AI Observability Agent

Small agentic AI project scaffold for learning:

• FastAPI
• LangChain
• LangGraph
• LangSmith tracing
• Pinecone
• HuggingFace embeddings
• MCP tools/server
• React frontend

Project layout

• backend/: FastAPI app, agent graph, RAG, tools, and MCP server
• frontend/react-chat/: React chat UI placeholder
• data/: local sample docs and logs for early development
• scripts/: helper scripts such as ingestion and local setup

Suggested build order

1. Implement RAG ingestion and retrieval
2. Add FastAPI chat endpoint
3. Build LangGraph flow
4. Add tools and MCP server
5. Add React UI

Quick start

Backend

1. Create a virtual environment

    python3.11 -m venv .venv
    source .venv/bin/activate

2. Install dependencies from backend/requirements.txt

    pip install --upgrade pip
    pip install -r backend/requirements.txt

3. Copy .env.example to .env
4. Start the API:

uvicorn backend.main:app --reload --app-dir .

5. Ingest documents:

python -m backend.rag.ingest

Frontend

cd frontend/react-chat
npm install
npm run dev

Pinecone

https://app.pinecone.io/organizations/

Deployment path

This repo now includes a Docker-first deployment baseline:

• Dockerfile for packaging the backend
• .dockerignore to keep the image lean
• .github/workflows/ci.yml for tests plus Docker build validation
• .github/workflows/cd.yml for publishing a container to GHCR and triggering Render
• render.yaml as a starter Render blueprint

Run locally with Docker

docker build -t ai-observability-agent .
docker run --rm -p 8000:8000 --env-file .env ai-observability-agent

Then verify:

curl http://localhost:8000/api/health

CI

CI runs on pull requests and pushes to main:

• installs backend dependencies
• runs pytest
• builds the Docker image

CD

CD runs on pushes to main:

• builds and pushes ghcr.io/<owner>/ai-observability-agent
• tags the image with latest and the Git SHA
• triggers Render through RENDER_DEPLOY_HOOK_URL

Render setup

1. Create a Render web service from an existing image.
2. Point it at ghcr.io/<your-user-or-org>/ai-observability-agent:latest.
3. Add the environment variables from .env.example.
4. Set the health check path to /api/health.
5. Add RENDER_DEPLOY_HOOK_URL as a GitHub Actions secret.

GitHub Actions secrets

You only need one repository secret for the current CD flow:

• RENDER_DEPLOY_HOOK_URL

The workflow uses the built-in GITHUB_TOKEN to push to GHCR.