How msilver22/agentic_mbd works
Project type and purpose
• This is a full-stack AI-agent orchestration project. It provides three “agents” (Feed Builder, Social Prompter, and a combined MBD Agent that includes a small-talk capability) implemented in Python, exposed via Streamlit UIs, and wired together using LangGraph workflows. It integrates with the Farcaster social network (via the mbd API), Groq for LLM inference, and Langfuse for observability.
File structure
• Top-level files and folders:
• .devcontainer/ – Docker/devcontainer definition for reproducible environments
• requirements.txt – Python dependencies: LangGraph, Groq SDK, mbd client, Streamlit, Langfuse, etc.
• src/ – core business logic implementing each agent (feed_builder.py, social_prompter.py, mbd_agent.py, and a v2 of mbd_agent)
• streamlit/ – one Streamlit script per agent to serve a browser-based UI
Readme insights
• The README describes three LangGraph-based agents with contextual memory, tool-enabled LLM calls, and real-time data fetch from Farcaster.
• It also highlights the use of Groq LLMs for certain nodes (summarization, planning, tool execution) and Langfuse for tracing.