pipecat-ai / pipecat

Open Source framework for voice and multimodal conversational AI
BSD 2-Clause "Simplified" License
3.29k stars 308 forks source link

class LangchainProcessor() do it supports multi agentic workflow ...sort of Langgraph? #302

Open geekofycoder opened 3 months ago

geekofycoder commented 3 months ago

If we want to connect multiple chatbots with different prompt context,multiple agents is something that comes to mind.Can we achieve this presently through any means or we have to wait for langgraph processor.

Basically my question is that is there any way to achieve multi agentic workflow in pipecat as of now?

TomTom101 commented 3 months ago

I am working on a Langgraph processor as we speak. I can push this into a feature branch this week, for a full PR I still need to create a test case and example. Stay tuned!

TomTom101 commented 3 months ago

For the time being, check out https://github.com/TomTom101/pipecat/tree/feat/langgraph Instantiate a processor and place it in the pipe with a compiled graph:

workflow = StateGraph(AgentState)
graph: CompiledGraph = workflow.compile()
lg = LanggraphProcessor(graph)

Let me know how well this works for you, I have only really used and tested it for a specific use case so far.

geekofycoder commented 3 months ago

sure waiting to use multi agentic workflow 😃

TomTom101 commented 3 months ago

Well, your agents are all in the graph, see e.g. https://github.com/langchain-ai/langgraph/blob/main/examples/multi_agent/agent_supervisor.ipynb?ref=blog.langchain.dev

geekofycoder commented 3 months ago

Well, your agents are all in the graph, see e.g. https://github.com/langchain-ai/langgraph/blob/main/examples/multi_agent/agent_supervisor.ipynb?ref=blog.langchain.dev

any updates on the Langgraph processor. I had a chatbot and I wanted to replicate with pipecat.

geekofycoder commented 3 months ago

Just for the quick start the chatbot I am following is https://github.com/langchain-ai/langgraph/blob/main/examples/chatbots/information-gather-prompting.ipynb you can show me some light of how to automating the above.