temurjonm/ai-search-rag

How temurjonm/ai-search-rag works

This is a FastAPI RAG search service: users upload tenant-scoped PDFs or text files, then ask questions and receive source-cited answers. Upload processing extracts text (with OCR fallback), chunks and embeds it, and persists document/chunk records; search embeds the question, fuses vector and keyword candidates, and generates an answer from retrieved sources. The README also documents a browser UI with SSE progress and defensive upload handling. Relationships below are based on the supplied excerpts and README; unsampled UI behavior is documented, not source-inspected.

Main entry points: main.py HTTP routes · ingestion.py Upload processing · retrieval.py Hybrid retrieval · db.py Database sessions

Components 16

API and UI

Web UI src/ai_search/static/index.html
HTTP routes src/ai_search/main.py
API schemas src/ai_search/schemas.py

Document ingestion

Upload processing src/ai_search/ingestion.py
Text extraction OCR src/ai_search/document_loader.py
Text chunking src/ai_search/chunking.py
Uploaded files storage

Retrieval and answers

Embedding client src/ai_search/openai_client.py
Hybrid retrieval src/ai_search/retrieval.py
Answer generation src/ai_search/openai_client.py

Data and configuration

Database sessions src/ai_search/db.py
Document and chunk models src/ai_search/models.py
Runtime settings src/ai_search/config.py

Other components

User
OpenAI API
PostgreSQL pgvector

Connections 24

User → Web UI: uses README.md
Web UI → HTTP routes: uploads and searches README.md
User → HTTP routes: calls API README.md
HTTP routes → Upload processing: dispatches uploads src/ai_search/main.py
HTTP routes → Embedding client: embeds query src/ai_search/main.py
HTTP routes → Hybrid retrieval: runs search src/ai_search/main.py
HTTP routes → Answer generation: generates answer src/ai_search/main.py
HTTP routes → Database sessions: initializes database src/ai_search/main.py
Upload processing → Text extraction OCR: extracts text src/ai_search/ingestion.py
Upload processing → Text chunking: chunks text src/ai_search/ingestion.py
Upload processing → Embedding client: embeds chunks src/ai_search/ingestion.py
Upload processing → Uploaded files: writes uploads src/ai_search/ingestion.py
Upload processing → Document and chunk models: creates records src/ai_search/ingestion.py
Upload processing → Runtime settings: reads chunk settings src/ai_search/ingestion.py
Embedding client → OpenAI API: requests embeddings src/ai_search/openai_client.py
Answer generation → OpenAI API: requests completion src/ai_search/openai_client.py
Hybrid retrieval → Document and chunk models: queries records src/ai_search/retrieval.py
Hybrid retrieval → API schemas: builds sources src/ai_search/retrieval.py
HTTP routes → API schemas: validates and returns src/ai_search/main.py
Database sessions → Runtime settings: reads database URL src/ai_search/db.py
Database sessions → Document and chunk models: creates schema src/ai_search/db.py
Database sessions → PostgreSQL pgvector: opens sessions src/ai_search/db.py
Document and chunk models → PostgreSQL pgvector: maps tables README.md
Hybrid retrieval → PostgreSQL pgvector: runs ranked queries src/ai_search/retrieval.py

Drawn by GitDiagram on November 10, 2026 from the repository's file tree, README and source files. Read as Markdown ·