vector-graph-rag

Project · ★263 · python · MIT

Project / 项目 · In the catalogue / 在目录中查看

What Jev decides here · Jev 在这里做什么决定

Decision patterns / 决策模式
Search & ranking (检索与排序)

Primitives / 原语
No reading of its primitives is recorded. 没有记录原语的阅读结果。

Evidence / 证据
a person read the call site 有人读过调用点 · 2026-09-22 · Cited file / 引用的文件: src/vector_graph_rag/llm/jev.py

Summary · 摘要

Graph RAG that stores entities and relations as vectors in Milvus, no graph database. Two-stage Jev reranking asks a noul per candidate relation against the query, then scores the passages of the top relations; by the author's run, 86.61% average Recall@5 on MuSiQue and 2Wiki, 1,000 questions each.written by a model

把实体和关系作为向量存进 Milvus 的 Graph RAG，无需图数据库。两阶段 Jev 重排先针对查询给每条候选关系提一个 noul，再给排名靠前的关系所指向的段落打分；按作者的测试，在 MuSiQue 和 2Wiki（各 1,000 题）上平均 Recall@5 为 86.61%。模型撰写机翻

Launches · 发布记录

Not launched yet. 还没有发布过。 Launch it / 发布它 ↗ · How a product launches / 怎样发布一个产品

Makers · Maker

zilliztech

Made this? Confirm or correct its entry / 这是你做的？确认或更正条目 ↗

The same form takes the product's one line and a demo link, in its maker's own words. 同一个表单也可以补上产品的一句话和演示链接，用 Maker 自己的话。

Badge for its README · README 徽章

[![awesome-jev](https://kydlikebtc.github.io/awesome-jev/badges/vector-graph-rag.svg)](https://kydlikebtc.github.io/awesome-jev/products/vector-graph-rag.html)

One line for the product's README, linked to this page. The picture says that it is listed; while it is on the launch board, that it is launching; after a week in the top three, its place. It is rebuilt with the site, so the line never needs changing. 放进产品 README 的一行，链接到本页。图片显示已收录；在发布榜上时显示正在发布；进过某周前三后显示名次。它随网站一起重建，这一行不用再改。

Also under this pattern · 同一模式的其他产品

• AutoGPT TypeSafe blocks

  Seven production blocks — choice, score, yes/no, ask-many, route, pick-best, filter — with a UTF-8 byte budget, verbatim wire capture and eleven test files.

• OpenViking: retrieval reranking

  One Noul per candidate document in a single batched request, with the yes-probability used directly as the relevance score.

• oh-my-pi

  ⌥ Coding agent with the IDE wired in

• FastMCP jev_search transform

  Two-stage MCP tool search: a wide Choice coarse-ranks the whole catalogue, then a shortlist gets full descriptions plus one Noul each to decide whether it does the job at all.

• jcode: memory recall without embeddings

  Replaces the whole retrieval stack for memory recall — no embeddings, no BM25, no reranker — with one batched Noul per candidate memory.