| name | pipecat-frame-processor |
| description | Wire a new FrameProcessor into a pipecat pipeline — create the class, refine it, and connect it in app.py |
/pipecat-frame-processor
When to use
Use this when you need to intercept or transform frames in a pipecat pipeline — e.g., intent routing, transcription filtering, or any pre/post-LLM processing step that requires a named position in the pipeline.
Steps
- Create the processor class — Append a new
FrameProcessor subclass to the appropriate module (e.g., agent/intent.py). Implement process_frame to handle the target frame type and pass others through.
- Refine the implementation — Replace any thin/placeholder version with the full implementation: proper
asyncio passthrough, correct frame type guards, and any side-effect logic (e.g., intent tagging, routing flags).
- Update imports in
app.py — Add the new class to the import block from its module.
- Wire into the pipeline — Insert the processor at the correct position in the pipeline constructor (e.g., between
audio_tags and user_agg for pre-LLM transcription interception).
- Verify ordering — Confirm the processor sits at the right stage: before the LLM aggregator for input hooks, after for output hooks.
Examples
- Intent routing: Create
IntentRouterProcessor in agent/intent.py, intercept TranscriptionFrame, classify intent, attach metadata, then pass the frame downstream — inserted between audio_tags and user_agg in app.py.
- Transcription filter: Create a
ProfanityFilterProcessor that inspects TranscriptionFrame text and either drops or modifies it before the LLM aggregator sees it.
