Search

The Search endpoint performs semantic search across your analyzed videos using OpenAI text-embedding-3-small embeddings and Turso’s vector_distance_cos() function.

Results are module-diverse — the ranking algorithm uses round-robin interleaving across modules to prevent any single module from dominating results.

---

POST /search

Search video segments using natural language.

Auth: Required

Body Field	Type	Required	Description
query	string	Yes	Natural language search query
job_id	string	No	Search within a specific video
job_ids	string[]	No	Search within multiple specific videos
modules	string[]	No	Filter by module types (e.g. ["transcription", "visualelements"])
limit	integer	No	Max results to return (default: 20, max: 200)
threshold	float	No	Minimum similarity score 0.0–1.0 (default: 0.4)

If neither job_id nor job_ids is provided, all completed videos for the user are searched.

Response: SearchResult[]

[
  {
    "id": "42",
    "job_id": "a1b2c3d4-...",
    "module": "visualelements",
    "content": "Objects Detected: person, laptop, desk; Colors: blue, white; ...",
    "start_time": 15.0,
    "end_time": 18.0,
    "metadata": null,
    "similarity": 0.8234
  },
  {
    "id": "78",
    "job_id": "a1b2c3d4-...",
    "module": "transcription",
    "content": "Let me show you how the dashboard works on this laptop.",
    "start_time": 16.2,
    "end_time": 19.5,
    "metadata": null,
    "similarity": 0.7891
  }
]

Results are sorted by similarity (descending). A similarity of 1.0 is a perfect match; the default threshold of 0.4 filters out weak matches.

Deduplication: Results are deduplicated per (job_id, module, 5-second window) to avoid near-duplicate segments.

curl -X POST https://api.lynxvizion.com/api/search \
  -H "Authorization: $LVZ_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "person giving a presentation",
    "limit": 10,
    "threshold": 0.5
  }'

Search within a specific video

curl -X POST https://api.lynxvizion.com/api/search \
  -H "Authorization: $LVZ_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "brand logo visible on screen",
    "job_id": "a1b2c3d4-...",
    "modules": ["logos", "visualelements"],
    "limit": 20
  }'