How abdulkaiz-007/predators works
This is an enterprise RAG web application for ingesting documents and web pages, querying them with grounded answers and citations, and inspecting retrieval quality and audit history. The main workflow runs from the chat UI through the Express API into retrieval and answer generation, returning evidence-backed results. Separate product branches manage document ingestion, corpus exploration, engine settings, and benchmark evaluation. The excerpts show in-memory corpus state; some UI and server route details are partial, so uncertain internal wiring is omitted.
Main entry points: server.ts HTTP API · App.tsx App Shell · ragEngine.ts Grounded Answering · CitationsTab.tsx Citation Explorer
Components 18
User Interface
- App Shell
src/App.tsx - RAG Chat
src/components/ChatTab.tsx - Document Manager
src/components/DocumentsTab.tsx - Citation Explorer
src/components/CitationsTab.tsx - Evaluation & Audit
src/components/EvaluationTab.tsx - Engine Settings
src/components/SettingsModal.tsx - Chunk Viewer
src/components/DocumentModal.tsx
Application API
- HTTP API
server.ts
RAG Engine
- Document Ingestion
server/ragEngine.ts - Hybrid Retrieval
server/ragEngine.ts - Grounded Answering
server/ragEngine.ts
Corpus & Evaluation
- In-Memory Corpus
server/ragEngine.ts - Engine Configuration
server/ragEngine.ts - Benchmark Runner
server/ragEngine.ts
Other components
- User
- Gemini API
- TinyFish Agent
- Web Sources
Connections 32
- User → App Shell: uses
- App Shell → RAG Chat: shows
- App Shell → Document Manager: shows
- App Shell → Citation Explorer: shows
- App Shell → Evaluation & Audit: shows
- App Shell → Engine Settings: opens
- App Shell → Chunk Viewer: opens
- RAG Chat → HTTP API: posts query
- HTTP API → Grounded Answering: dispatches query
- Grounded Answering → Hybrid Retrieval: requests passages
- Hybrid Retrieval → In-Memory Corpus: reads chunks
- Hybrid Retrieval → Grounded Answering: returns evidence
- Grounded Answering → Gemini API: generates answer
- Grounded Answering → In-Memory Corpus: writes audit
- Grounded Answering → HTTP API: returns response
- HTTP API → RAG Chat: returns answer
- Document Manager → HTTP API: uploads or crawls
- HTTP API → Document Ingestion: dispatches ingest
- Document Ingestion → TinyFish Agent: extracts page
- Document Ingestion → Web Sources: fetches page
- Document Ingestion → In-Memory Corpus: stores documents
- HTTP API → In-Memory Corpus: lists or deletes
- Citation Explorer → HTTP API: loads chunks
- Citation Explorer → HTTP API: tests retrieval
- HTTP API → Hybrid Retrieval: dispatches retrieval
- HTTP API → Engine Configuration: reads or updates
- Engine Settings → HTTP API: saves settings
- HTTP API → Benchmark Runner: runs benchmark
- Benchmark Runner → Grounded Answering: queries cases
- Evaluation & Audit → HTTP API: requests evaluation
- Evaluation & Audit → HTTP API: loads audit
- HTTP API → In-Memory Corpus: reads audit