knowledge-graph on tomrochette.com

Code-Graph-RAG

tom@tomrochette.com (Tom Rochette) — Wed, 07 Oct 2026 00:00:00 +0000

Code-Graph-RAG is an MIT-licensed Python system that parses a multi-language codebase with tree-sitter plus compiler-grade frontends into a Memgraph knowledge graph and answers natural-language questions and edit requests through an agent that writes Cypher.

It pushes code retrieval past similarity search entirely: functions, classes, calls, and imports live in a graph, so the questions it answers well are structural ones (what calls this, what breaks if this changes) that embeddings index cannot address at all.

What it is
#

A cgr CLI (PyPI code-graph-rag) with two components. The parser layer reads every source file with tree-sitter, sharpens facts with compiler frontends where a toolchain exists (libclang for C and C++, go/types for Go, opt-in Roslyn, javac, and Jedi for C#, Java, and Python), and can overlay runtime behavior by tracing a test run or pulling production eBPF profiles into the graph. The RAG layer turns natural language into Cypher over Memgraph, retrieves the matching code, and drives AST-based surgical patching with a diff preview. Python, TypeScript, JavaScript, Rust, Go, Java, C, C++, C#, PHP, Lua, and Dart are fully supported, with more languages on a pluggable ast-grep tier, under one language-agnostic schema across a monorepo. Made by an independent developer, with an enterprise support and services badge on the README.

Status
#

Actively shipping and early: 5.2k stars in sixteen months with almost no independent discussion. 5,233 stars and 699 forks since 2025-06-16, pushed 2026-10-07, with 587 open issues (GitHub API, as of 2026-10-07).

PyPI shows version 0.1.38 released 2026-09-29 across 27 releases. Two Show HNs (February and March 2026) drew 1 point each, so the star count ran far ahead of any independent technical debate, the same pattern Knowhere and open-codebase-index in this section show.

Strengths
#

Blast-radius questions are its home turf: call graphs, reference walks, and dead-code discovery from entry points are native graph queries, not retrieval approximations.
The layered parser is the right architecture: tree-sitter for coverage, compiler frontends for precision where available, runtime traces for the dispatch static analysis cannot see.
One unified graph schema across a polyglot monorepo is the case where per-repo embeddings pipelines degrade worst.

Cautions
#

Zero independent discussion and 587 open issues against 5.2k stars means the only quality evidence is the project’s own demos and CI.
It brings a graph database: Memgraph is a separate dependency with its own licensing and operational cost, which the one-line install pitch understates.
0.1.x with 27 releases means churn, and the Cypher-generation quality depends on the model you attach, which no benchmark here measures.
Indexing is per-repo work (cgr start --update-graph), so freshness after agent edits needs the incremental update path to actually hold.

Pricing
#

Free and open source under MIT, with enterprise support and services quoted by the developer. Your costs are the Memgraph deployment and your own LLM bills for query answering, so no prices belong in this note and no price history table applies.

Compared to
#

open-codebase-index: the embeddings-plus-BM25 sibling over MCP, host-neutral and lighter; choose it for concept lookup, Code-Graph-RAG for structural and blast-radius questions.
Aider’s graph-ranked repo map (Aider): the no-store middle ground, ranking tree-sitter symbols by references without a database; choose it when a Memgraph deployment is too much.
Tree-sitter code chunking: the parsing layer alone, feeding a vector store; Code-Graph-RAG keeps the parse and ditches the store.

Bottom line
#

Recommended for monorepo teams whose retrieval questions are structural, who will operate Memgraph, and who will validate answer quality on their own repo before trusting it. Not for concept-level search, not as a dependency-free tool, and not for anyone needing independent evidence first, because none exists. My disagreeable claim: this is the version of code retrieval the embeddings era skipped, and its near-zero discussion footprint says more about how few teams measure retrieval structurally than about the tool.

Changes
#

2026-10-07 - Created in the daily refresh’s retrieval entrant scan from the awesome-rag-production source.

References
#

https://api.github.com/repos/vitali87/code-graph-rag - 5,233 stars, 699 forks, MIT, created 2025-06-16, pushed 2026-10-07, 587 open issues (GitHub API, as of 2026-10-07)
https://raw.githubusercontent.com/vitali87/code-graph-rag/main/README.md - the two-component architecture, compiler frontends, runtime tracing, language support, Memgraph storage
https://raw.githubusercontent.com/vitali87/code-graph-rag/main/NEWS.md - the feature-news ledger behind the active-maintenance claim
https://pypi.org/pypi/code-graph-rag/json - version 0.1.38 (2026-09-29), 27 releases
https://hn.algolia.com/api/v1/search?query=Code-Graph-RAG%20knowledge%20graph&tags=story - the footprint scan: two Show HNs at 1 point each (2026-02 and 2026-03)

GraphRAG

tom@tomrochette.com (Tom Rochette) — Wed, 07 Oct 2026 00:00:00 +0000

GraphRAG is Microsoft Research’s MIT-licensed Python pipeline that builds an LLM-extracted entity knowledge graph over a text corpus and answers queries from community summaries, the reference implementation of the graph-based RAG pattern.

It founded the graph-RAG family and answered the question vector RAG answers worst (what are the themes across this whole corpus), and its own README now declares the project largely in maintenance mode, which makes it a pattern to study rather than a dependency to adopt.

What it is
#

A pip install graphrag CLI and library that chunks a corpus, prompts an LLM to extract entities, relationships, and claims per chunk, detects communities over the graph, and pre-generates community summaries for global questions. It answers two query styles: local search (entity-anchored neighborhood questions) and global search (map-reduce over community summaries). The method paper (arXiv 2404.16130) frames it as moving from local to global sense-making on narrative private data. Made by Microsoft Research, first released July 2024, optional Azure OpenAI or local model support.

Status
#

Active by the commit clock, retired by its own README. 36,241 stars and 3,834 forks since 2024-03-27, pushed 2026-10-05, 49 open issues (GitHub API, as of 2026-10-07).

PyPI shows version 3.2.0 released 2026-09-23 across 50 releases, so bugfix and dependency trains still run. The README’s warning is the status fact that matters: the project is largely in maintenance mode, it will not accept new PRs or implement new features, and it names the dramatic change in frontier-model capabilities and a diversified research portfolio as the cause. The launch thread drew 282 points in July 2024, the largest graph-RAG footprint on Hacker News.

Strengths
#

Global sense-making is its product: community summaries answer “what are the themes across this corpus” questions that similarity search structurally cannot.
It is the citation root of the pattern: LightRAG, nano-graphrag, and most graph-RAG work name it.
MIT licensed with a CLI, a Python API, and docs deep enough to replicate the pipeline.

Cautions
#

Maintenance mode means the feature freeze is explicit: budget to own whatever you build on it, because upstream has said it is done.
Indexing is expensive by the project’s own warning, an LLM call set per chunk over your whole corpus.
Version bumps carry breaking config changes, and the README tells you to re-init between minor versions.
Nothing agent-facing: no MCP server, no retrieval tools for a coding harness, the consumer is your own application.

Pricing
#

Free and open source under MIT. The costs are your own LLM bills for indexing and querying, which the project’s documentation warns can be large on big corpora, so no prices belong in this note and no price history table applies.

Compared to
#

LightRAG: the actively maintained graph-RAG counterweight, dual-level retrieval instead of community summaries; pick it when you want graph RAG with a living release train.
RAGFlow: the deployable engine with citations and a UI; graph structure is one artifact kind among its outputs, not the core.
LlamaIndex: property-graph indexes give you the graph option inside a general framework, at the cost of assembling the pipeline yourself.

Bottom line
#

Recommended for studying what graph RAG is and for one-shot theme extraction over a stable narrative corpus you can afford to index. Not for code, not for fast-changing corpora, and not as a dependency bet now that the maker has declared maintenance mode. My disagreeable claim: the maintenance-mode declaration is graph RAG’s candid status report, the frontier ate the gap it exploited, and the pattern’s heirs now argue about which parts survive as plain long-context questions.

Changes
#

2026-10-07 - Created in the daily refresh’s retrieval entrant scan from the awesome-rag-production source.

References
#

https://api.github.com/repos/microsoft/graphrag - 36,241 stars, 3,834 forks, MIT, created 2024-03-27, pushed 2026-10-05, 49 open issues (GitHub API, as of 2026-10-07)
https://raw.githubusercontent.com/microsoft/graphrag/main/README.md - the maintenance-mode warning, the expensive-indexing warning, the init-between-versions rule
https://arxiv.org/abs/2404.16130 - the method paper: From Local to Global, graph RAG as query-focused summarization
https://www.microsoft.com/en-us/research/blog/graphrag-unlocking-llm-discovery-on-narrative-private-data/ - the Microsoft Research announcement grounding community summaries and the local/global query split
https://microsoft.github.io/graphrag/ - official docs root: local and global search, CLI quickstart
https://pypi.org/pypi/graphrag/json - version 3.2.0 (2026-09-23), 50 releases, MIT
https://hn.algolia.com/api/v1/search?query=GraphRAG%20is%20now%20on%20GitHub&tags=story - the July 2024 launch thread, 282 points, 49 comments

LightRAG

tom@tomrochette.com (Tom Rochette) — Wed, 07 Oct 2026 00:00:00 +0000

LightRAG is the HKU Data Intelligence Lab’s MIT-licensed Python framework that extracts entities and relations into a knowledge graph and answers queries with dual-level retrieval, the EMNLP 2025 paper behind the most-starred graph-RAG implementation.

It took the graph-RAG pattern Microsoft parked in maintenance mode and turned it into a maintained framework with pluggable storage and a growing multimodal stack, and that velocity is both its appeal and its risk.

What it is
#

A pip install lightrag-hku framework: an LLM pass extracts entities and relations into a graph, and queries run at two levels, low-level retrieval over specific entities and high-level retrieval over topics and themes. Storage is pluggable across Neo4j, PostgreSQL, MongoDB, and OpenSearch (the unified backend added March 2026), with a WebUI that visualizes the graph, citation support, a default reranker for mixed queries, and four selectable chunking strategies (fixed, recursive, vector, paragraph). RAG-Anything merged in May 2026 brought multimodal parsing and extraction through MinerU or Docling services, and role-specific LLM configuration (extraction, query, keywords, vision) landed the same month. Made by HKUDS, the lab that also ships RAG-Anything, VideoRAG, and MiniRAG.

Status
#

Actively maintained with a large, fast-moving community. 40,003 stars and 5,665 forks since 2024-10-02, pushed 2026-10-03, with 359 open issues (GitHub API, as of 2026-10-07).

PyPI shows version 1.5.7 released 2026-09-02 across 94 releases, and the README’s news line runs to July 2026 (Smart Heading recognition for Word documents). The paper (arXiv 2410.05779) was published at EMNLP 2025. The launch thread drew 82 points in July 2024 and little HN activity since, so the star count grew through tutorials and the lab’s ecosystem rather than launch-driven debate.

Strengths
#

The complete graph-RAG loop in one install: extraction, dual-level query, graph storage of your choice, visualization, and citations, where GraphRAG gives you the pipeline and stops.
Storage portability is unusual for the family: one framework over Neo4j, PostgreSQL, MongoDB, or OpenSearch keeps the graph out of a vendor.
An academic paper plus a large community means the method is documented and the failure modes are discussed in public issues.

Cautions
#

The lab’s breadth is the roadmap risk: RAG-Anything, VideoRAG, and MiniRAG share one maintainership, and 359 open issues say the core absorbs more than it resolves.
The accuracy story is self-run; I found no independent benchmark replicating the paper’s numbers on a different corpus.
The feature surface churns (chunking strategies, merged multimodal parsing, role-specific models), so pin versions and re-test on upgrades.
Setup is a stack: an embedding model, an LLM, a graph or unified store, and optionally parsing services.

Pricing
#

Free and open source under MIT. The costs are your own model bills for extraction and querying plus the storage you attach, so no prices belong in this note and no price history table applies.

Compared to
#

GraphRAG: the origin in maintenance mode; LightRAG is where the pattern’s active development lives.
RAGFlow: the deployable engine with a UI and cloud tiers; choose it when you want citations and chunk inspection without assembling storage yourself.
Graphify: the context-engine angle, building graph structure an agent reads, rather than a queryable RAG framework.

Bottom line
#

Recommended as the default graph-RAG framework when the graph layer genuinely earns its indexing cost, which for most corpora means theme and multi-hop questions, not lookup. Not for teams that need a frozen dependency or independent accuracy evidence, because it offers neither. My disagreeable claim: LightRAG’s star count measures the graph-RAG idea’s appeal, not adoption depth, and the 359 open issues are the truer count of how much work the pattern still demands.

Changes
#

2026-10-07 - Created in the daily refresh’s retrieval entrant scan from the awesome-rag-production source.

References
#

https://api.github.com/repos/HKUDS/LightRAG - 40,003 stars, 5,665 forks, MIT, created 2024-10-02, pushed 2026-10-03, 359 open issues (GitHub API, as of 2026-10-07)
https://raw.githubusercontent.com/HKUDS/LightRAG/main/README.md - dual-level retrieval, storage backends, chunking strategies, the RAG-Anything merge, the news timeline
https://arxiv.org/abs/2410.05779 - the EMNLP 2025 paper: LightRAG, simple and fast retrieval-augmented generation
https://pypi.org/pypi/lightrag-hku/json - version 1.5.7 (2026-09-02), 94 releases, MIT
https://learnopencv.com/lightrag/ - the widely-cited third-party guide grounding the tutorial-driven adoption pattern
https://hn.algolia.com/api/v1/search?query=LightRAG&tags=story - the footprint scan: the 82-point 2024 launch thread and thin activity since

knowledge-graph on tomrochette.com

Code-Graph-RAG

What it is
#

Status
#

Strengths
#

Cautions
#

Pricing
#

Compared to
#

Bottom line
#

Changes
#

See also
#

References
#

GraphRAG

What it is
#

Status
#

Strengths
#

Cautions
#

Pricing
#

Compared to
#

Bottom line
#

Changes
#

See also
#

References
#

LightRAG

What it is
#

Status
#

Strengths
#

Cautions
#

Pricing
#

Compared to
#

Bottom line
#

Changes
#

See also
#

References
#

knowledge-graph on tomrochette.com

Code-Graph-RAG

What it is #

Status #

Strengths #

Cautions #

Pricing #

Compared to #

Bottom line #

Changes #

See also #

References #

GraphRAG

What it is #

Status #

Strengths #

Cautions #

Pricing #

Compared to #

Bottom line #

Changes #

See also #

References #

LightRAG

What it is #

Status #

Strengths #

Cautions #

Pricing #

Compared to #

Bottom line #

Changes #

See also #

References #

What it is
#

Status
#

Strengths
#

Cautions
#

Pricing
#

Compared to
#

Bottom line
#

Changes
#

See also
#

References
#

What it is
#

Status
#

Strengths
#

Cautions
#

Pricing
#

Compared to
#

Bottom line
#

Changes
#

See also
#

References
#

What it is
#

Status
#

Strengths
#

Cautions
#

Pricing
#

Compared to
#

Bottom line
#

Changes
#

See also
#

References
#