A LangGraph agent with local function tools, guardrails, memory, human-in-the-loop, and optional MCP server connections.

Features

• Local tools: filesystem, network, web search, math, bash
• Optional MCP server connections (math, perf)
• Configurable LLM providers: Ollama (default), OpenAI, Google Gemini, Anthropic, Microsoft Azure, Meta, DeepSeek, Mistral, xAI, OpenRouter
• Conversation memory via LangGraph checkpointing (SQLite)
• Guardrails: input validation, LLM-as-judge, output PII redaction, bash command denylist
• Human-in-the-loop tool call confirmation
• Graceful error handling for LLM connection failures
• HTTP API via FastAPI with streaming support
• Playground UI at /playground
• React frontend under frontend/
• LangSmith tracing support

Getting Started

uv sync
./run-cli-agent.sh

For a different LLM provider:

A production-ready WebSocket client/server app using FastAPI.

docker run --rm --name websocket-python-example -p 8080:8080 -it jecklgamis/websocket-python-example:main

Tech Stack

• FastAPI - async web framework with WebSocket support
• websockets - WebSocket client library
• Pydantic v2 - validation and settings management
• Docker - containerized deployment (Python 3.12-slim)
• Helm - Kubernetes deployment chart
• Ruff - linting and formatting
• pytest - testing with async support

Project Structure

├── app/
│   ├── __init__.py        # Global logging init (INFO level)
│   ├── main.py            # FastAPI application entry point
│   ├── config.py          # Environment-specific settings (dev/test/prod)
│   └── routers/           # API route handlers
│       └── websocket.py   # WebSocket /ws endpoint
├── tests/                 # Test suite
├── deployment/
│   └── k8s/helm/          # Helm chart and deploy Makefile
├── websocket_client.py    # WebSocket client
├── run-server.sh          # Start the server
├── run-client.sh          # Start the client
├── Dockerfile
├── Makefile
└── pyproject.toml

Getting Started

Prerequisites

• Python 3.12+
• uv (recommended)
• Docker (for containerized deployment)

# Clone the repository
git clone git@github.com:jecklgamis/websocket-python-example.git && cd websocket-python-example

# Install dependencies
make install-deps

# Start the dev server (uses .env.dev by default)
make run-dev

The server will be available at http://localhost:8080. This will reload on file changes.