Processes the input from one of the following: - file: an uploaded file (image, audio, PDF or text) - content: a text string - url: a

Render the custom testing interface.

Run all available tests.

Run the full agent loop until completion

Run the narrowing accuracy integration test.

Run the complete pipeline accuracy integration test.

Run the selection accuracy integration test.

Run the TUI application

Run unit tests using pytest.

Save the dataset to local files in the specified format. Note: Images cannot be saved to JSON/CSV formats, only parquet preserves the

Save evaluation results to disk.

Generate video summary from transcript using BAML Returns: VideoSummary with bullet points, topics, and takeaways

Test YouTube API connection by fetching channel info

Test YouTube API connection by fetching channel info

Test Zoom API connection by fetching user info

Test Zoom API connection by fetching user info

Initialize the hybrid narrowing strategy. Args: embedding_service: The module's embedding service. use_vector_store:

Add guests to the session via email.

Build cache of existing categories for fast lookup.

Build BAML TypeBuilder for dynamic categories. Args: categories: The categories to build the TypeBuilder for. Returns:

Build lookup maps for efficient hierarchy navigation.

Check if a slug is available for use. Args: slug: The slug to check Returns: True if the slug is av

Click the 'Session' option in the new session dropdown.

Collect all categories involved in the pipeline analysis.

Return start/end time dicts for a clip, or null values if not computable.

Create a new collection with current settings.

Create start and end times for an event on the given date. Events are always 10:15-11:15 AM PST. Args: event_da

Create the data for the pipeline analysis table.

Run text through deslop CLI via uvx. Falls back to original on failure.

Run text through the deslop CLI via uvx. Falls back to original text on failure.

Download Zoom recording with caching

Download Zoom recording with caching

Evaluate correctness using flexible definition and return results.

Fill in the session description field.

Fill in the session name field.

Return the character offset of *anchor* in *transcript*. Tries exact match first, then case-insensitive, then a trimmed first-15-word

Format a URL suffix into a valid slug. Args: luma_url_suffix: The URL suffix to format Returns: For

Call BAML for a batch of talks; return {talk_number: description}.

Generate embeddings for a batch of texts.

Generate the episode README using BAML and the example template

Generate the updated root README

Get Zoom access token from stored credentials

Get Zoom access token from stored credentials

Generate cache filename for a recording

Generate cache filename for a recording

Return duration in seconds using ffprobe.

Find a Luma event that matches both the Zoom recording date AND contains the Zoom meeting ID in its URL/description. Returns the matc

Calculate the date to look back from today. Args: months: Number of months to look back (default: LOOKBACK_MONTHS)

Get transcript from Zoom recording

Get transcript from Zoom recording

Load the template from file (cached after first load). Returns: The template string Raises:

Load the ChromaDB vector store. Args: auto_create: Whether to create the vector store if it doesn't exist.

Load YouTube API credentials from the existing OAuth setup

Load YouTube API credentials from the existing OAuth setup

In-memory similarity search. Args: text: The text to narrow categories based on. categories: The categories to narrow

Use vector store for fast similarity search. Args: text: The text to narrow categories based on. max_results: Maximum

Navigate the calendar to the target month.

Navigate to Planner and open the new session form.

Parse category.txt into Category objects. Args: file_path: The path to the category file. Returns: The categ

Parse a single category line. Args: line: The line to parse. Returns: The category.