AI Utils API

FastAPI service for ML-powered text, PDF, and audio processing with OpenAI integration.

Quick Start

# Install dependencies
uv sync

# Run locally
source .venv/bin/activate
uv run fastapi dev src/aiutils/main.py

# Or run with Docker
docker build -t ai-utils .
docker run -p 80:80 ai-utils

Features

• Text Processing: Convert text to embeddings and tokens using multilingual-e5-small model
• PDF Processing: Extract and embed text from PDF documents
• Audio Processing: Transcribe MP3 files using OpenAI Whisper
• OpenAI Integration: Access GPT models for advanced processing
• Vector Search: Vespa integration for similarity search

API Endpoints

• GET /health/live - Liveness check
• GET /health/ready - Readiness check
• POST /text/to-tensor/ - Text to embedding vectors
• POST /text/to-tokens/ - Text tokenization
• POST /pdf/* - PDF processing endpoints
• POST /mp3/* - Audio transcription endpoints
• POST /openai/* - OpenAI-powered endpoints

Requirements

• Python 3.12+
• System dependencies: ffmpeg, poppler-utils (included in Docker image)
• Environment variables: OpenAI API key and other configs in .env

Architecture

src/
├── aiutils/
│   ├── main.py          # FastAPI app with lifespan context
│   ├── routes/          # API endpoints by processing type
│   ├── services/        # Business logic and integrations
│   ├── schemas/         # Pydantic models
│   └── core/            # Config, security, utilities
└── tests/               # Test suite

Models are pre-loaded at startup for optimal performance.