mcp_demo

Local setup ```sh # Get {YOUR_GOOGLE_API_KEY} at https://console.cloud.google.com/apis/credentials # Tips: Restrict the API key so it can only access the Google Custom Search API. # Create and get {YOUR_SEARCH_ENGINE_PROJECT_ID} at https://programmablesearchengine.google.com/controlpanel/all # Tips: Add the NewsArticle schema to the page restrictions so you only retrieve articles. # First run with migration ennabled RUN_MIGRATION=1 \ API_KEY={YOUR_GOOGLE_API_KEY} \ PROJECT_ID={YOUR_SEARCH_ENGINE_PROJECT_ID} \ docker compose -f dockercompose.yml up --build # From the second time, just need to run up --build docker compose -f dockercompose.yml up --build ``` Run local LLM model ```sh # download Ollama at https://ollama.com/download # run qwen2.5:3b model (good for testing on local) ollama run qwen2.5:3b ``` Run MCP client - pipx install poetry - For tts usage: install any player mpv | mpg123 | aplay ```sh cd src/projects/host_chat poetry install # LLM models # ollama/qwen2.5:3b # openai/gpt-4o-mini # For debug export DEBUG=1 # For Open AI models export OPEN_AI_API={YOUR_API} FINDXAI_MCP_CONNECTION=http://localhost:8080/sse \ TTS_REST_CONNECTION=http://localhost:8081 \ REDIS_URL=redis://localhost:6379?db=3 \ LLM_MODEL=ollama/qwen2.5:3b \ python main.py ``` For remote MCP server ```sh export FINDXAI_MCP_CONNECTION=[http/https]://<ip>:<port>/sse ```