Repository navigation
Home
Rishu edited this page Oct 21, 2026
·
2 revisions
Noema is a Python library for token-efficient RAG. It parses text into a graph of entities and relations, merges duplicate facts, and gives your LLM only the facts a question needs, in a compact notation.
Larry Page founded Google in 1998. Sundar Pichai is the CEO of Google.
becomes
#1 Larry Page :PERSON
#2 Google :ORG
#3 Sundar Pichai :PERSON
#1 -found-> #2
#3 -ceo_of-> #2
| If you want to... | Read |
|---|---|
| Install Noema | Installation |
| See it work in five minutes | Quick Start |
| Understand the model | Core Concepts |
| Page | What it covers |
|---|---|
| Compressing Text | One-off compression with noema.compress()
|
| Knowledge Bases | Ingesting, saving, loading and updating a knowledge base |
| Retrieval |
kb.retrieve(), token budgets, top_k and hops
|
| Notation Format | The compact notation grammar, encoding and decoding |
| Round-Trip Validation | Confidence scores for compressed notation |
| Advanced NLP | Coreference resolution, transformer relations, NLI validation |
| Large Datasets and Backends | Streaming ingestion, Neo4j and RDF |
| Integrations | LangChain, LlamaIndex and token counting |
| CLI Reference | The noema command |
| Benchmarks | Running and reading the benchmark suite |
| Page | What it covers |
|---|---|
| API Reference | Public functions, classes and data types |
| Troubleshooting | Common errors and fixes |
| Contributing | Development setup, tests and linting |