Vanijya AI / TradeSight Codebase Dossier

Vanijya AI is a trade-intelligence platform combining structured-data analysis and an advanced RAG workspace. The rag/ application provides the inspected FastAPI/React system; vanijya_ai_use_case.md documents its integration with an Analyst API for code-based structured insights.

Read first

• Product overview
• Agentic architecture
• Backend deep dive
• Data and retrieval
• Unique engineering

System map

flowchart TD
  classDef default fill:#1e293b,stroke:#38bdf8,stroke-width:2px,color:#f8fafc
  classDef highlight fill:#065f46,stroke:#34d399,stroke-width:2px,color:#f0fdf4
  User["Trade analyst"]:::highlight --> FE["Unified React workspace"]
  FE --> Analyst["Analyst API / structured data"]
  FE --> RAG["FastAPI RAG API"]
  RAG --> Qdrant["Qdrant dense vectors"]
  RAG --> Meili["Meilisearch sparse search"]
  RAG --> Redis["Redis memory"]
  RAG --> PG["PostgreSQL metadata/history"]
  RAG --> MinIO["MinIO documents"]
  linkStyle default stroke:#64748b,stroke-width:2px

Top facts

• Five use cases are scoped through project IDs and specialized prompts.
• Hybrid retrieval merges dense and sparse results, then reranks.
• Document ingestion supports PDF/layout extraction, chunking, embeddings, summaries, and suggestions.
• SSE streams answer tokens and pipeline/debug events.
• Redis supplies sliding-window conversation memory; PostgreSQL stores durable history.
• JWT SSO is shared across structured and unstructured workspaces.
• Docker Compose runs Postgres, Redis, MinIO, Qdrant, Meilisearch, backend, worker, and frontend.