[ML Assessment] Transcript Analysis API - Matias Martos

[matiasmartos]

Summary

Implementation of a REST API for analyzing coaching transcripts using OpenAI's structured outputs. Built following hexagonal architecture principles.

Requirements Completed

Point 1 (Mandatory) ✅

• POST /transcripts/analyze - Analyze a single transcript
• GET /transcripts/{id} - Retrieve analysis by ID
• Input validation (empty transcript handling)
• In-memory storage
• Unique ID generation per analysis

Point 2 (Optional) ✅

• POST /transcripts/analyze/batch - Concurrent analysis of multiple transcripts
• Async processing with asyncio.gather()

Architecture

app/
├── adapters/       # OpenAI adapter, InMemoryRepository
├── api/            # FastAPI routes, schemas, dependencies
├── application/    # Services, DTOs
├── domain/         # Domain models
└── ports/          # Abstract interfaces (LLM, Repository)

How to Run

# Install dependencies
poetry install

# Configure environment
cp .env.example .env
# Add your OPENAI_API_KEY

# Start server
uvicorn app.main:app --reload

# Run tests
pytest

API Documentation

Available at http://localhost:8000/docs (Swagger UI)

Tests

• Unit tests with mocked LLM adapter
• All endpoints covered
• pytest -v → 10 tests passing