# Zleap-AI/SAG

A new SOTA for RAG — an original retrieval architecture and an open-source knowledge base for humans and agents.

Repository: https://github.com/Zleap-AI/SAG
Canonical: https://ross.abutalabs.com/products/sag
Homepage: https://zleap.com
Language: Python
License: MIT
License Family: permissive
Topics: ai, knowledge-graph, llm, rag, vector-search, data-engineering, graphrag, knowledge-base, knowledge-graphs, agent, sag
Last push: 2026-08-25T19:25:25+00:00

## Health v2 (maintenance only)
Score: 83/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 99, longevity 21
- inputs: {"age_days": 299, "days_push": 8, "days_rel": 9, "gap_med": 0.0, "n_releases_24m": 29}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2425, forks 150 (observed 2026-08-28T04:06:50.747917+00:00)

## What it is
SAG is an open-source retrieval architecture and knowledge base application that replaces both traditional RAG and GraphRAG with event-entity indexing and query-time dynamic hyperedges. It ships as a Python package (zleap-sag), a desktop app, and a CLI that mounts a Knowledge MCP server into agents like Codex and Claude Code.

## Use cases
- build a searchable, source-traceable knowledge base from scattered documents
- answer multi-hop questions with better recall than RAG or GraphRAG
- connect a knowledge base to coding agents via MCP
- migrate and back up knowledge bases across instances
- extract entities and relations from documents into a knowledge graph
- give agents shared context with citations

## When to choose
- you need semantic retrieval plus relational reasoning in one system
- you want a self-hosted knowledge base with traceable, cited answers
- you want to plug a knowledge base into LLM agents through MCP
- you are benchmarking retrieval on multi-hop QA datasets

## When to avoid
- you need a battle-tested, widely-adopted enterprise RAG stack
- you only need simple keyword or plain vector search without graph reasoning
- you require non-Python/non-Node runtimes or managed cloud deployment
- you need long-term stability guarantees from a mature project

## Facets
- artifact type: application
- maturity: active
- function: rag, search-engine, vector-database, agent-framework, mcp, nlp, machine-learning
- domain: artificial-intelligence, large-language-models, databases
- platform: python, cross-platform, self-hosted, cli
- tags: graphrag, knowledge-graph, event-entity-indexing, multi-hop-qa, retrieval-architecture, mcp-server, desktop-app, cli-tool, pypi-package, knowledge-base, retrieval-augmented-generation, ai-agents, search, data-engineering, natural-language-processing, nodejs, desktop

## Member repositories
- Zleap-AI/SAG (main) score 83

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:50.747917+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:31:49.819135+00:00, confidence not recorded.
  - readme: https://github.com/Zleap-AI/SAG (fetched 2026-08-28T04:06:50.747917+00:00, sha 90b3420fb488)
  - homepage: https://zleap.com (fetched 2026-08-29T10:13:06.754809+00:00, sha b63d55a8bce7)
- Data as of 2026-08-30T08:39:29.467469+00:00.