[INFRA-AI-06] Text AI Parser Containerization and :11437 Port Isolation

Infrastructure / MLOps

Published

October 14, 2026

NoteTask Brief
  • Repository / Component: ai_voice_text_parser.
  • Stack: Docker, Docker Compose, Ollama (Qwen).
  • Task Type: DevOps / Infra.
  • Status: Ready for Implementation

Task Description

It is necessary to package the Python text structuring service into a Docker container, configure its isolation from other AI components on port 11437, and automate the downloading of the lightweight LLM model weights for qwen2.5:3b.

Step-by-Step Instructions

  1. Network Port Isolation: In the launch file (uvicorn.run) and the Dockerfile configuration, fix the port to 11437 so that the service can run concurrently with the receipt gateway (11435) and the STT gateway (11436).
  2. Dockerfile Manifest Development: Define the container build based on python:3.10-slim, configuring the installation of fastapi, requests, and pydantic packages.
  3. Orchestration in Docker Compose: Add the ai-voice-parser service to the shared configuration file. Map ports 11437:11437 and link it to the ollama-core container using the shared OLLAMA_HOST environment variable.
  4. Model Seeding Automation: Add an initialization step (ollama run qwen2.5:3b) to the deployment script so that the model is automatically pulled into local volume storage upon the environment’s first launch.