ContextQMD
Back to Weknora

`search chunks` — hybrid retrieval (no LLM)

cli/skills/weknora-rag-search/references/search-chunks.md

0.6.21.9 KB
Original Source

search chunks — hybrid retrieval (no LLM)

Raw vector + keyword retrieval against one knowledge base. Returns ranked chunks for you to reason over; it does NOT synthesize an answer (use chat for that).

Command & flags

weknora search chunks "<query>" --kb <name-or-id> [flags]
FlagDefaultMeaning
--kb(required)KB name or UUID
--limit, -L8max chunks returned (1..1000); 8 is tuned for an LLM context window
--vector-threshold0 (off)min vector similarity, per-channel pre-fusion
--keyword-threshold0 (off)min keyword score, per-channel pre-fusion
--no-vectorfalsedisable the vector channel (keyword-only)
--no-keywordfalsedisable the keyword channel (vector-only)

You cannot disable both channels. --limit is a hard cap on returned chunks applied client-side (the server may internally retrieve a larger pool for recall, then the CLI trims).

Output (--format json)

data is an array of chunk objects; meta.count is the number returned.

jsonc
{"ok":true,"meta":{"count":3},"data":[
  {"id":"chunk_…","content":"…","knowledge_id":"doc_…","knowledge_title":"…",
   "chunk_index":4,"score":0.82,"match_type":"hybrid","chunk_type":"text"}
]}
  • score is the fused rank; match_type indicates which channel(s) hit.
  • knowledge_id / knowledge_title attribute the chunk to its source document.
  • Project just what you need with --jq, e.g. weknora search chunks "q" --kb eng --jq '.data[] | {score,content}'.

When to use vs alternatives

  • Need an answer → chat (it does retrieval + synthesis internally).
  • Building your own prompt/context from sources → search chunks (this).
  • Finding which documents exist by keyword → search docs --kb <kb>.
  • Inspecting/debugging a specific document's chunks → chunk list --doc <id>.