How adarshh9/docqa-aws works
Project type and purpose
• This is a microservices‐based Document Q&A system (“DocQA AI Agent”) built on FastAPI, Docker, and machine learning.
• Its goal is to let users upload PDF/TXT documents, extract content, build embeddings, and answer questions via an LLM, all orchestrated by an API gateway.
File structure
• Top-level services (each with its own Dockerfile, main.py, requirements.txt):
• api-gateway
• ingestion-service
• vector-service
• llm-service
• memory-service
• Shared code in shared (models.py, utils.py) used by multiple services
• Orchestration via docker-compose.yml
• Helpers/tests: qdrant_debug.py, workspace_dumper.py, test_compete_system.py
• A simple index.html and package.json (likely for a minimal frontend or health checks)
README for additional insights
• Ports: 8000 (API Gateway), 8001–8004 for individual services
• Key features: document upload, embedding, similarity search, session memory, LLM generation
• Production notes: swap in Redis/PostgreSQL, external LLMs, proper vector DBs, observability
Detailed capture of essential elements
• Shared code as a package imported by each service
• Health check endpoints on each service