How sunnysavita10/document_portal works
Project type and purpose
• This is a full-stack “Document Portal” service combining a web frontend, a Python API backend, and a suite of document-analysis/chat pipelines.
• The README calls out LLM and embedding models, vector databases, and API keys – so the service ingests user documents, indexes them, and offers single- or multi-document chat/analysis via external LLMs.
File structure
Top-level directories/files:
• api/ ← API entrypoint (api/main.py) and persisted data
• src/ ← Core business logic: ingestion, retrieval, analysis, comparison, chat
• model/ ← Model definitions (data classes or ORM models)
• exception/, logger/, utils/ ← Cross-cutting concerns: errors, logging, config, I/O, model loading
• static/, templates/ ← Web UI assets (CSS, HTML)
• config/config.yaml ← Deployment/runtime config
• Dockerfile, .github/workflows/*.yml ← Containerization & CI/CD to AWS
• notebook/ ← Exploratory data science and logging experiments
• prompt/ ← Prompt engineering library
Patterns:
• Layered architecture: Web UI → API layer → Service layer (document pipelines) → Persistence/External services
• Modular pipelines under src/document_analyzer, src/document_compare, src/document_chat, etc.
README for additional insights
• Lists supported LLMs and embedding models.
• Emphasizes vector DB choice (in-memory, on-disk, cloud).
Elements
– Clear labels for each component and sub-component
• Blues for internal services
• Greens for data stores