#Issue: memory-core plugin ignores chunking config.

1 messages · Page 1 of 1 (latest)

knotty crown
#

Behavior: Sends full 9k+ token files to embedding model (limit 8192).

Impact: Fails on both OpenAI and Local providers.

Fix: Needs patch in @openclaw/plugin-memory/indexer.ts to enforce chunkSize before API call.

maiden stratusBOT
#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

knotty crown
#

Summary of what we found: Its actaully affecting all models:

  • Core issue: OpenClaw's memory-core plugin completely ignores chunking.tokens config
  • Affects all providers: OpenAI, Gemini, Ollama - same behavior
  • Root cause: Indexer sends full files (9k+ tokens) instead of 500-token chunks
  • Impact: Semantic search completely broken across the board
split willow
#

@knotty crown Any bandage solution? Had the same issue, wondering if disable memory plugin, will remote embedding still work?

knotty crown