NISIRA Assistant is a full-stack Retrieval-Augmented Generation (RAG) conversational assistant built for organizations that need accurate, document-grounded answers. Users submit natural-language questions; the system retrieves the most relevant document chunks via hybrid search, passes them to an LLM, and returns a cited response — all within a React chat interface backed by a Django REST API.Documentation Index
Fetch the complete documentation index at: https://mintlify.com/HugoX2003/nisira-assistant/llms.txt
Use this file to discover all available pages before exploring further.
Introduction
Learn what NISIRA Assistant is, how it works, and when to use it.
Quickstart
Get NISIRA Assistant running locally in minutes with Docker.
Architecture
Explore the full system architecture: backend, RAG pipeline, and frontend.
API Reference
Browse every REST endpoint with parameters, request bodies, and responses.
Key Features
Hybrid RAG Pipeline
60% semantic search via pgvector/ChromaDB + 40% lexical BM25-style search for maximum recall.
Adaptive Retrieval
Dynamically selects 3–15 documents per query based on complexity scoring.
Multi-Provider LLMs
Plug in Gemini 2.0 Flash, OpenRouter, or Groq — swap providers via environment variables.
Google Drive Sync
Automatically ingest PDF, DOCX, PPTX, XLSX, and TXT files from a Drive folder.
RAG Evaluation Metrics
Built-in Precision@k, Recall@k, Faithfulness, Hallucination Rate, and WER tracking.
Admin Panel
React-based dashboard for managing documents, embeddings, and viewing system metrics.
Get Up and Running
Configure environment variables
Copy the example environment file and set your API keys and database credentials.At minimum, set
GOOGLE_API_KEY (or another LLM provider key) and SECRET_KEY. See Environment Variables for the full reference.Start with Docker Compose
8000, and the React frontend on port 3000.For production deployments with PostgreSQL and pgvector, see the Docker Production and DigitalOcean guides.