Repository Issue Activity (beta)

#3681 [Bug]: FaissVectorDBStorage.delete is eager — quadratic merge cost and a delete lost across writers

#3664 RFC: Add C strategy for Server-injected custom chunking callbacks

#3603 Per-chunk text "breadcrumb" / metadata hook — `chunking_func` is silently bypassed on `/documents/scan` & `/documents/upload`

#3706 Per-document as-of date / metadata channel — `create_time` is ingestion, not content; the date has nowhere to live

#3679 [Bug]:[Performance] O(n²) ingestion slowdown: per-relation delete() copies the entire vector matrix (NanoVectorDB)

#3695 Gemini binding silently drops the generic max_tokens kwarg

#3694 [Question]: Security vulnerability

#3657 [Feature Request]: Add an `openai_responses` LLM binding (OpenAI Responses API)

#3691 [Feature Request]: Optional pre-merge extraction validator hook (post-extract, before merge)

#3688 [Bug]: index_done_callback's unconditional reload silently drops applied-but-unsaved ops (Nano/Faiss)

#1957 [Question]: Merge scale performance issues

#3609 Chunk-tracking reseed: `has_stored_data` truthiness check conflates an absent row with an empty one

#3612 get_knowledge_graph: BFS same-depth truncation is not degree/label ordered on neo4j, memgraph, mongo, opensearch

#285 LightRAG Chat (微信交流群)

#1866 [Bug]:ERROR: graph "space1_chunk_entity_relation" already exists in `Server`

#3673 amerge_entities loses relation evidence-count: weight merges max instead of tracking distinct sources

#3059 [Question]:Support for separate LLM configurations for indexing and query phases

#3425 [Feature Request]:Native YouTube Video and Transcript Ingestion Pipeline

#3671 [RFC] ENABLE_API_DOCS switch for /docs, /redoc, /openapi.json exposure + move WebUI API tab to a header icon

#3666 [Question]:服务启动后有swagger接口文档暴露问题

#3659 RFC: single-pass deferred KG rebuild for batch document deletion

#3516 [RFC] Authorized API access and document artifact downloads

#2340 [Bug]:Error:'utf-8' codec can't encode character '\udc9a' in position 201: surrogates not allowed，上传所有文件失败

#3647 Stable extension point for the server-mode LLM completion call-site — or a first-class Codex (ChatGPT OAuth) binding?

#3579 Proposal: build a verified cold-migration utility from `PGGraphStorage` (Apache AGE) to `PGTableGraphStorage` (plain PostgreSQL tables)