vector search cost anomaly error query loop spikes bill fails

Not specified by the original author.

TL;DR: Vector search is slow or returns junk. The causes are almost always index config: dimensions, candidates, readiness, or filter fields. Too low and recall collapses (good docs never surface); way too high and latency balloons. Confirm Run the same query with numCandidates raised 10x.

The fix

TL;DR: Vector search is slow or returns junk. The causes are almost always index config: dimensions, candidates, readiness, or filter fields. Too low and recall collapses (good docs never surface); way too high and latency balloons. Confirm Run the same query with numCandidates raised 10x.

The error

`$vectorSearch` is slow, returns irrelevant results, or errors.

The fix

1. getSearchIndexes(): is the index READY? BUILDING explains errors and empty results.
2. numDimensions in the index vs your embedding length. A mismatch breaks relevance silently or errors loudly; count the vector you send.
3. numCandidates vs limit. Too low and recall collapses (good docs never surface); way too high and latency balloons.
4. filter fields: are they declared type: "filter" in the index? Undeclared filters either error or get ignored.

When to use this

• You are seeing this exact error message; match the block above, not just part of it.
• The failing call matches the scenario in the title: Atlas.
• You want the fastest verified fix before digging through logs.

When not to use this

• Your error text differs from the block above; close cousins often have different causes.
• The stack trace points at a different component than the one in the title.
• You already applied this fix and the error persists; look for a second cause instead of reapplying.

Compatibility

• Not pinned to a specific version; follows current Atlas behavior.

Also seen as

• `in the index? Undeclared filters either error or get ignored.

Confirm

Run the same query with`

Vector search index options (numDimensions, similarity, filter fields) and query options (numCandidates, limit) per the MongoDB Atlas Search docs at https://www.mongodb.com/docs/search/indexes/manage-indexes/, verified 2026-09-26.

Not specified by the original author.

TL;DR: Symptom: querying by vector id returns odd results or fails. Query-by-id compares one stored vector against the index; it is not a fetch, and it has its own limits. Confirm it 1. Run fetch on the id: if fetch works, the record exists and the issue is query semantics.

The error

querying by vector id returns odd results or fails. Query-by-id compares one stored vector against the index; it is not a fetch, and it has its own limits. Confirm it 1. Run fetch on the id: if fetch works, the record exists and the issue is query semantics.

The fix

1. Run fetch on the id: if fetch works, the record exists and the issue is query semantics.
2. Check the namespace on the query call; query-by-id still scopes to a namespace.
3. Check whether a filter is excluding the expected neighbors.
4. For point lookups use fetch, not query(id=...).
5. For find-similar-to-this-record, keep query(id=...) but set namespace, top_k, and filter deliberately.
6. If you need the vector values back too, pass include_values=True; matches do not include values by default.
7. This covers exactly what the title says: Pinecone diagnosis.
8. You are setting this up for the first time, or auditing an existing setup.
9. You want the key gotchas in one place before you start.
10. You are doing a different workflow with Pinecone; these steps are specific to the title above.
11. You need the full reference docs; this is the short path, not the manual.
12. Not pinned to a specific version; follows current Pinecone behavior.

When to use this

• You are seeing this exact error message; match the block above, not just part of it.
• The failing call matches the scenario in the title: Pinecone diagnosis.
• You want the fastest verified fix before digging through logs.

When not to use this

• Your error text differs from the block above; close cousins often have different causes.
• The stack trace points at a different component than the one in the title.
• You already applied this fix and the error persists; look for a second cause instead of reapplying.

Pinecone's querying-by-id limitations troubleshooting page and the query guide define the find-similar semantics and its limits versus fetch. Verified against live docs 2026-09-26.

Not specified by the original author.

SingleStore vector search needs the query vector as a runtime constant

TL;DR

If your SingleStore vector search errors or plans badly when the query vector comes from application code, make it a runtime constant first. Do SET @queryvec = JSONARRAYPACK('[0.44, 0.554, 0.34, 0.62]'); then SELECT EUCLIDEANDISTANCE(vec, @queryvec) AS score FROM vectorsb ORDER BY score ASC LIMIT 10. Pack the constant once per session, not per row or per query execution.

Use this when you see this exact error.

When not to use this skill: a different error message. The steps below target this specific failure.

Compatibility

The steps above apply to the commands named in them. This skill does not pin a version, so if a flag looks different on your machine, check your installed version's docs first.

Details

Also set vectortypeproject_format = JSON when you want vector columns to print readably in results instead of binary.

Context: Official docs (SingleStore EUCLIDEANDISTANCE reference): documents a query-shape gotcha that trips agents building parameterized search. The query vector in a vector search (@queryvector) must be a runtime constant. The documented pattern is SET @queryvec = JSONARRAY_PACK('[...]') once, then SELECT ... WHERE ... ORDER BY score LIMIT k against that variable. Inline JSON constants in the SELECT work for one-off queries, but for repeated queries SingleStore recommends casting the JSON string to a VECTOR once and reusing it.

Source: Official docs (SingleStore EUCLIDEANDISTANCE reference): documents a query-shape gotcha that trips agents building parameterized search. The query vector in a vector search (@queryvector) must be a runtime constant. The documented pattern is SET @queryvec = JSONARRAY_PACK('[...]') once, then SELECT ... WHERE ... ORDER BY score LIMIT k against that variable. Inline JSON constants in the SELECT work for one-off queries, but for repeated queries SingleStore recommends casting the JSON string to a VECTOR once and reusing it.

Verified resolution: If your SingleStore vector search errors or plans badly when the query vector comes from application code, make it a runtime constant first. Do SET @queryvec = JSONARRAYPACK('[0.44, 0.554, 0.34, 0.62]'); then SELECT EUCLIDEANDISTANCE(vec, @queryvec) AS score FROM vectorsb ORDER BY score ASC LIMIT 10. Pack the constant once per session, not per row or per query execution. Also set vectortypeproject_format = JSON when you want vector columns to print readably in results instead of binary.

Source URL: https://docs.singlestore.com/cloud/reference/sql-reference/vector-functions/euclidean-distance/