graph-rag · LightRAG service and data flow

Persistent entity/relationship extraction with profile-driven graph, vector, and chunk retrieval

graph-rag LightRAG service and data flowDocuments are extracted into persistent LightRAG graph and vector stores, then queried through keyword and answer model roles. ATLAS COMPOSE STACK · LIGHTRAG SERVICE + RAG PLUGININDEXING · PERSISTENT KNOWLEDGE CONSTRUCTIONQUERY · PROFILE-SELECTED MODE AND BUDGETS LIGHTRAG PERSISTENT STORES optional TEI adapter bypass8 · answer-only OpenAI response returns through plugin and LiteLLM Full documentsAtlas ingestion job LightRAG /documentschunk + extraction orchestration EXTRACT model roleentities · relations · summaries Neo4jentity/relation graph Vector storageentities · relations · chunks KV / doc statustext units · pipeline state Graph/vector lookupprofile-bounded evidence CallerOpen WebUI / API LiteLLMgraph-rag profile alias Graph pluginselect profile · normalize LightRAG /querylocal · global · hybrid · mix KEYWORD rolequery decomposition TEI adapteroptional rerank QUERY model roleevidence → answer 1request2profile alias 3profile params4/query 5keywords6contexts7ranked MESSAGE CONTRACT, STATE, AND TUNING 0 · EXTRACT + PERSISTfull docs → chunks → entities/relationshipsGraph is created during ingestion and reused.Rebuild when extraction inputs/config change. 1–4 · PROFILE DISPATCHLiteLLM alias identifies a query profile.Mode, top-k, chunk-k, and token budgets vary.The profiles share one ingested workspace. 5–7 · RETRIEVE + RERANKLightRAG selects graph/vector evidence internally.No showcase-controlled fixed k-hop traversal.The query profile selects optional TEI reranking. 8 · GENERATE + RETURNQUERY role synthesizes the final answer.Exact retrieved contexts are not exposed.Plugin returns answer and service metrics.