mirror of
https://github.com/wassname/HumanAgent-MCP.git
synced 2026-08-20 12:00:30 +08:00
Fix tool call timeout by implementing chunked response with keepalive
- Added special handling for HumanAgent_Chat tool calls to prevent Node.js undici 5-minute timeout - Send HTTP headers immediately with Transfer-Encoding: chunked to stop headersTimeout - Implement 4-minute keepalive mechanism writing space character to reset bodyTimeout - Allows indefinite wait times for human responses in tool calls - Tested successfully with 35+ minute wait time without timeout Resolves issue where VS Code MCP client would cancel tool calls after ~5-6 minutes with error 'The user cancelled the tool call' due to undici hardcoded timeouts.
This commit is contained in:
B Harper
@@ -0,0 +1,182 @@
|
||||
# Taking Back Control: Why Human-in-the-Loop AI Beats Fully Autonomous Agents
|
||||
|
||||
*How HumanAgent MCP transforms AI coding from guesswork to conversation*
|
||||
|
||||
|
||||
|
||||
# The Frustration We All Know
|
||||
|
||||
Picture this: You ask your AI coding assistant a simple question expecting some discussion and instead of answering or planning with you it confidently rewrites half your codebase, breaks your tests, and introduces subtle bugs that take hours to debug. Generally starting with the dreaded "Sure Let me..."
|
||||
|
||||
Sound familiar?
|
||||
|
||||
The promise of AI coding assistants was supposed to be increased productivity and reduced cognitive load. Instead, many developers find themselves in a constant battle:
|
||||
|
||||
AI agents making decisions and changes without asking based on outdated training data or just a whim.
|
||||
Trying so hard to please that a simple request for discussion results in 15 minutes of "Let Me, no wait, actually if i try this, no that didnt work" and never undoing any of the half baked ideas that are now littered around in code.
|
||||
|
||||
We're stuck between two extremes: passive suggestion tools that can't engage, or overly aggressive agents that won't stop to ask. There's no middle ground—until now.
|
||||
|
||||
Keep reading till the end - there is a free ready made solution for you detailed in this article.
|
||||
|
||||
# Why AI Assistants Keep Guessing Wrong
|
||||
|
||||
The fundamental problem isn't the AI's intelligence—it's the communication model.
|
||||
Worse still is VSCode's Copilot Chat's built-in default prompt which is so strongly and poorly worded that it forces the agent to spare no effort and not accept defeat or bother you with any discussion before coming back with a complete solution that you did not ask for.
|
||||
|
||||
Following are some real and very problematic statements taken directly from GitHub Copilot Chat's main system prompt:
|
||||
* "implement changes rather than only suggesting them."
|
||||
* "If the user’s intent is unclear, infer the most useful likely action and proceed"
|
||||
* “Continue working until the user’s request is completely resolved before ending your turn and yielding back to the user. Only terminate your turn when you are certain the task is complete.”
|
||||
|
||||
|
||||
When AI can't ask questions, it fills the gaps with assumptions:
|
||||
- Should I use TypeScript or JavaScript? *Guess.*
|
||||
- Should I optimize for speed or readability? *Guess.*
|
||||
- Should I update the tests too? *Guess.*
|
||||
|
||||
These guesses waste expensive API calls, generate incorrect code, and burn your time debugging. More importantly, they break the collaborative flow that makes human pair programming so effective.
|
||||
|
||||
# Enter Human-in-the-Loop: The Best of Both Worlds
|
||||
|
||||
What if your AI assistant could:
|
||||
- Ask clarifying questions before making changes?
|
||||
- Discuss approach options and get your input?
|
||||
- Confirm destructive actions before executing them?
|
||||
- Have real conversations instead of one-way commands?
|
||||
|
||||
This is the human-in-the-loop approach—and it's where AI tooling is heading.
|
||||
|
||||
# The Model Context Protocol (MCP) Changes Everything
|
||||
|
||||
Anthropic's [Model Context Protocol](https://modelcontextprotocol.io/) provides the infrastructure for this new paradigm. MCP allows AI models to interact with external tools and services in a standardized way—including asking humans for input.
|
||||
|
||||
Instead of AI making isolated decisions, MCP enables:
|
||||
- Bidirectional communication between AI and humans
|
||||
- Real-time interaction without leaving your development environment
|
||||
- Rich context sharing that keeps everyone on the same page
|
||||
|
||||
# Introducing HumanAgent MCP
|
||||
|
||||
HumanAgent MCP is a VS Code extension that brings true human-in-the-loop AI to your development workflow. Built on the Model Context Protocol, it creates a persistent chat interface where your AI assistant can ask questions, discuss options, and get real-time guidance.
|
||||
|
||||
# How It Works
|
||||
|
||||
Install the HumanAgent MCP server - no configuration required. [Install HumanAgent MCP](vscode:extension/3DTek-xyz.humanagent-mcp)
|
||||
Start your agent discussion in the usual manner with something resembling "Lets chat via Human Agent Tool".
|
||||
|
||||
|
||||
What happens next:
|
||||
1. A chat panel opens in your VS Code sidebar
|
||||
2. The AI's question appears with context
|
||||
3. You respond directly in the chat (or use quick reply buttons)
|
||||
4. The AI receives your answer and continues with clarity
|
||||
|
||||
None of the back and forth interaction with human agent chat are treated as premium requests $$$!
|
||||
Yes thats correct - you can run an entire days planning & coding session on 1 premium request.
|
||||
|
||||
The agent is encouraged to keep the discussion up until a plan is reached and its time to code. Its also encouraged to come back with any extra questions it needs answered.
|
||||
An additional prompt can be setup to be included in your every response such as "Remember to read the copilot-instructions file regularly and dont iterate and change your direction - if its not working come back and we can make a new plan"
|
||||
|
||||
# Key Features
|
||||
|
||||
* Native VS Code Integration + Web Interface
|
||||
- Dockable chat panel that stays visible while coding
|
||||
- Multiple session support for parallel AI in multiple vscode workspaces
|
||||
- Web interface option for monitoring and interacting with all running VSCode workspaces. Manage many VSCode agents working on multiple projects from one place
|
||||
|
||||
* Quick Replies
|
||||
- Pre-defined responses for common questions
|
||||
- Customizable via workspace settings
|
||||
- "Yes Please Proceed" / "Explain in more detail please" by default
|
||||
|
||||
* Smart Notifications
|
||||
- Tab indicators show which sessions need attention
|
||||
- Visual cues for pending AI questions
|
||||
- Sound notifications (optional)
|
||||
|
||||
* Flexible Configuration
|
||||
- Per-workspace tool customization
|
||||
- Override default behavior for specific projects
|
||||
- Session naming for easy tracking
|
||||
|
||||
|
||||
|
||||
# The Benefits: More Than Just Better Code
|
||||
|
||||
Reduced Costs
|
||||
Stop wasting expensive API calls on AI generating code based on wrong assumptions. One quick question saves thousands of tokens of regeneration.
|
||||
|
||||
Fewer Errors
|
||||
Clarification before action means less incorrect code, fewer bugs, less debugging time and less orphaned AI code littered around.
|
||||
|
||||
Faster Development Cycles
|
||||
A 10-second clarification beats an hour of debugging wrong guesses. Keep your flow state intact.
|
||||
|
||||
Better Collaboration
|
||||
Turns one-way instructions into dialogue. You stay in control while the AI handles the grunt work.
|
||||
|
||||
|
||||
|
||||
# Getting Started
|
||||
|
||||
# Installation
|
||||
|
||||
1. Install from VS Code Marketplace: [Install HumanAgent MCP](vscode:extension/3DTek-xyz.humanagent-mcp)
|
||||
2. The MCP server starts automatically on extension activation
|
||||
3. That's it - nothing further. No MCP server setup, MCP.json files. Just install and start working.
|
||||
|
||||
|
||||
|
||||
# Using with GitHub Copilot
|
||||
|
||||
GitHub Copilot in VS Code automatically discovers MCP servers. Once HumanAgent MCP is installed, Copilot can invoke the chat tool when it needs input.
|
||||
|
||||
Try asking Copilot: "Refactor my authentication system, but check with me before making any database schema changes"
|
||||
|
||||
# The Future is Collaborative, Not Autonomous
|
||||
|
||||
The vision of fully autonomous AI coding—where you describe what you want and AI builds it without supervision—is seductive but flawed. Software development isn't just about writing code; it's about:
|
||||
|
||||
- Making informed tradeoffs (speed vs quality, simple vs scalable)
|
||||
- Understanding business context (deadlines, user needs, technical debt)
|
||||
- Maintaining coherent architecture across teams and time
|
||||
- Learning and adapting from mistakes and feedback
|
||||
|
||||
These require human judgment, but they don't require humans to write every line of code.
|
||||
|
||||
Human-in-the-loop AI hits the sweet spot:
|
||||
- AI handles the repetitive, time-consuming coding work
|
||||
- Humans make the strategic decisions and provide domain knowledge
|
||||
- Conversation ensures alignment and catches errors early
|
||||
- Both parties learn and improve from the interaction
|
||||
|
||||
This isn't a compromise—it's the optimal collaboration model.
|
||||
|
||||
|
||||
# Join the Conversation
|
||||
|
||||
HumanAgent MCP is open source and actively developed. We're building features based on real developer feedback:
|
||||
|
||||
- GitHub: [3DTek-xyz/HumanAgent-MCP](https://github.com/3DTek-xyz/HumanAgent-MCP)
|
||||
- VS Code Marketplace: [HumanAgent MCP](https://marketplace.visualstudio.com/items?itemName=3DTek-xyz.humanagent-mcp)
|
||||
- Documentation: See the README for advanced configuration
|
||||
|
||||
# What's Next
|
||||
|
||||
We're working on:
|
||||
- Image support for pasting screenshots and diagrams
|
||||
- Multi-agent agentic orchestration for complex workflows. Multiple agents with narrow scope working together.
|
||||
- Team collaboration features for shared AI sessions
|
||||
|
||||
# Try It Today
|
||||
|
||||
Stop fighting with autonomous AI that guesses wrong. Stop context-switching to ChatGPT. Start having real conversations with your AI coding assistant.
|
||||
|
||||
Install HumanAgent MCP and experience what coding with AI should feel like—a true collaboration between human insight and machine productivity.
|
||||
|
||||
---
|
||||
|
||||
*HumanAgent MCP is free and open source. Built by developers, for developers who want AI to augment their skills—not replace their judgment.*
|
||||
|
||||
Tags: #AI #Development #VSCode #ModelContextProtocol #Copilot #HumanInTheLoop
|
||||
@@ -13,4 +13,8 @@ For more information see https://github.com/3DTek-xyz/HumanAgent-MCP/blob/master
|
||||
|
||||
|
||||
|
||||
*Complete demonstration of the HumanAgent MCP extension in action - showing real-time human-AI collaboration*
|
||||
*Complete demonstration of the HumanAgent MCP extension in action - showing real-time human-AI collaboration*
|
||||
|
||||
## Medium Article
|
||||
https://medium.com/@harperbenwilliam/stop-the-ai-chaos-why-human-in-the-loop-beats-fully-autonomous-coding-agents-eeb0ae17fde9
|
||||
|
||||
|
||||
@@ -691,3 +691,90 @@
|
||||
2025-11-18T02:22:23.278Z - RESPONSE ENDPOINT CALLED - RequestID: 4-1763432483320
|
||||
2025-11-18T02:26:48.410Z - RESPONSE ENDPOINT CALLED - RequestID: 3-1763432790349
|
||||
2025-11-18T02:27:14.305Z - RESPONSE ENDPOINT CALLED - RequestID: 4-1763432819507
|
||||
2025-11-18T02:39:58.506Z - RESPONSE ENDPOINT CALLED - RequestID: 4-1763433568966
|
||||
2025-11-18T02:53:30.480Z - RESPONSE ENDPOINT CALLED - RequestID: 3-1763434329608
|
||||
2025-11-18T02:56:15.460Z - RESPONSE ENDPOINT CALLED - RequestID: 4-1763434424942
|
||||
2025-11-18T03:38:45.925Z - RESPONSE ENDPOINT CALLED - RequestID: 5-1763434725081
|
||||
2025-11-18T03:40:19.015Z - RESPONSE ENDPOINT CALLED - RequestID: 3-1763437187183
|
||||
2025-11-18T03:41:59.378Z - RESPONSE ENDPOINT CALLED - RequestID: 4-1763437242001
|
||||
2025-11-18T03:43:01.827Z - RESPONSE ENDPOINT CALLED - RequestID: 5-1763437369466
|
||||
2025-11-18T06:14:00.728Z - RESPONSE ENDPOINT CALLED - RequestID: 6-1763437423019
|
||||
2025-11-18T06:18:02.912Z - RESPONSE ENDPOINT CALLED - RequestID: 3-1763446658929
|
||||
2025-11-18T06:23:47.631Z - RESPONSE ENDPOINT CALLED - RequestID: 3-1763447021849
|
||||
2025-11-18T06:25:57.558Z - RESPONSE ENDPOINT CALLED - RequestID: 4-1763447105247
|
||||
2025-11-18T06:27:25.839Z - RESPONSE ENDPOINT CALLED - RequestID: 5-1763447231081
|
||||
2025-11-18T06:32:46.263Z - RESPONSE ENDPOINT CALLED - RequestID: 6-1763447483088
|
||||
2025-11-18T06:34:23.905Z - RESPONSE ENDPOINT CALLED - RequestID: 7-1763447622207
|
||||
2025-11-18T06:35:14.021Z - RESPONSE ENDPOINT CALLED - RequestID: 8-1763447690581
|
||||
2025-11-18T06:36:31.582Z - RESPONSE ENDPOINT CALLED - RequestID: 9-1763447750484
|
||||
2025-11-18T06:37:44.404Z - RESPONSE ENDPOINT CALLED - RequestID: 10-1763447818173
|
||||
2025-11-18T06:38:38.546Z - RESPONSE ENDPOINT CALLED - RequestID: 11-1763447875648
|
||||
2025-11-18T06:39:48.141Z - RESPONSE ENDPOINT CALLED - RequestID: 12-1763447932022
|
||||
2025-11-18T06:43:40.511Z - RESPONSE ENDPOINT CALLED - RequestID: 13-1763448001173
|
||||
2025-11-18T06:45:17.457Z - RESPONSE ENDPOINT CALLED - RequestID: 14-1763448230522
|
||||
2025-11-18T06:46:27.529Z - RESPONSE ENDPOINT CALLED - RequestID: 15-1763448329021
|
||||
2025-11-18T06:48:28.407Z - RESPONSE ENDPOINT CALLED - RequestID: 16-1763448405443
|
||||
2025-11-18T06:50:29.838Z - RESPONSE ENDPOINT CALLED - RequestID: 17-1763448526213
|
||||
2025-11-18T06:52:12.050Z - RESPONSE ENDPOINT CALLED - RequestID: 18-1763448650631
|
||||
2025-11-19T01:24:09.277Z - RESPONSE ENDPOINT CALLED - RequestID: 3-1763515019650
|
||||
2025-11-19T01:25:33.021Z - RESPONSE ENDPOINT CALLED - RequestID: 3-1763515517453
|
||||
2025-11-19T01:26:25.579Z - RESPONSE ENDPOINT CALLED - RequestID: 4-1763515548791
|
||||
2025-11-19T01:27:28.483Z - RESPONSE ENDPOINT CALLED - RequestID: 5-1763515595531
|
||||
2025-11-19T01:30:18.613Z - RESPONSE ENDPOINT CALLED - RequestID: 6-1763515656709
|
||||
2025-11-19T01:32:02.270Z - RESPONSE ENDPOINT CALLED - RequestID: 7-1763515831756
|
||||
2025-11-19T01:35:55.654Z - RESPONSE ENDPOINT CALLED - RequestID: 8-1763515932210
|
||||
2025-11-19T01:36:23.176Z - RESPONSE ENDPOINT CALLED - RequestID: 9-1763516165498
|
||||
2025-11-19T01:37:00.835Z - RESPONSE ENDPOINT CALLED - RequestID: 10-1763516192706
|
||||
2025-11-19T01:39:10.905Z - RESPONSE ENDPOINT CALLED - RequestID: 11-1763516230904
|
||||
2025-11-19T01:39:49.004Z - RESPONSE ENDPOINT CALLED - RequestID: 12-1763516367404
|
||||
2025-11-19T01:40:17.739Z - RESPONSE ENDPOINT CALLED - RequestID: 13-1763516398238
|
||||
2025-11-19T01:41:36.756Z - RESPONSE ENDPOINT CALLED - RequestID: 14-1763516481876
|
||||
2025-11-19T01:42:38.294Z - RESPONSE ENDPOINT CALLED - RequestID: 15-1763516526387
|
||||
2025-11-19T01:45:01.737Z - RESPONSE ENDPOINT CALLED - RequestID: 16-1763516585390
|
||||
2025-11-19T01:45:35.148Z - RESPONSE ENDPOINT CALLED - RequestID: 17-1763516718714
|
||||
2025-11-19T01:46:43.539Z - RESPONSE ENDPOINT CALLED - RequestID: 18-1763516759168
|
||||
2025-11-19T01:47:16.740Z - RESPONSE ENDPOINT CALLED - RequestID: 19-1763516833761
|
||||
2025-11-19T01:51:18.447Z - RESPONSE ENDPOINT CALLED - RequestID: 20-1763517065160
|
||||
2025-11-19T01:53:24.021Z - RESPONSE ENDPOINT CALLED - RequestID: 21-1763517148133
|
||||
2025-11-19T01:55:22.395Z - RESPONSE ENDPOINT CALLED - RequestID: 22-1763517220485
|
||||
2025-11-19T01:55:56.343Z - RESPONSE ENDPOINT CALLED - RequestID: 23-1763517336062
|
||||
2025-11-19T02:01:39.996Z - RESPONSE ENDPOINT CALLED - RequestID: 24-1763517693982
|
||||
2025-11-19T02:02:56.332Z - RESPONSE ENDPOINT CALLED - RequestID: 25-1763517710116
|
||||
2025-11-19T02:03:13.157Z - RESPONSE ENDPOINT CALLED - RequestID: 26-1763517783066
|
||||
2025-11-19T02:05:42.415Z - RESPONSE ENDPOINT CALLED - RequestID: 27-1763517881460
|
||||
2025-11-19T02:10:16.438Z - RESPONSE ENDPOINT CALLED - RequestID: 28-1763517957632
|
||||
2025-11-19T02:11:54.688Z - RESPONSE ENDPOINT CALLED - RequestID: 29-1763518226131
|
||||
2025-11-19T02:12:39.655Z - RESPONSE ENDPOINT CALLED - RequestID: 30-1763518329883
|
||||
2025-11-19T02:15:30.961Z - RESPONSE ENDPOINT CALLED - RequestID: 31-1763518367194
|
||||
2025-11-19T02:16:41.402Z - RESPONSE ENDPOINT CALLED - RequestID: 32-1763518543112
|
||||
2025-11-19T02:17:43.201Z - RESPONSE ENDPOINT CALLED - RequestID: 33-1763518614727
|
||||
2025-11-19T02:19:21.697Z - RESPONSE ENDPOINT CALLED - RequestID: 34-1763518676927
|
||||
2025-11-19T02:20:18.802Z - RESPONSE ENDPOINT CALLED - RequestID: 35-1763518774294
|
||||
2025-11-19T02:43:24.964Z - RESPONSE ENDPOINT CALLED - RequestID: 3-1763520157655
|
||||
2025-11-19T02:45:01.206Z - RESPONSE ENDPOINT CALLED - RequestID: 4-1763520219515
|
||||
2025-11-19T02:46:19.104Z - RESPONSE ENDPOINT CALLED - RequestID: 5-1763520313847
|
||||
2025-11-19T02:57:37.343Z - RESPONSE ENDPOINT CALLED - RequestID: 3-1763521014023
|
||||
2025-11-19T02:59:52.681Z - RESPONSE ENDPOINT CALLED - RequestID: 4-1763521113125
|
||||
2025-11-19T03:02:14.607Z - RESPONSE ENDPOINT CALLED - RequestID: 5-1763521251826
|
||||
2025-11-19T03:03:54.164Z - RESPONSE ENDPOINT CALLED - RequestID: 6-1763521353229
|
||||
2025-11-19T03:06:19.281Z - RESPONSE ENDPOINT CALLED - RequestID: 7-1763521454858
|
||||
2025-11-19T03:07:37.298Z - RESPONSE ENDPOINT CALLED - RequestID: 8-1763521598807
|
||||
2025-11-19T03:11:39.988Z - RESPONSE ENDPOINT CALLED - RequestID: 9-1763521816638
|
||||
2025-11-19T03:13:32.720Z - RESPONSE ENDPOINT CALLED - RequestID: 10-1763521924945
|
||||
2025-11-19T03:15:58.163Z - RESPONSE ENDPOINT CALLED - RequestID: 11-1763522057738
|
||||
2025-11-19T03:17:35.340Z - RESPONSE ENDPOINT CALLED - RequestID: 12-1763522181228
|
||||
2025-11-19T03:18:55.097Z - RESPONSE ENDPOINT CALLED - RequestID: 13-1763522268824
|
||||
2025-11-19T03:19:49.257Z - RESPONSE ENDPOINT CALLED - RequestID: 14-1763522377448
|
||||
2025-11-19T03:21:18.464Z - RESPONSE ENDPOINT CALLED - RequestID: 15-1763522444120
|
||||
2025-11-19T03:23:03.579Z - RESPONSE ENDPOINT CALLED - RequestID: 16-1763522504241
|
||||
2025-11-19T03:23:50.363Z - RESPONSE ENDPOINT CALLED - RequestID: 17-1763522596596
|
||||
2025-11-19T03:25:51.736Z - RESPONSE ENDPOINT CALLED - RequestID: 18-1763522683032
|
||||
2025-11-19T03:36:07.974Z - RESPONSE ENDPOINT CALLED - RequestID: 3-1763523299064
|
||||
2025-11-19T03:37:27.924Z - RESPONSE ENDPOINT CALLED - RequestID: 4-1763523384331
|
||||
2025-11-19T03:38:54.188Z - RESPONSE ENDPOINT CALLED - RequestID: 5-1763523460761
|
||||
2025-11-19T03:45:59.295Z - RESPONSE ENDPOINT CALLED - RequestID: 3-1763523836543
|
||||
2025-11-19T03:46:34.788Z - RESPONSE ENDPOINT CALLED - RequestID: 4-1763523964130
|
||||
2025-11-19T03:52:41.873Z - RESPONSE ENDPOINT CALLED - RequestID: 3-1763524321397
|
||||
2025-11-19T04:29:21.690Z - RESPONSE ENDPOINT CALLED - RequestID: 4-1763524368377
|
||||
2025-11-19T04:29:57.436Z - RESPONSE ENDPOINT CALLED - RequestID: 5-1763526575920
|
||||
2025-11-19T04:32:45.235Z - RESPONSE ENDPOINT CALLED - RequestID: 3-1763526722087
|
||||
|
||||
@@ -636,6 +636,42 @@ export class McpServer extends EventEmitter {
|
||||
this.debugLogger.log('HTTP', `Added sessionId ${sessionId} to MCP message params`);
|
||||
}
|
||||
|
||||
// Special handling for HumanAgent_Chat tool to prevent undici timeout
|
||||
// This tool can wait indefinitely for human response, so we need to:
|
||||
// 1. Send HTTP headers immediately (stops undici's 5-minute headersTimeout)
|
||||
// 2. Send keepalive data every 4 minutes (resets undici's 5-minute bodyTimeout)
|
||||
if (message.method === 'tools/call' && message.params?.name === 'HumanAgent_Chat') {
|
||||
this.debugLogger.log('HTTP', 'HumanAgent_Chat detected - using streaming response to prevent timeout');
|
||||
|
||||
// Send headers immediately to stop headersTimeout
|
||||
res.statusCode = 200;
|
||||
res.setHeader('Content-Type', 'application/json');
|
||||
res.setHeader('Transfer-Encoding', 'chunked');
|
||||
this.debugLogger.log('HTTP', 'Sent immediate headers for HumanAgent_Chat');
|
||||
|
||||
// Start keepalive to reset bodyTimeout every 4 minutes
|
||||
// Note: We write a space character which is valid JSON whitespace and gets ignored
|
||||
const keepaliveInterval = setInterval(() => {
|
||||
if (!res.destroyed) {
|
||||
res.write(' '); // Write whitespace to reset bodyTimeout
|
||||
this.debugLogger.log('HTTP', 'Sent keepalive for HumanAgent_Chat');
|
||||
} else {
|
||||
clearInterval(keepaliveInterval);
|
||||
}
|
||||
}, 4 * 60 * 1000); // 4 minutes (undici timeout is 5 minutes)
|
||||
|
||||
// Wait for human response (this can take any amount of time now)
|
||||
const response = await this.handleMessage(message);
|
||||
|
||||
// Cleanup keepalive and send final response
|
||||
clearInterval(keepaliveInterval);
|
||||
this.debugLogger.log('HTTP', 'HumanAgent_Chat response received, sending to client');
|
||||
const responseJson = JSON.stringify(response);
|
||||
res.end(responseJson); // This sends the actual JSON response
|
||||
return;
|
||||
}
|
||||
|
||||
// Normal handling for all other tools/methods
|
||||
const response = await this.handleMessage(message);
|
||||
this.debugLogger.log('HTTP', 'Response from handleMessage:', response);
|
||||
|
||||
|
||||
Reference in New Issue
Block a user