🧠 Ollama SlackBot — LangGraph-Based Chatbot for Slack

an intelligent, traceable assistant built with LangGraph, LangChain, ChromaDB, Ollama, and Langfuse. It integrates directly with Slack to answer questions using a structured RAG (Retrieval-Augmented Generation) pipeline.

🔧 Tech Stack

• Slack Bot (Bolt Socket Mode) — Interacts with users in DM and threads
• LangGraph — Orchestrates the workflow using a stateful graph
• LangChain — Connects to Ollama LLM + Chroma vector store
• Ollama — Local or remote LLM + embedding model
• ChromaDB — Fast, lightweight vector database
• Langfuse — Observability with trace logging for each session

---

🚀 Quickstart

1. Clone & Setup Environment

   git clone <this-repo>
   cd supervisor_api
   cp .env.example .env

2. Update .env with your credentials

   • Slack tokens
   • Ollama / Chroma configuration
   • Langfuse API keys

3. Run the stack

   docker-compose up --build

4. Test the API

   curl -X POST http://localhost:8080/query \
     -H "Content-Type: application/json" \
     -d '{"query": "How do I update our time-off policy?"}'

---

🗺️ System Overview

graph TD
    A[Slack Message] -->|DM or Thread| B[Slack Bolt App]
    B --> C[LangGraph Supervisor]
    C --> D{Intent?}
    D -->|Small Talk| E[LLM Only]
    D -->|Retrieve| F[ChromaDB Lookup]
    F --> G[Rank + Compose]
    G --> H[LLM Generate]
    H --> I[Slack Response]
    E --> I
    C -->|Trace Events| J[Langfuse]

Loading

---

🧠 Core Logic

• main.py — FastAPI entrypoint
• slack_integration.py — Slack event router
• langgraph_supervisor.py — LangGraph node pipeline:
  • classify → retriever → grader → response
• supervisor_agent.py — LangChain-based helpers (Ollama, Chroma)
• config.py — Environment-aware config for all services

---

🛠️ Dev Notes

• Chroma must be reachable via REST if running in Docker (CHROMA_HOST, CHROMA_PORT)
• Ollama must be reachable via REST if running in Docker (OLLAMA_HOST, OLLAMA_PORT)
• Langfuse requires valid public/secret keys and environment setup
• Responses are logged to Langfuse with full graph execution trace
• Slack integration supports DM and thread replies (notifies user in DM if public channel)

---

✅ To-Do

• Improve confidence-based grading logic
• Add document upload ingestion pipeline
• Integrate memory or Redis for long-term thread state

---

📄 License

MIT — free to use, modify, and contribute.

---