docs: RAG tuning guide — jina-embeddings-v5-text-small + qwen3-vl-8b SmartExtraction#304
Open
ordiy wants to merge 1 commit intoCortexReach:masterfrom
Open
docs: RAG tuning guide — jina-embeddings-v5-text-small + qwen3-vl-8b SmartExtraction#304ordiy wants to merge 1 commit intoCortexReach:masterfrom
ordiy wants to merge 1 commit intoCortexReach:masterfrom
Conversation
…3-vl-8b - Benchmark comparison: jina-embeddings-v3 vs v5-text-small on Chinese domain corpus (entity recall +3.1%, keyword phrase recall +14.8%) - SmartExtraction LLM comparison: qwen-2.5-7b vs qwen3-14b vs qwen3-vl-8b-instruct (qwen3-vl-8b: 35 completion tokens, 5x cheaper, Chinese fidelity preserved) - Documents the autoRecallTimeoutMs patch for hardcoded 3s timeout bug - Documents the default llm.model silent-failure bug (gpt-oss-120b on Jina baseURL) - Full model migration procedure including the reembed double-insert pitfall - Cross-version compatibility table (2026.3.8 vs 2026.3.13) - Covers both macOS (workspace/plugins) and Linux (extensions/) deployment paths Tested on: OpenClaw 2026.3.8 (Linux) and 2026.3.13 (macOS), memory-lancedb-pro v1.1.0-beta.9, Jina AI API, OpenRouter Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
This PR adds a practical tuning guide to
docs/based on real production deployments ofmemory-lancedb-prov1.1.0-beta.9 in OpenClaw 2026.3.8/2026.3.13 environments.What's included
New file:
docs/rag-tuning-jina-v5-openclaw.mdTopics covered
Embedding model comparison — jina-embeddings-v3 vs jina-embeddings-v5-text-small on a Chinese-language domain corpus
SmartExtraction LLM comparison — qwen-2.5-7b vs qwen3-14b vs qwen3-vl-8b-instruct
Embedding migration procedure — step-by-step including the
reembed --forcedouble-insert pitfallTwo bugs documented with patches:
AUTO_RECALL_TIMEOUT_MS = 3_000hardcoded — too tight for cloud embedding APIs (sed patch included)llm.model = "openai/gpt-oss-120b"silently fails against non-OpenAI base URLsCross-version compatibility table — 2026.3.8 vs 2026.3.13 CLI command availability
Testing
Benchmarks run on actual Jina API responses with real memory data. LLM comparisons measured via OpenRouter with identical prompts.
Style
Follows the existing
docs/openclaw-integration-playbook.mdformat: numbered sections, code blocks, practical rules over theory.