🔥 Phoenix AI Services is MCP server

Unified agentic framework to run dynamic RAG APIs and utility tools like calculator, date, and python evaluator.

phoenix_ai_services

phoenix_ai_services is a modular Python package for managing RAG endpoints, tool-based APIs, and plug-and-play AI utilities using FastAPI & Poetry. It supports RESTful control for registering, updating, and deleting endpoints for RAG, tools, and more.

🔧 Setup

poetry install
poetry run python phoenix_ai_services/main.py

🧩 MCP Server (Latest Standard)

Phoenix AI Services now exposes an MCP-compatible SSE server mounted at /mcp by default.

• SSE endpoint: GET /mcp/sse
• Message endpoint: POST /mcp/messages/

You can change the mount path via PHOENIX_MCP_MOUNT_PATH.

Example: Run with Uvicorn (Production)

PHOENIX_HOST=0.0.0.0 PHOENIX_PORT=8003 uvicorn phoenix_ai_services.main:app

🚀 API Endpoints

RAG

• POST /rag/endpoints/{name} – Register RAG
• PUT /rag/endpoints/{name} – Update RAG
• DELETE /rag/endpoints/{name} – Remove RAG
• GET /rag/query/{name} – Ask RAG agent

Tools

• GET /tool/calculator?input_data=2+3*4
• GET /tool/system_time
• GET /tool/python?input_data=round(3.14159, 2)

Admin

• GET /rag/endpoints – View all registered endpoints

Health

• GET /healthz – Liveness check
• GET /elk/query?q=message:error – Query ELK with a query string

⚙️ Production Configuration

Environment variables:

• PHOENIX_HOST (default: 0.0.0.0)
• PHOENIX_PORT (default: 8003)
• PHOENIX_LOG_LEVEL (default: info)
• PHOENIX_MCP_MOUNT_PATH (default: /mcp)
• PHOENIX_ELK_BASE_URL (example: https://your-elasticsearch:9200)
• PHOENIX_ELK_INDEX (example: logs-*)
• PHOENIX_ELK_API_KEY (optional, API key for Elasticsearch)

🔍 ELK MCP Integration Examples

Configure ELK connection

export PHOENIX_ELK_BASE_URL="https://your-elasticsearch:9200"
export PHOENIX_ELK_INDEX="logs-*"
export PHOENIX_ELK_API_KEY="your_api_key"

REST query example

curl "http://localhost:8003/elk/query?q=service:api%20AND%20level:error&size=5"

MCP tool example (client-side)

Use your MCP client to call the mcp_query_elk tool:

tool: mcp_query_elk
args:
  query: "service:api AND level:error"
  size: 5

🧠 Powered by Phoenix Agentic AI Framework