Semantic Search: Query Your Transcripts by Meaning

Search across your entire transcript library by meaning, not just exact keywords. Our search index understands context — find transcripts by topic, speaker, or concept even when the exact words differ.

Endpoint

bash

GET /api/v1/library/search?q={query}&limit={n}&minScore={threshold}
Scope: library:search

Library search with a typed query — Library → search by meaning across all your transcripts

Example

const results = await client.library.search('product launch metrics', {
  limit: 5,
  minScore: 0.8,
});

for (const r of results) {
  console.log(r.transcriptionId, r.score.toFixed(2), r.snippet);
}

json

{
  "success": true,
  "data": {
    "query": "product launch metrics",
    "results": [
      {
        "transcriptionId": 42,
        "score": 0.91,
        "snippet": "...our MRR hit $12k on launch day, with 340 signups in the first 24 hours...",
        "sourceUrl": "https://youtube.com/watch?v=...",
        "title": "Episode 17: Launch Debrief",
        "timestampSeconds": 892
      },
      {
        "transcriptionId": 17,
        "score": 0.87,
        "snippet": "...the activation rate was 62%, which beat our benchmark by 14 points...",
        "sourceUrl": "https://youtube.com/watch?v=...",
        "title": "Founders Roundtable S2",
        "timestampSeconds": 2341
      }
    ],
    "total": 2
  }
}

Parameters

q — natural language query (required). No boolean syntax needed.
limit — max results (default 10, max 50).
minScore — relevance threshold 0–1 (default 0.7). Raise to 0.85+ for tighter, more precise matches.

How it works

When a transcription completes, it is automatically indexed for semantic search — no extra step needed.
When you search, your query is matched against all indexed passages owned by your API key by meaning, not by keyword.
Results are ranked by relevance score and de-duplicated to one snippet per transcription.
New transcriptions become searchable moments after processing completes.

FAQ

How is semantic search different from keyword search?

Keyword search finds exact word matches. Semantic search understands intent — searching for “product revenue” will also surface passages that mention “MRR”, “sales figures”, or “quarterly earnings” without those words appearing in your query.

How quickly are transcripts indexed after processing?

Indexing runs as part of the transcription pipeline. By the time a job reaches completed status, the transcript is already searchable.

What languages does search support?

Semantic search works across all 100+ languages supported by the transcription API. Query in any language — results will be scored by meaning regardless of the language of the underlying transcript.