontendId, task, trigger, mutate }) => {
const replyContent = useRef<TaskContent>(null);
const [showUnchangedWarning, setShowUnchangedWarning] = useState(false);
const taskType = TaskTypes.find((taskType) => taskType.type === task.type);
const rootEl = useRef<HTMLDivElement>(null);
const taskType = TaskTypes.find((taskType) => taskType.type === task.type && taskType.mode === task.mode);
const { trigger: sendRejection } = useSWRMutation("/api/reject_task", post, {
onSuccess: async () => {
@@ -89,6 +91,7 @@ export const Task = ({ frontendId, task, trigger, mutate }) => {
content: replyContent.current,
});
setTaskStatus("SUBMITTED");
scrollToTop(rootEl.current);
break;
}
default:
@@ -138,7 +141,7 @@ export const Task = ({ frontendId, task, trigger, mutate }) => {
}
return (
<div>
<div ref={rootEl}>
{taskTypeComponent()}
<TaskControls
task={task}
@@ -164,3 +167,10 @@ export const Task = ({ frontendId, task, trigger, mutate }) => {
</div>
);
};
const scrollToTop = (element: HTMLElement) => {
while (element) {
element.scrollTop = 0;
element = element.parentElement;
}
};
website/src/components/Tasks/Task/index.tsx
+1
View File
@@ -0,0 +1 @@
export * from "./Task";
website/src/components/Tasks/TaskHeader/TaskHeader.tsx
+35
View File
@@ -0,0 +1,35 @@
import { HStack, IconButton, Link, Stack, Text, useColorModeValue } from "@chakra-ui/react";
import { FiHelpCircle } from "react-icons/fi";
import type { TaskInfo } from "src/components/Tasks/TaskTypes";
interface TaskHeaderProps {
/**
* The `TaskInfo` representing how we present the task to a user.
*/
taskType: TaskInfo;
}
/**
* Presents the Task label, instructions, and help link
*/
const TaskHeader = ({ taskType }: TaskHeaderProps) => {
const labelColor = useColorModeValue("gray.600", "gray.400");
const titleColor = useColorModeValue("gray.800", "gray.300");
return (
<Stack spacing="1">
<HStack>
<Text fontSize="xl" fontWeight="bold" color={titleColor}>
{taskType.label}
</Text>
<Link href={taskType.help_link} isExternal>
<IconButton variant="ghost" aria-label="More Information" icon={<FiHelpCircle />} />
</Link>
</HStack>
<Text fontSize="md" color={labelColor}>
{taskType.overview}
</Text>
</Stack>
);
};
export { TaskHeader };
website/src/components/Tasks/TaskHeader/index.tsx
+1
View File
@@ -0,0 +1 @@
export * from "./TaskHeader";
website/src/components/Tasks/TaskTypes.tsx
+55 -2
View File
@@ -11,6 +11,8 @@ export interface TaskInfo {
category: TaskCategory;
pathname: string;
type: string;
help_link: string;
mode?: string;
overview?: string;
instruction?: string;
update_type: string;
@@ -25,6 +27,7 @@ export const TaskTypes: TaskInfo[] = [
desc: "Help us improve Open Assistant by starting a random task.",
category: TaskCategory.Tasks,
pathname: "/tasks/random",
help_link: "https://projects.laion.ai/Open-Assistant/docs/guides/prompting",
type: "random",
update_type: "random",
},
@@ -34,6 +37,7 @@ export const TaskTypes: TaskInfo[] = [
desc: "Write initial prompts to help Open Assistant to try replying to diverse messages.",
category: TaskCategory.Create,
pathname: "/create/initial_prompt",
help_link: "https://projects.laion.ai/Open-Assistant/docs/guides/prompting",
type: "initial_prompt",
overview: "Create an initial message to send to the assistant",
instruction: "Provide the initial prompt",
@@ -44,6 +48,7 @@ export const TaskTypes: TaskInfo[] = [
desc: "Chat with Open Assistant and help improve it’s responses as you interact with it.",
category: TaskCategory.Create,
pathname: "/create/user_reply",
help_link: "https://projects.laion.ai/Open-Assistant/docs/tasks/reply_as_user",
type: "prompter_reply",
overview: "Given the following conversation, provide an adequate reply",
instruction: "Provide the user's reply",
@@ -54,6 +59,7 @@ export const TaskTypes: TaskInfo[] = [
desc: "Help Open Assistant improve its responses to conversations with other users.",
category: TaskCategory.Create,
pathname: "/create/assistant_reply",
help_link: "https://projects.laion.ai/Open-Assistant/docs/tasks/reply_as_assistant",
type: "assistant_reply",
overview: "Given the following conversation, provide an adequate reply",
instruction: "Provide the assistant's reply",
@@ -65,6 +71,8 @@ export const TaskTypes: TaskInfo[] = [
category: TaskCategory.Evaluate,
desc: "Help Open Assistant improve its responses to conversations with other users.",
pathname: "/evaluate/rank_user_replies",
help_link: "https://projects.laion.ai/Open-Assistant/docs/guides/prompting",
overview: "Given the following User replies, sort them from best to worst, best being first, worst being last.",
type: "rank_prompter_replies",
update_type: "message_ranking",
unchanged_title: "Order Unchanged",
@@ -75,6 +83,9 @@ export const TaskTypes: TaskInfo[] = [
desc: "Score prompts given by Open Assistant based on their accuracy and readability.",
category: TaskCategory.Evaluate,
pathname: "/evaluate/rank_assistant_replies",
help_link: "https://projects.laion.ai/Open-Assistant/docs/guides/prompting",
overview:
"Given the following Assistant replies, sort them from best to worst, best being first, worst being last.",
type: "rank_assistant_replies",
update_type: "message_ranking",
unchanged_title: "Order Unchanged",
@@ -85,19 +96,23 @@ export const TaskTypes: TaskInfo[] = [
desc: "Score prompts given by Open Assistant based on their accuracy and readability.",
category: TaskCategory.Evaluate,
pathname: "/evaluate/rank_initial_prompts",
help_link: "https://projects.laion.ai/Open-Assistant/docs/guides/prompting",
overview: "Given the following inital prompts, sort them from best to worst, best being first, worst being last.",
type: "rank_initial_prompts",
update_type: "message_ranking",
unchanged_title: "Order Unchanged",
unchanged_message: "You have not changed the order of the prompts. Are you sure you would like to continue?",
},
// label
// label (full)
{
label: "Label Initial Prompt",
desc: "Provide labels for a prompt.",
category: TaskCategory.Label,
pathname: "/label/label_initial_prompt",
help_link: "https://projects.laion.ai/Open-Assistant/docs/guides/prompting",
overview: "Provide labels for the following prompt",
type: "label_initial_prompt",
mode: "full",
update_type: "text_labels",
},
{
@@ -105,8 +120,10 @@ export const TaskTypes: TaskInfo[] = [
desc: "Provide labels for a prompt.",
category: TaskCategory.Label,
pathname: "/label/label_prompter_reply",
overview: "Given the following discussion, provide labels for the final promp",
help_link: "https://projects.laion.ai/Open-Assistant/docs/tasks/label_prompter_reply",
overview: "Given the following discussion, provide labels for the final prompt",
type: "label_prompter_reply",
mode: "full",
update_type: "text_labels",
},
{
@@ -114,8 +131,44 @@ export const TaskTypes: TaskInfo[] = [
desc: "Provide labels for a prompt.",
category: TaskCategory.Label,
pathname: "/label/label_assistant_reply",
help_link: "https://projects.laion.ai/Open-Assistant/docs/tasks/label_assistant_reply",
overview: "Given the following discussion, provide labels for the final prompt.",
type: "label_assistant_reply",
mode: "full",
update_type: "text_labels",
},
// label (simple)
{
label: "Classify Initial Prompt",
desc: "Provide labels for a prompt.",
category: TaskCategory.Label,
pathname: "/label/label_initial_prompt",
help_link: "https://projects.laion.ai/Open-Assistant/docs/guides/prompting",
overview: "Read the following prompt and then answer the question about it.",
type: "label_initial_prompt",
mode: "simple",
update_type: "text_labels",
},
{
label: "Classify Prompter Reply",
desc: "Provide labels for a prompt.",
category: TaskCategory.Label,
pathname: "/label/label_prompter_reply",
help_link: "https://projects.laion.ai/Open-Assistant/docs/guides/prompting",
overview: "Read the following conversation and then answer the question about the last prompt in the discussion.",
type: "label_prompter_reply",
mode: "simple",
update_type: "text_labels",
},
{
label: "Classify Assistant Reply",
desc: "Provide labels for a prompt.",
category: TaskCategory.Label,
pathname: "/label/label_assistant_reply",
help_link: "https://projects.laion.ai/Open-Assistant/docs/guides/prompting",
overview: "Read the following conversation and then answer the question about the last prompt in the discussion.",
type: "label_assistant_reply",
mode: "simple",
update_type: "text_labels",
},
];
website/src/lib/auth.ts
+3 -2
View File
@@ -1,11 +1,12 @@
import type { NextApiRequest, NextApiResponse } from "next";
import { getToken, JWT } from "next-auth/jwt";
import { Role } from "src/components/RoleSelect";
/**
* Wraps any API Route handler and verifies that the user does not have the
* specified role. Returns a 403 if they do, otherwise runs the handler.
*/
const withoutRole = (role: string, handler: (arg0: NextApiRequest, arg1: NextApiResponse, arg2: JWT) => void) => {
const withoutRole = (role: Role, handler: (arg0: NextApiRequest, arg1: NextApiResponse, arg2: JWT) => void) => {
return async (req: NextApiRequest, res: NextApiResponse) => {
const token = await getToken({ req });
if (!token || token.role === role) {
@@ -20,7 +21,7 @@ const withoutRole = (role: string, handler: (arg0: NextApiRequest, arg1: NextApi
* Wraps any API Route handler and verifies that the user has the appropriate
* role before running the handler. Returns a 403 otherwise.
*/
const withRole = (role: string, handler: (arg0: NextApiRequest, arg1: NextApiResponse) => void) => {
const withRole = (role: Role, handler: (arg0: NextApiRequest, arg1: NextApiResponse) => void) => {
return async (req: NextApiRequest, res: NextApiResponse) => {
const token = await getToken({ req });
if (!token || token.role !== role) {
website/src/lib/oasst_api_client.ts
+10 -4
View File
@@ -16,7 +16,13 @@ export class OasstError {
}
export class OasstApiClient {
constructor(private readonly oasstApiUrl: string, private readonly oasstApiKey: string) {}
oasstApiUrl: string;
oasstApiKey: string;
constructor(oasstApiUrl: string, oasstApiKey: string) {
this.oasstApiUrl = oasstApiUrl;
this.oasstApiKey = oasstApiKey;
}
private async post(path: string, body: any): Promise<any> {
const resp = await fetch(`${this.oasstApiUrl}${path}`, {
@@ -107,7 +113,7 @@ export class OasstApiClient {
type: taskType,
user: {
id: userToken.sub,
display_name: userToken.name || userToken.email,
display_name: userToken.name,
auth_method: "local",
},
});
@@ -140,7 +146,7 @@ export class OasstApiClient {
type: updateType,
user: {
id: userToken.sub,
display_name: userToken.name || userToken.email,
display_name: userToken.name,
auth_method: "local",
},
task_id: taskId,
@@ -170,7 +176,7 @@ export class OasstApiClient {
const params = new URLSearchParams();
params.append("max_count", max_count.toString());
// The backend API uses different query paramters depending on the
// The backend API uses different query parameters depending on the
// pagination direction but they both take the same cursor value.
// Depending on direction, pick the right query param.
if (cursor !== "") {
website/src/pages/404.tsx
+3 -7
View File
@@ -1,14 +1,10 @@
import { Box, Button, Center, Link, Text, useColorModeValue } from "@chakra-ui/react";
import { Box, Button, Center, Link, Text } from "@chakra-ui/react";
import Head from "next/head";
import { useRouter } from "next/router";
import { FiAlertTriangle } from "react-icons/fi";
import { PageEmptyState } from "src/components/EmptyState";
import { EmptyState } from "src/components/EmptyState";
import { getTransparentHeaderLayout } from "src/components/Layout";
function Error() {
const router = useRouter();
const backgroundColor = useColorModeValue("white", "gray.800");
return (
<>
<Head>
@@ -16,7 +12,7 @@ function Error() {
<meta name="404" content="Sorry, this page doesn't exist." />
</Head>
<Center flexDirection="column" gap="4" fontSize="lg" className="subpixel-antialiased">
<PageEmptyState />
<EmptyState text="Sorry, the page you are looking for does not exist." icon={FiAlertTriangle} />
<Box display="flex" flexDirection="column" alignItems="center" gap="2" mt="6">
<Text fontSize="sm">If you were trying to contribute data but ended up here, please file a bug.</Text>
<Button
website/src/pages/500.tsx
+25 -11
View File
@@ -1,32 +1,46 @@
import { Button, Link, Stack } from "@chakra-ui/react";
import { Box, Button, Center, Link, Text } from "@chakra-ui/react";
import Head from "next/head";
import NextLink from "next/link";
import { FiAlertTriangle } from "react-icons/fi";
import { EmptyState } from "src/components/EmptyState";
import { getTransparentHeaderLayout } from "src/components/Layout";
export default function Error() {
function ServerError() {
return (
<>
<Head>
<title>500 - Open Assistant</title>
<meta name="404" content="Sorry, this page doesn't exist." />
</Head>
<main className="flex h-3/4 items-center justify-center overflow-hidden subpixel-antialiased text-xl">
<Stack>
<p>Sorry, We encountered a server error. We&apos;re not sure what went wrong</p>
<p>Please file a but below and describe what you were trying to accomplish</p>
<Button leftIcon={<FiAlertTriangle className="text-blue-500" aria-hidden="true" />} variant="solid">
<Center flexDirection="column" gap="4" fontSize="lg" className="subpixel-antialiased">
<EmptyState
text="Sorry, we encountered a server error. We're not sure what went wrong."
icon={FiAlertTriangle}
/>
<Box display="flex" flexDirection="column" alignItems="center" gap="2" mt="6">
<Text fontSize="sm">If you were trying to contribute data but ended up here, please file a bug.</Text>
<Button
width="fit-content"
leftIcon={<FiAlertTriangle className="text-blue-500" aria-hidden="true" />}
variant="solid"
size="xs"
>
<Link
as={NextLink}
key="Report a Bug"
href="https://github.com/LAION-AI/Open-Assistant/issues/new/choose"
aria-label="Report a Bug"
className="flex items-center"
_hover={{ textDecoration: "none" }}
isExternal
>
Report a Bug
</Link>
</Button>
</Stack>
</main>
</Box>
</Center>
</>
);
}
ServerError.getLayout = getTransparentHeaderLayout;
export default ServerError;
website/src/pages/_app.tsx
+12 -2
View File
@@ -3,16 +3,24 @@ import "focus-visible";
import type { AppProps } from "next/app";
import { SessionProvider } from "next-auth/react";
import { appWithTranslation } from "next-i18next";
import { FlagsProvider } from "react-feature-flags";
import { getDefaultLayout, NextPageWithLayout } from "src/components/Layout";
import flags from "src/flags";
import { SWRConfig, SWRConfiguration } from "swr";
import nextI18NextConfig from "../../next-i18next.config.js";
import { Chakra, getServerSideProps } from "../styles/Chakra";
type AppPropsWithLayout = AppProps & {
Component: NextPageWithLayout;
};
const swrConfig: SWRConfiguration = {
revalidateOnFocus: false,
revalidateOnMount: true,
};
function MyApp({ Component, pageProps: { session, cookies, ...pageProps } }: AppPropsWithLayout) {
const getLayout = Component.getLayout ?? getDefaultLayout;
const page = getLayout(<Component {...pageProps} />);
@@ -20,10 +28,12 @@ function MyApp({ Component, pageProps: { session, cookies, ...pageProps } }: App
return (
<FlagsProvider value={flags}>
<Chakra cookies={cookies}>
<SessionProvider session={session}>{page}</SessionProvider>
<SWRConfig value={swrConfig}>
<SessionProvider session={session}>{page}</SessionProvider>
</SWRConfig>
</Chakra>
</FlagsProvider>
);
}
export { getServerSideProps };
export default MyApp;
export default appWithTranslation(MyApp, nextI18NextConfig);

Some files were not shown because too many files have changed in this diff Show More