74 Commits

Author SHA1 Message Date
yangdx
44daf51501 Fix edge_degree to handle non-existent nodes 2025-06-27 01:37:10 +08:00
yangdx
6e1ca8f461 Merge branch 'patch-1' 2025-06-26 18:45:23 +08:00
yangdx
da46b341dc feat: Optimize document deletion performance
- To enhance performance during document deletion, new batch-get methods, `get_nodes_by_chunk_ids` and `get_edges_by_chunk_ids`, have been added to the graph storage layer (`BaseGraphStorage` and its implementations). The [`adelete_by_doc_id`](lightrag/lightrag.py:1681) function now leverages these methods to avoid unnecessary iteration over the entire knowledge graph, significantly improving efficiency.
- Graph storage updated: Networkx, Neo4j, Postgres AGE
2025-06-25 12:37:57 +08:00
pablo-statsig
4a8202138c
Remove graspologic pip installation that no longer seems to be used
As far as I can tell this is no longer actually used and its usage was removed in this commit:
83353ab9a6 (diff-a346bcfb05aab0cc0c0baa6018976f4efab339e8cade9f6f8fb658fcbd54ae2e)

Our systems are flagging this package as having a dependency on a package with a less permissive license so I would appreciate if it can be removed if its no longer needed.  Let me know if that is not the case.
2025-06-05 18:01:00 -07:00
yangdx
9ec7f5c8b3 Fix degree sorting problem in BFS 2025-04-25 11:25:29 +08:00
yangdx
bd11bcae32 Fix linting 2025-04-25 09:43:18 +08:00
yangdx
c620f9c4f2 Prioritize high-degree neighbors in BFS traversal for NetoworkX storage 2025-04-25 09:22:53 +08:00
yangdx
d4c4a40c53 Fix M.env AX_GRAPH_NODES not working problem 2025-04-17 01:28:22 +08:00
yangdx
745301ea13 Deleted node2vec implementation 2025-04-11 18:41:45 +08:00
yangdx
83353ab9a6 Remove unused node embedding functionality from graph storage
- Deleted embed_nodes() method implementations
2025-04-11 18:34:48 +08:00
yangdx
ad087073aa Optimize logger for storage 2025-04-10 01:07:06 +08:00
yangdx
84d6dee34d Fix linting 2025-04-03 00:14:33 +08:00
yangdx
6d942da338 Fix BFS depth control in NetworkX graph traversal 2025-04-02 23:59:19 +08:00
yangdx
72132ee1d6 Added info logs when graph is truncated 2025-04-02 23:19:41 +08:00
yangdx
82c4baba70 Add is_truncated to graph query for NetworkX graph db 2025-04-02 22:12:20 +08:00
yangdx
4ceafb7cbc Optimize NetworkX subgraph query 2025-04-02 21:41:24 +08:00
yangdx
95a8ee27ed Fix linting 2025-03-31 23:22:27 +08:00
yangdx
1772e7a887 Add delete support to all storage implementation 2025-03-31 16:21:20 +08:00
yangdx
1df4b777d7 Add drop funtions to storage implementations 2025-03-30 15:17:57 +08:00
yangdx
15e060f854 Fix share storage update status handling problem of in memeory storage 2025-03-25 10:48:15 +08:00
yangdx
ff9cb2138d Ensure thread safety in storage update callbacks
- Added storage lock in index_done_callback
- Fixed potential race conditions
2025-03-24 02:11:59 +08:00
yangdx
c641a6290c Ensure unique edge IDs for NetworkX 2025-03-18 00:00:56 +08:00
yangdx
6d5c04d561 Reduce max_depth and update edge type in NetworkXStorage.
- Decreased max_depth from 5 to 3
- Changed edge type from "RELATED" to "DIRECTED"
2025-03-06 20:54:02 +08:00
yangdx
5b29e760f0 Add dynamic parameter handling for storage 2025-03-06 14:29:48 +08:00
yangdx
0ca7c0e230 Fix linting 2025-03-05 15:07:11 +08:00
yangdx
4e2a9b1994 Apply min_degree filter for full graph query 2025-03-05 13:21:17 +08:00
yangdx
ca9e958cad Fix subgraph filtering bugs 2025-03-05 13:13:46 +08:00
yangdx
5f12aa3ee4 Fix linting 2025-03-05 12:32:23 +08:00
yangdx
5e40e4107d Added min_degree exception for connected nodes 2025-03-05 11:48:04 +08:00
yangdx
1fddc8552e Added minimum degree filter for graph queries
- Introduced min_degree parameter in graph query
- Updated UI to include minimum degree setting
- Modified API to handle min_degree parameter
- Updated graph query logic in LightRAG
2025-03-05 11:37:55 +08:00
yangdx
002948d342 Added search mode and min degree filtering for NetworkX
- Implemented exact and inclusive search modes
- Added min degree filtering for nodes
- Updated API to parse label for search options
2025-03-04 16:08:05 +08:00
yangdx
11fdb60fe5 Remove Chinese comments 2025-03-03 01:30:41 +08:00
yangdx
465737efed Fix linting 2025-03-02 17:32:25 +08:00
yangdx
68bf02abb6 refactor: improve graph querying with label substring matching and security fixes 2025-03-02 16:20:37 +08:00
yangdx
0f1eb42c8d Add node limit and prioritization for knowledge graph retrieval
• Add MAX_GRAPH_NODES limit from env var
• Prioritize nodes by label match & connection
2025-03-02 15:39:14 +08:00
yangdx
1ca6837219 Add max nodes limit for graph retrieval of networkX
• Set MAX_GRAPH_NODES env var (default 1000)
• Change edge type to "RELATED"
2025-03-02 12:52:25 +08:00
yangdx
e3a40c2fdb Fix linting 2025-03-01 16:23:34 +08:00
yangdx
d4f6dcfd54 Improve multi-process data synchronization and persistence in storage implementations
• Remove _get_client() or _get_graph() from index_done_callback
• Add return value for index_done_callback
2025-03-01 12:41:30 +08:00
yangdx
d3de57c1e4 Add multi-process support for vector database and graph storage with lock flags
• Implement storage lock mechanism
• Add update flag handling
• Add cross-process reload detection
2025-03-01 10:37:05 +08:00
yangdx
cd7648791a Fix linting 2025-02-28 01:25:59 +08:00
yangdx
291e0c1b14 revert vector and graph use local data(single process) 2025-02-28 01:14:25 +08:00
yangdx
05d03638ec Clean up logging output and remove redundant log messages 2025-02-27 20:17:28 +08:00
yangdx
64f22966a3 Fix linting 2025-02-27 19:05:51 +08:00
yangdx
1699b10a25 Refactor direct client/graph access to reduce redundant get calls in vector/graph ops 2025-02-27 15:14:54 +08:00
yangdx
f007ebf006 Refactor initialization logic for vector, KV and graph storage implementations
• Add try_initialize_namespace check
• Move init code out of storage locks
• Reduce redundant init conditions
• Simplify initialization flow
• Make init thread-safer
2025-02-27 14:55:07 +08:00
yangdx
7436c06f6c Fix linting 2025-02-26 18:11:16 +08:00
yangdx
145bacc773 Add empty graph creation logging in NetworkXStorage 2025-02-26 17:42:30 +08:00
yangdx
2752a764ae Refactor storage implementations to support both single and multi-process modes
• Add shared storage management module
• Support process/thread lock based on mode
2025-02-26 05:38:38 +08:00
yangdx
a642bb3190 refactor: use shared manager from main process for storage implementations. 2025-02-25 12:08:49 +08:00
yangdx
362321204f Merge branch 'main' into add-multi-worker-support 2025-02-25 11:15:12 +08:00