Skip to content

feat facade agent [WIP]#73

Open
Jesus-Ban wants to merge 10 commits into
mainfrom
feat-facadeagent
Open

feat facade agent [WIP]#73
Jesus-Ban wants to merge 10 commits into
mainfrom
feat-facadeagent

Conversation

@Jesus-Ban
Copy link
Copy Markdown
Collaborator

@Jesus-Ban Jesus-Ban commented Feb 26, 2025

https://github.com/supercog-ai/agentic/blob/main/examples/oss_deep_research.py#L53

Create a LangChain Wrapper

langchain_wrapper = LangChainWrapperAgent(
name="Research Agent",
model=GPT_4O_MINI,
langchain_agent=langchain_agent
)

Create a BrowserUseAgent

browser_agent = BrowserUseAgent(
name="Browser Assistant",
model=GPT_4O_MINI
)

Create an operator agent that can use both agents

operator = Agent(
name="Operator",
tools=[langchain_wrapper, browser_agent],
instructions="""

  You are an operator agent that can use both research and browser automation to help users.
  
  You have access to two specialized agents:
  1. Research Agent: Use this for general information gathering and research tasks.
  2. Browser Assistant: Use this for tasks that require browser automation.
  
  Choose the appropriate agent based on the user's request.
  """

, model=GPT_4O_MINI
)

@Jesus-Ban Jesus-Ban self-assigned this Feb 26, 2025
@scottpersinger
Copy link
Copy Markdown
Collaborator

Cool. This is a good start. Note that we have two clients for "Agentic agents": the AgentRunner for the repl, and the FastAPI endpoints in DynamicFastAPIHandler. THIS code which wraps Langchain agents should result in an object that we can pass into either of those clients and get the right behavior.

So if you look at AgentRunner and DynamicFastAPIHandler, you will see these calls to the RayFacadeAgent:

start_request and get_events

those are the key methods that you will want to implement for the LangChainFacade.

@Jesus-Ban
Copy link
Copy Markdown
Collaborator Author

Cool. This is a good start. Note that we have two clients for "Agentic agents": the AgentRunner for the repl, and the FastAPI endpoints in DynamicFastAPIHandler. THIS code which wraps Langchain agents should result in an object that we can pass into either of those clients and get the right behavior.

So if you look at AgentRunner and DynamicFastAPIHandler, you will see these calls to the RayFacadeAgent:

start_request and get_events

those are the key methods that you will want to implement for the LangChainFacade.

I've upgraded vendor_agent.py according to your feedback. Plz check and teach me..

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants