How pipecat-ai/pipecat-flows works
Project type and purpose
• Pipecat-Flows is a hybrid open-source toolkit:
• A Python conversation-flow framework (FlowManager, FlowsFunctionSchema, adapters, actions, context strategies)
• A standalone web-based visual flow editor (built with Vite, Tailwind, vanilla JS)
• Purpose: let developers design, manage, and run both static and dynamic LLM-driven conversational flows, with optional TTS, event hooks, and external integrations.
File structure
Top‐level:
• src/pipecat_flows/ → core Python library modules: manager.py, actions, adapters, types, exceptions
• editor/ → self‐contained frontend app (index.html, js/, css/, build config, examples)
• examples/ → sample Python scripts (static & dynamic flows) and assets demonstrating real‐world usage
• tests/ → pytest suite verifying core library behavior
• . github/workflows/ → CI/CD pipelines for testing and publishing to PyPI
• pyproject.toml, requirements.txt → packaging & dependencies
Patterns & configs: plugin architecture (adapters, actions), config-driven flows (JSON or dict), LLM function‐calling abstractions.
README for insights
• Architecture outline: FlowManager core, function schemas, node vs edge functions, context strategies
• Supported LLM providers (OpenAI, Anthropic, Gemini), TTS, transport (Daily), built-in & custom actions
• Usage scenarios: static vs dynamic flows, pre/post actions, examples in examples/ and editor/examples.