Skip to content

fix: cherry-pick critical fixes from HKUDS/LightRAG v1.4.9.11#12

Merged
cleo-ia merged 3 commits intomainfrom
006-upstream-cherry-picks
Jan 16, 2026
Merged

fix: cherry-pick critical fixes from HKUDS/LightRAG v1.4.9.11#12
cleo-ia merged 3 commits intomainfrom
006-upstream-cherry-picks

Conversation

@cleo-ia
Copy link
Copy Markdown

@cleo-ia cleo-ia commented Jan 16, 2026

Summary

Cherry-picks des correctifs critiques de la release v1.4.9.11 du repo officiel HKUDS/LightRAG.

Bug fixes

Configuration (optionnel)

# PostgreSQL HA - valeurs par défaut augmentées
POSTGRES_CONNECTION_RETRY_ATTEMPTS=10
POSTGRES_CONNECTION_RETRY_BACKOFF=3.0
POSTGRES_CONNECTION_RETRY_BACKOFF_MAX=30.0

Test plan

  • Tous les tests passent
  • Vérifier le chargement des options LLM depuis les variables d'environnement
  • Tester la reconnexion PostgreSQL en cas de failover

🤖 Generated with Claude Code

cleo-ia and others added 3 commits January 16, 2026 10:36
Cherry-picked from HKUDS/LightRAG PR HKUDS#2585.

- Fix GlobalArgsProxy to support vars() for binding options extraction
- Refactor argument parsing for binding options to reduce duplication

Before: OPENAI_LLM_TEMPERATURE and similar env vars were silently ignored
After: Environment variables properly loaded and applied to LLM calls

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Cherry-picked from HKUDS/LightRAG PR HKUDS#2562.

Backend changes:
- connection_retry_attempts: max 10 → 100, default 3 → 10
- connection_retry_backoff: max 5s → 300s, default 0.5s → 3s
- connection_retry_backoff_max: max 60s → 600s, default 5s → 30s

Frontend changes:
- Context-aware timeouts: default 30s, uploads 120s, scans 90s

This enables seamless recovery during ~30s PostgreSQL HA switchovers
without manual intervention.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Cherry-picked from HKUDS/LightRAG PR HKUDS#2591.

Content duplicates now create trackable FAILED records instead of
being silently ignored:
- status="FAILED" with is_duplicate=true in metadata
- original_doc_id and original_track_id for cross-reference
- Enables clients to understand what happened to uploaded files

Before: Content duplicates returned empty track_status
After: Content duplicates are trackable with clear duplicate info

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
@cleo-ia cleo-ia merged commit 345ec49 into main Jan 16, 2026
3 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant