wavecap-search

Search WaveCap transcriptions with text queries and filters. Use when the user wants to find specific transcriptions, search for keywords, or query transcription history.

⭐5stars🔀1forks📁View on GitHub🕐Updated Jan 11, 2026⬇️2downloads

When & Why to Use This Skill

The WaveCap Search skill enables Claude to perform advanced, filtered queries across transcription histories. It allows users to locate specific keywords, analyze audio stream content over precise time intervals, and manage large-scale transcription data through pagination and confidence-score filtering, effectively turning recorded audio into a searchable knowledge repository.

Use Cases

  • 1. Incident Investigation: Rapidly identify critical events by searching for keywords like 'emergency', 'fire', or 'incident' across multiple live or archived streams.
  • 2. Compliance and Auditing: Retrieve specific transcription records within a defined ISO8601 date range to verify protocol adherence or historical accuracy.
  • 3. Transcription Quality Assurance: Filter results by confidence thresholds to isolate low-confidence segments for manual review and correction.
  • 4. Timeline Analysis: Use chronological sorting (asc/desc) and time-based filters to reconstruct the sequence of events from recorded audio data.
namewavecap-search
descriptionSearch WaveCap transcriptions with text queries and filters. Use when the user wants to find specific transcriptions, search for keywords, or query transcription history.

WaveCap Search Skill

Use this skill to search through transcription history in WaveCap.

Search Transcriptions for a Stream

curl -s "http://localhost:8000/api/streams/{STREAM_ID}/transcriptions?search=fire&limit=20" | jq

Search Parameters

Parameter Type Default Description
search string - Text to search for in transcriptions
limit int 100 Maximum results to return
before ISO8601 - Get transcriptions before this timestamp
after ISO8601 - Get transcriptions after this timestamp
order string desc Sort order: asc or desc

Search with Time Range

# Search in the last hour
AFTER=$(date -u -v-1H +"%Y-%m-%dT%H:%M:%SZ")
curl -s "http://localhost:8000/api/streams/{STREAM_ID}/transcriptions?search=emergency&after=$AFTER" | jq
# Search within a specific date range
curl -s "http://localhost:8000/api/streams/{STREAM_ID}/transcriptions?search=incident&after=2025-01-15T00:00:00Z&before=2025-01-16T00:00:00Z" | jq

Pagination

The response includes pagination hints:

{
  "transcriptions": [...],
  "hasMoreBefore": true,
  "hasMoreAfter": false
}

To get the next page:

# Get the timestamp of the last result, then use it as 'before'
LAST_TIMESTAMP="2025-01-15T12:30:00Z"
curl -s "http://localhost:8000/api/streams/{STREAM_ID}/transcriptions?before=$LAST_TIMESTAMP&limit=20" | jq

Search Across All Streams

To search across all streams, first get the stream list, then query each:

# Get all stream IDs
STREAMS=$(curl -s "http://localhost:8000/api/streams?includeTranscriptions=false" | jq -r '.[].id')

# Search each stream
for STREAM_ID in $STREAMS; do
  echo "=== Stream: $STREAM_ID ==="
  curl -s "http://localhost:8000/api/streams/$STREAM_ID/transcriptions?search=YOUR_QUERY&limit=5" | jq '.transcriptions[] | {timestamp, text}'
done

Format Search Results

Show just text and timestamps

curl -s "http://localhost:8000/api/streams/{STREAM_ID}/transcriptions?search=fire" | \
  jq '.transcriptions[] | "\(.timestamp): \(.text)"' -r

Show with confidence scores

curl -s "http://localhost:8000/api/streams/{STREAM_ID}/transcriptions?search=ambulance" | \
  jq '.transcriptions[] | {time: .timestamp, text, confidence: (.confidence * 100 | floor | tostring + "%")}'

Filter by confidence threshold

curl -s "http://localhost:8000/api/streams/{STREAM_ID}/transcriptions?limit=50" | \
  jq '[.transcriptions[] | select(.confidence > 0.8)] | length'

Transcription Fields

Each transcription contains:

{
  "id": "uuid",
  "streamId": "stream-id",
  "text": "transcribed text",
  "timestamp": "2025-01-15T12:30:00Z",
  "confidence": 0.95,
  "duration": 30.5,
  "segments": [...],
  "recordingUrl": "/recordings/file.wav",
  "reviewStatus": "pending",
  "correctedText": null,
  "alerts": [...]
}

Tips

  • The search is case-insensitive
  • Use order=asc to get oldest results first (useful for timeline analysis)
  • Combine after and before for precise time windows
  • Large result sets should use pagination to avoid timeouts