multimodal_fin.runners package

Submodules

multimodal_fin.runners.base module

Base interface for runner classes in the multimodal_fin CLI.

All runners must inherit from this class and implement the run() method.

class multimodal_fin.runners.base.Runner[source]

Bases: ABC

Abstract base class for all runners used in the CLI.

abstractmethod run(**kwargs)[source]

Execute the runner’s logic.

Raises:

NotImplementedError – Must be implemented by subclasses.

Return type:

None

multimodal_fin.runners.downloads_runner module

Runner for downloading conference data (transcripts and audio) from earningscall.biz.

It uses a subset of S&P500 companies grouped by sector.

class multimodal_fin.runners.downloads_runner.DataAdquisitionRunner(settings)[source]

Bases: Runner

Runner responsible for fetching earnings call transcripts and audio.

run(**kwargs)[source]

Download data for S&P500 companies from earningscall.biz.

This includes scraping the main page, parsing the company table, and triggering download for transcripts and audio files.

Return type:

None

multimodal_fin.runners.embeds_runner module

Runner for generating multimodal hierarchical embeddings from enriched JSON files.

Embeddings are generated using both node-level and conference-level encoders.

class multimodal_fin.runners.embeds_runner.EmbedRunner(settings, emb_cfg)[source]

Bases: Runner

Runner for generating embeddings from enriched transcript JSONs.

run(paths, **kwargs)[source]

Generate and log embeddings for each input JSON file.

Args: paths (List[str]): List of file paths to enriched transcript JSONs.

Return type:

None

multimodal_fin.runners.process_runner module

Runner for executing the conference processing pipeline.

This runner handles classification, enrichment, and other processing steps based on a provided configuration.

class multimodal_fin.runners.process_runner.ProcessRunner(settings)[source]

Bases: Runner

Runner responsible for executing the main conference processing pipeline.

run(**kwargs)[source]

Run the full processing pipeline: classification and metadata enrichment.

Return type:

None

Module contents

Factory method for obtaining CLI runners based on the selected mode.

This module exposes a single function get_runner that returns an instance of the appropriate runner class based on the CLI command invoked.

multimodal_fin.runners.get_runner(mode, config)[source]

Return the appropriate runner instance based on the selected mode.

Parameters:
  • mode (str) – One of ‘process’, ‘embed’, or ‘download’.

  • config (FullConfig) – Aggregated pipeline configuration.

Returns:

Instantiated runner object.

Return type:

Runner

Raises:

ValueError – If the configuration is missing or the mode is unknown.