Repository navigation
feat(llm): extract standalone graph design and extraction SDK - #387
Merged
Merged
Conversation
- Integrate the merged MCP feature and documentation cleanup. - Adopt the current upstream Ruff workflow. - Preserve fork history with a merge commit and an identical upstream tree.
Release factory-owned clients in bounded groups so long batches do not exhaust descriptors. Drain sibling clients when aclose itself cancels, while preserving caller ownership and cleanup-thread semantics. Distinguish class construction from async instance calls and reject explicit edge frequency/sort-key/nullable-key contradictions before inference. Preserve omitted defaults and legacy custom-ID metadata. Add 35 regression cases. Core: 646 passed. Isolated Python 3.10-3.13 wheels: 591 passed, 55 skipped per version. Legacy CI selection: 1334 passed, 3 skipped, 18 deselected; coverage 63.62%. Addresses the four PR #84 comments published before this repair.
Remove the Tenacity wrapper that only retried generator creation. Clarify streaming delta and replay behavior, and cover transient provider and callback failures without changing non-streaming retries. Record the second independent review, published comments, and source and installed-wheel validation.
Drain shielded default-executor futures before propagating cancellation, preserving ContextVars and retrieving late failures. Cover real-resource ownership, repeated cancellation, and asyncio.run shutdown. Document cancellation latency and record independent Python 3.10-3.13 installed-wheel validation.
imbajin
marked this pull request as ready for review
October 28, 2026 07:47
MrJs133
approved these changes
Oct 28, 2026
This was referenced Oct 28, 2026
Closed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Before → After
Before: graph schema design and extraction require importing the application and global configuration.
After: provide an independently installable
hugegraph-llm-coreSDK with a Pydantic-only base dependency. Share one implementation through compatibility adapters while exposing synchronous/asynchronous design, extraction, validation and bounded repair with caller-supplied models and workflow steps. Keep UI, persistence and application configuration outside the SDK.Validation boundary
The author reports core, isolated-wheel Python 3.10–3.13, optional-adapter and legacy compatibility tests. Base-wheel environments intentionally skip optional-provider cases. The author now reports a source-built pycgraph ABI fix and local HTTP container smoke checks; these have not been independently rerun. Retain draft status. No package publication, external model calls or graph writes are claimed.
Source: 🔗 hugegraph/hugegraph-ai #84
Visual overview
Standalone graph-design/extraction SDK and application boundaries. The author reports an ABI fix and local HTTP container smoke checks; these were not independently rerun. No package publication, real-model invocation, graph write or RC acceptance is implied.
怎么用,怎么改
传入文本和自己的模型,即可生成 schema 草案;业务修改并校验 schema 后,再抽取属性图。同步与异步入口都支持,返回图数据和按块排列的问题,便于业务决定接受、重试或人工处理。
可以直接使用完整流程,也可以只替换模型、提示词、校验规则或抽取/审阅/修复步骤。默认执行本地校验与有限审阅修复;关闭审阅后,结果明确标记为

validated,方便调用方区分质量保障程度。接入指南与 editable 开发方式 · 最小调用示例