find_similar_standards

Use find_similar_standards to retrieve standards whose stored embeddings are closest to a source standard. The tool runs a pgvector cosine search and returns nearest neighbors; it does not decide alignment quality or explain matches. Use it to seed crosswalk review, find duplicate or near-duplicate standards, or give an agent candidate standards before applying stricter operator rules.

Parameters

Name
source_standard_id
Type
string
Description
The WKW standard identifier to use as the embedding query. It must refer to an indexed standard; the tool compares other standards against this source vector.
Name
target_framework_guids
Type
string[]
Description
Optional list of framework GUIDs to limit the neighbor search. Omit this field to search across all indexed target frameworks; pass one or more GUID strings to restrict results to those frameworks.
Name
k
Type
number
Description
Maximum number of neighbors to return. Defaults to 10; use a small value for agent workflows that will review or rank candidates downstream.

Example

{
  "source_standard_id": "std_ccss_math_4_nf_a_1",
  "target_framework_guids": [
    "8d68f0d8-7f9a-4c1f-a6e6-9c3d4e1b6d90",
    "3b1f4a2c-2dd6-49e0-9f2b-5ef0df7f0a11"
  ],
  "k": 5
}

Common pitfalls

Do not treat high similarity as an approved alignment; this is vector proximity only, with no rubric or LLM judgment.
Use framework GUIDs, not framework names or codes, in target_framework_guids.
If the source standard has no embedding or the target framework is not indexed, the result set may be empty.