Skip to content
HotKVDocs

Ask the HotKV docs

Answers come only from these docs and can be wrong; check the linked pages before you rely on them.

RAG.SEARCH

Last updated 
RAG.SEARCH pipeline query_text [K n] [ALPHA weight] [VECTOR vector] [FILTER key value [key value ...]]
Group
RAG
Complexity
O(N) where N is the number of chunks
Arity
-3 (at least 3 words, counting the command name)
Flags
read-only, slow
Keys
first 1, last 1, step 1
HotKV Cloud
Performance and Scale
On-Premise
requires the rag-pipeline license feature

A RAG pipeline splits your documents into chunks, keeps each chunk with the embedding you computed for it, and answers searches that combine keyword (BM25) relevance with vector similarity, all inside HotKV. RAG.SEARCH scores chunks by keyword relevance to query_text and, when you pass VECTOR, by similarity to the query embedding, then returns the best K chunks ranked by the combined score.

Options

Section titled “Options”
Option Effect
K n Number of results. Must be positive. Default 5.
ALPHA weight How much the vector score counts against the keyword score, from 0.0 (keywords only) to 1.0 (vectors only). Default 0.5.
VECTOR vector The query embedding as one argument of comma-separated numbers, with the pipeline’s dimensions.
FILTER key value ... Only chunks whose metadata has every given key with exactly the given value.

Return value

Section titled “Return value”
Case RESP2 RESP3
Always Array of results, best first. Empty for an unknown pipeline or no matches. Same

Each result is an array of bulk strings: the chunk id, the document id, the chunk text, the combined score, the vector score, the keyword (BM25) score, followed by the chunk’s metadata as alternating keys and values. Scores are decimal strings with six digits after the point; higher is better.

Errors:

  • ERR invalid K if K is not a positive integer.
  • ERR invalid ALPHA (must be 0.0 to 1.0) if ALPHA is out of range or not a number.
  • ERR expected N dimensions, got M if VECTOR has the wrong number of values.
  • ERR wrong number of arguments for 'rag.search' command if the argument count is wrong.
  • NOLICENSE rag-pipeline requires a HotKV license. Contact sales@hotkv.com if your license does not include the rag-pipeline feature.

Examples

Section titled “Examples”
127.0.0.1:6379> RAG.SEARCH faq "how do I reset my password" K 2 VECTOR 0.88,0.12,0.01
1) 1) "doc:1:0"
   2) "doc:1"
   3) "To reset your password, click Forgot password on the login screen."
   4) "1.000000"
   5) "0.999700"
   6) "1.386294"
   7) "source"
   8) "help-center"
2) 1) "doc:1:1"
   2) "doc:1"
   3) "Billing questions go to the billing team."
   4) "0.255100"
   5) "0.510250"
   6) "0.000000"
   7) "source"
   8) "help-center"

The score values depend on your documents and embeddings.

Compatibility

Section titled “Compatibility”

HotKV-only command. Redis and Valkey have no equivalent.

See also

Section titled “See also”

RAG.CREATE, RAG.INGEST, RAG.STATS, RAG.DROP, RAG.LIST, RERANK.GET, AI features guide