VideoMCP

# Nanobot MCP Video Service A Model Context Protocol (MCP) service for Nanobot, accessible via HTTP, that forwards multimodal inference requests to the vLLM backend. ## Features - Based on Anthropic MCP Python SDK (`mcp`) - Utilizes Streamable HTTP transport (`/mcp`) - MCP tool `analyze_video(prompt, video_url)` - Forwards requests to vLLM `/v1/chat/completions` - Provides Dockerfile and docker-compose for deployment ## API Protocol From Nanobot client to this MCP service (tool input): - `prompt`: Text instruction - `video_url`: Video link The service forwards these two parameters to vLLM and returns the model-generated text result. ## Environment Variables Copy the example file and modify as needed: ```powershell copy .env.example .env ``` Common configurations: - `VLLM_BASE_URL`: vLLM service address (e.g., `http://YOUR_VLLM_HOST:PORT`) - `VLLM_MODEL`: Model name - `VLLM_API_KEY`: API Key (if vLLM requires authentication) - `VLLM_VIDEO_FPS`: Video frame rate sampling (optional) ## Local Deployment ```powershell cd nanobot-mcp python -m venv .venv .\.venv\Scripts\activate pip install -r requirements.txt copy .env.example .env uvicorn main:app --host 0.0.0.0 --port 8000 ``` Health check: `http://localhost:8000/healthz` MCP service address: `http://localhost:8000/mcp` ## Docker Deployment ```powershell cd nanobot-mcp docker build -t nanobot-mcp:latest . docker run -d --name nanobot-mcp -p 8000:8000 --env-file .env nanobot-mcp:latest ``` ## Docker Compose Deployment ```powershell cd nanobot-mcp docker compose up -d --build ```