Skip to main content

Knowledge Search

Use Knowledge Search to query FluxLoop documentation directly from VSCode without leaving your editor.

How to Use​

Method 1: Integration View Button​

  1. Open the FluxLoop activity bar
  2. Navigate to the Integration view
  3. Click Knowledge Search button at the top
  4. Enter your question in the input box

Method 2: Command Palette​

  1. Open Command Palette (Cmd+Shift+P or Ctrl+Shift+P)
  2. Type and select: FluxLoop: Open Knowledge Search
  3. Enter your question

What You Can Ask​

Knowledge Search is backed by the FluxLoop MCP server, which indexes:

  • Official FluxLoop documentation
  • SDK and CLI guides
  • Framework-specific recipes
  • Code examples and samples

Example Queries​

Integration Questions:

  • "How do I integrate FastAPI with FluxLoop?"
  • "What's the best runner pattern for Express.js?"
  • "How to configure simulation.yaml for WebSocket handlers?"

Troubleshooting:

  • "Why are my traces not showing up?"
  • "How to fix recording mode errors?"
  • "What does 'runner target not found' mean?"

Configuration:

  • "How do I set up replay_args in simulation.yaml?"
  • "What evaluation metrics are available?"
  • "How to configure LLM input generation?"

Understanding Results​

Knowledge Search returns:

  1. Answer: Synthesized response based on indexed documentation
  2. Citations: Source file paths and sections where information was found
  3. Output: Full response appears in the FluxLoop output channel

Example Output​

==========================================================
[MCP FAQ] How do I integrate FastAPI with FluxLoop?
----------------------------------------------------------
Answer:

To integrate FluxLoop SDK into a FastAPI application:

1. Install the SDK:
   pip install fluxloop

2. Add the @fluxloop.trace() decorator to your endpoint:
   
   from fluxloop import fluxloop
   
   @app.post("/chat")
   @fluxloop.trace()
   async def chat_endpoint(request: ChatRequest):
       # Your handler code
       pass

3. Configure configs/simulation.yaml with runner target:
   
   runner:
     target: "app.main:chat_endpoint"

Citations:
- packages/website/docs-sdk/integration/fastapi.md
- packages/website/docs-cli/configuration/runner-targets.md

Tips for Better Results​

Be Specific​

❌ Too Vague: "How to use FluxLoop?"
✅ Better: "How do I trace a Next.js API route with FluxLoop SDK?"

Include Context​

❌ Minimal: "runner configuration"
✅ Better: "How to configure runner for an async generator in Python?"

Reference Your Stack​

❌ Generic: "How to set up tracing?"
✅ Better: "How to set up tracing in a NestJS controller with TypeScript?"

Viewing Search History​

Recent searches are stored in the Integration view under Recent Suggestions:

  1. Expand Recent Suggestions
  2. Click any entry to view the full result again
  3. Results are stored per-project

Clear History​

To clear suggestion history:

  1. Right-click on Recent Suggestions in the Integration view
  2. Select Clear History

Or use Command Palette: FluxLoop: Clear Integration Suggestion History

Troubleshooting​

"fluxloop-mcp package is not installed"​

Install the MCP package:

pip install fluxloop-mcp

Then click Connect MCP in the Integration view.

"MCP Index not found"​

Build the knowledge index:

fluxloop-mcp rebuild-index

Or if you cloned the repo:

cd packages/mcp
./scripts/rebuild_index.sh

"No response returned"​

Check the FluxLoop output channel for detailed error messages. Common causes:

  • Index not built or corrupted
  • Python environment issues
  • fluxloop-mcp not in PATH

Next Steps​