SafeQueryAI Documentation

SafeQueryAI is a privacy-first architecture for document question-answering. It lets users upload PDF and CSV files, ask natural-language questions, and receive grounded answers from a local LLM runtime.

Product Summary

SafeQueryAI addresses a common business need: extracting actionable information from private documents without exposing that data to external services.

The product enforces session-based processing:

• Documents are processed only within the active session.
• Data is held in temporary storage and in-memory indexes.
• Session cleanup removes files and indexes after timeout or manual clear.
• Ollama is restricted to a local loopback URL to prevent remote inference endpoints.

Business Value

• Reduces privacy risk for teams working with sensitive internal documents.
• Improves speed-to-insight by enabling direct natural-language querying of uploaded files.
• Provides a demonstrable implementation of privacy-first architecture using modern web and API stacks.
• Supports portfolio and assessment scenarios with traceable architecture, testing, and documentation.

High-Level Architecture

1. Frontend receives uploads and question input.
2. Backend stores files in temporary storage per session.
3. Text extraction and chunking prepare data for retrieval.
4. Embeddings and retrieval run through the local LLM runtime (Ollama).
5. Answers are generated using only session content and returned to the user.
6. Session timeout or manual clear removes temporary files and in-memory vectors.

See Architecture and Security & Privacy for full detail.

Environment Reference

Item	Default Value
Frontend URL	http://localhost:5173
Backend API	http://localhost:5000/api
Swagger UI (development)	http://localhost:5000/swagger
Ollama URL	http://localhost:11434
Supported file types	.pdf, .csv
Session timeout	60 minutes
Max file size	20 MB (25 MB absolute request ceiling)

Audience Paths

• Product and non-technical stakeholders: Business Overview
• Technical reviewers and assessors: Architecture, Security & Privacy, Testing
• Developers: Getting Started, API Reference, Frontend Guide, Deployment
• Recruiters and portfolio reviewers: Business Overview, Roadmap, FAQ

Documentation Map

Section	Focus
Business Overview	Problem statement, users, scope, risks, success criteria
Getting Started	Local setup and first run
Architecture	System flow, components, constraints
Security & Privacy	Data handling model and trust assumptions
API Reference	Endpoint contract and examples
Frontend Guide	UI structure and frontend integration points
Testing	Test strategy, suites, and execution commands
Deployment	Current deployment approach and operational notes
Roadmap	Planned improvements and delivery priorities
FAQ	Common operational and setup questions

---

Last updated: March 2026