# PrismRAG

One canonical identity. Human page: https://www.insightits.com/products/prismrag.html

| Field | Value |
|-------|-------|
| Slug | `prismrag` |
| Status | published |
| Family | aiRetrieval |
| Infra category | AI Retrieval |
| Record type | product |
| Parent | — |
| JSON | https://www.insightits.com/catalog/prismrag.json |
| Markdown | https://www.insightits.com/catalog/prismrag.md |
| GitHub | https://github.com/insightitsGit/prismrag |
| PyPI | https://pypi.org/project/prismrag-patch/0.2.1/ |
| Install | `pip install "prismrag-patch[graph]" (Apache-2.0, v0.2.1). No license key. Interactive demo: examples/demo_app.` |
| Version | 0.2.1 |

## What it is

Taxonomy-controlled Graph RAG library (PyPI: prismrag-patch). Client-defined word-to-category mapping, dual vectors, Louvain communities, BFS graph search, and auditable retrieval for pgvector, ChromaDB, Pinecone, and Weaviate.

## Problem

Flat top-k retrieval bleeds categories. Teams that own category rules need deterministic projection and graph search — not Microsoft GraphRAG-style document co-occurrence as the only control.

## Alternatives

- **Microsoft GraphRAG** — Published comparison page: PrismRAG is a library with user-defined taxonomy; Microsoft GraphRAG is a document co-occurrence graph approach. No comparative accuracy claim. ([source](https://www.insightits.com/compare/prismrag-vs-microsoft-graphrag.html))

## Features

- **Tier-1 Rules Mapping** — Auditable word→category assignment; 768-d semantic → 256-d personal projection. Same rules → same vectors every run.
- **Graph RAG Search** — Community seeding → BFS on your word graph → semantic re-rank. Direct fallback when the graph path runs dry.
- **Louvain Communities** — Automatic topic clusters with optional LLM labels. Bridge vectors link communities for cross-domain queries.

## Architecture

Tier-1 rules mapping (word→category) → 768-d semantic to 256-d personal projection → Louvain communities → BFS on the word graph with semantic re-rank → direct fallback when the graph path runs dry. MemoryStore for tests; PostgresStore for production.

## Use cases

- Healthcare, pharmacy, and finance taxonomies you own
- ChorusGraph RetrievalBackend over a customer database
- Teams that need the same mapping to produce the same vectors every run

## Benchmarks

The public evaluation is a repository test harness, not a neutral cross-vendor benchmark. No comparative accuracy or speed claim is made. Teams should evaluate retrieval quality on their own taxonomy, documents, embedding model, and false-positive budget.

### Disclosures

- The public evaluation is a repository test harness, not a neutral cross-vendor benchmark. No comparative accuracy or speed claim is made. Teams should evaluate retrieval quality on their own taxonomy, documents, embedding model, and false-positive budget.

### Evidence URLs

- https://github.com/insightitsGit/prismrag/blob/main/tests/test_lib_step06_evaluation.py
- https://github.com/insightitsGit/prismrag/blob/main/examples/demo_app/test_integration.py


## GitHub

https://github.com/insightitsGit/prismrag

## PyPI

https://pypi.org/project/prismrag-patch/0.2.1/

## Documentation

- [Taxonomy Graph RAG guide](https://www.insightits.com/guides/taxonomy-graph-rag.html)
- [PrismRAG vs Microsoft GraphRAG](https://www.insightits.com/compare/prismrag-vs-microsoft-graphrag.html)
- [Graph RAG tools](https://www.insightits.com/compare/graph-rag-tools.html)
- [INFO.md](https://github.com/insightitsGit/prismrag/blob/main/INFO.md)
- [Evaluation harness](https://github.com/insightitsGit/prismrag/blob/main/tests/test_lib_step06_evaluation.py)
- [howItWorks](https://www.insightits.com/products/prismrag.html#what-is-prismrag)
- [vsMicrosoft](https://www.insightits.com/products/prismrag.html#why-prismrag)
- [pricing](https://www.insightits.com/products/choruscontrol.html#pricing)
- [register](https://pypi.org/project/prismrag-patch/)
- [libDocs](https://github.com/insightitsGit/prismrag/tree/main/docs)
- [interactiveDemo](https://github.com/insightitsGit/prismrag/tree/main/examples/demo_app)
- [demoAppReadme](https://github.com/insightitsGit/prismrag/blob/main/examples/demo_app/README.md)

## Is not

- Hosted SaaS (retired — use the Apache-2.0 library or self-host)
- A neutral cross-vendor accuracy or speed claim
- VectorPrism (intent-gated 1024d retrieval) or VectorBridge (migration)

## Related products

- [chorusgraph](./chorusgraph.md)
- [python-libs](./python-libs.md)
- [vectorprism](./vectorprism.md)