How risasaragula/zaravya-knowledge-assist works
Zaravya Knowledge Assist provides two user workflows: questions over the bundled company knowledge base, and questions about an uploaded company document. The Flask app serves the web interface and routes requests; rag.py handles knowledge-base chunking, embedding, retrieval and Gemini answers, while uploaded-document questions use Gemini with the uploaded file. The README also documents product Q&A, document upload, and the company policy template. Source excerpts show the main request paths; template presentation details are not sampled.
Main entry points: main.py Flask routes · script.js Web client · rag.py Knowledge answering · company_policy_template.html Policy template
Components 14
Web interface
- Web client
script.js - Assistant page
index.html - Policy template
company_policy_template.html
Flask application
- Flask routes
main.py - Uploaded file
main.py
Knowledge assistant
- Zaravya knowledge base
zaravya.txt - Chunk preparation
rag.py - Embedding generation
rag.py - Vector retrieval
rag.py - Knowledge answering
rag.py
Company document assistant
- Uploaded-document answering
rag.py
External services
- Gemini API
- Local file storage
Other components
- User
Connections 18
- User → Web client: uses
README.md - Web client → Flask routes: posts questions
script.js - Flask routes → Knowledge answering: calls
main.py - Flask routes → Uploaded-document answering: calls
main.py - Web client → Flask routes: uploads file
script.js - Flask routes → Uploaded file: saves file
main.py - Flask routes → Local file storage: writes file
main.py - Flask routes → Policy template: serves template
main.py - Flask routes → Assistant page: serves page
main.py - Zaravya knowledge base → Chunk preparation: provides text
rag.py - Chunk preparation → Embedding generation: embeds chunks
rag.py - Embedding generation → Gemini API: requests embeddings
rag.py - Knowledge answering → Vector retrieval: requests matches
rag.py - Vector retrieval → Embedding generation: embeds question
rag.py - Knowledge answering → Gemini API: generates answer
rag.py - Uploaded-document answering → Uploaded file: uses saved path
main.py - Uploaded-document answering → Gemini API: uploads file and prompts
rag.py - User → Policy template: views policy template
README.md