Releases: pipecat-ai/pipecat-flows
Releases · pipecat-ai/pipecat-flows
v0.0.6
Added
- New FlowManager supporting both static and dynamic conversation flows
- Provider-specific examples demonstrating dynamic flows:
- OpenAI:
insurance_openai.py
- Anthropic:
insurance_anthropic.py
- Gemini:
insurance_gemini.py
- OpenAI:
- Type safety improvements:
FlowArgs
: Type-safe function argumentsFlowResult
: Type-safe function returns
Changed
- Simplified function handling:
- Automatic LLM function registration
- Optional handlers for edge nodes
- Updated all examples to use unified FlowManager interface
v0.0.5
Added
-
Added LLM support for:
- Anthropic
- Google Gemini
-
Added
LLMFormatParser
, a format parser to handle LLM provider-specific
messages and function call formats -
Added new examples:
movie_explorer_anthropic.py
(Claude 3.5)movie_explorer_gemini.py
(Gemini 1.5 Flash)travel_planner_gemini.py
(Gemini 1.5 Flash)
v0.0.4
Added
- New example
movie_explorer.py
demonstrating:- Real API integration with TMDB
- Node functions for API calls
- Edge functions for state transitions
- Proper function registration pattern
Changed
-
Renamed function types to use graph terminology:
- "Terminal functions" are now "node functions" (operations within a state)
- "Transitional functions" are now "edge functions" (transitions between states)
-
Updated function registration process:
- Node functions must be registered directly with the LLM before flow initialization
- Edge functions are automatically registered by FlowManager during initialization
- LLM instance is now required in FlowManager constructor
-
Added flexibility to node naming with the Editor:
- Start nodes can now use any descriptive name (e.g., "greeting")
- End nodes conventionally use "end" but support custom names
- Flow configuration's
initial_node
property determines the starting state
Updated
- All examples updated to use new function registration pattern
- Documentation updated to reflect new terminology and patterns
- Editor updated to support flexible node naming