Fathom Quote Extractor

An Apify Actor that extracts video chunk download URLs for specific quotes from Fathom meeting recordings.

Overview

This actor automates the extraction of direct video download URLs for precise moments in a Fathom.video recording. You provide a share URL and a list of quotes; it finds those quotes in the transcript and returns the corresponding video chunk URLs. It's built for developers needing to programmatically access specific clips from Fathom recordings.

Key Features

• Quote Extraction: Finds exact quotes within Fathom video transcripts.
• Batch Processing: Handles multiple quotes in a single execution.
• Fuzzy Matching: Intelligently handles minor transcription variations, filler words ("uh", "um"), and punctuation differences.
• Structured JSON Output: Provides clear results with metadata, including download URLs and chunk IDs.

How to Use

Input

Configure the actor run with a JSON input object. Only fathomShareUrl and quotes are required.

{
  "fathomShareUrl": "https://fathom.video/share/ABC123",
  "quotes": [
    "This is the first quote to find",
    "This is another quote to extract"
  ],
  "delayBetweenQuotes": 1.0,
  "pageLoadTimeout": 30
}

• fathomShareUrl (string, required): The Fathom share link. Must start with https://fathom.video/share/.
• quotes (array, required): An array of quote strings to locate in the transcript.
• delayBetweenQuotes (number, optional): Delay in seconds between processing each quote. Default is 1.0.
• pageLoadTimeout (number, optional): Page load timeout in seconds. Default is 30.

Output

The actor outputs a dataset with a single item containing all results.

{
  "fathomShareUrl": "https://fathom.video/share/ABC123",
  "results": [
    {
      "quote": "This is the first quote to find",
      "downloadURL": "https://fathom-recording-data.storage.googleapis.com/chunk/...",
      "found": true,
      "chunkId": "12345678_1_00042"
    },
    {
      "quote": "This is another quote to extract",
      "downloadURL": "",
      "found": false,
      "chunkId": null
    }
  ],
  "processedAt": "2024-01-15T10:30:00.000Z",
  "totalQuotes": 2,
  "quotesFound": 1,
  "quotesNotFound": 1
}

Each item in results indicates whether the quote was found and provides the direct download URL and chunk ID if successful.

How It Works

1. Navigate: Launches a headless browser and loads the provided Fathom share URL.
2. Load Transcript: Automatically clicks the TRANSCRIPT tab and waits for the content to load.
3. Find Quote: Searches through transcript segments using a combination of exact text matching and fuzzy logic (including a 70% word overlap threshold) to accommodate minor variations.
4. Capture URL: Clicks the matching transcript segment and intercepts the network request to capture the video chunk's direct download URL.
5. Output: Compiles and returns all results in the structured JSON format.

Local Development

Prerequisites: Python 3.10+, Chrome browser, Apify CLI.

1. Clone the repo and set up a virtual environment:
   bash python -m venv .venv source .venv/bin/activate # On Windows: .venv\Scripts\activate
2. Install dependencies:
   bash pip install -r requirements.txt
3. Place your input JSON in storage/key_value_stores/default/INPUT.json.
4. Run the actor: apify run
5. Check the output in storage/datasets/default/.

Deployment

Deploy to the Apify platform using the CLI:

apify login
apify push

The repository includes GitHub Actions workflows for linting/testing and for automatic deployment to Apify on pushes to the main branch. The APIFY_TOKEN secret is required for deployment.

Limitations

• Requires a valid, non-expired Fathom share URL.
• The quote must exist in the transcript, either verbatim or with very minor variations.
• Captured download URLs are time-limited (typically valid for about 6 hours).

License

MIT