Testimony summarization -- Just look at hearing_overview_pipeline.py

[danielblauvelt]

Summary

Add a short summary of the changes, and a reference to the original issue using # and the issue number, like #1
Created a pipeline that takes in a testimony hearing and provides an AI overview of what was discussed during the hearing.

Checklist

• On the frontend, I've made my strings translate-able.
• If I've added shared components, I've added a storybook story.
• I've made pages responsive and look good on mobile.

Screenshots

Add some screenshots highlighting your changes.

Known issues

If you've run against limitations or caveats, include them here. Include follow-up issues as well.

Steps to test/reproduce

For each feature or bug fix, create a step by step list for how a reviewer can test it out. E.g.:

1. Go to the home page
2. Click on a testimony
3. See that it's loaded with a loading spinner

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
maple-dev	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	Jul 2, 2025 0:14am
maple-prod	🔄 Building (Inspect)	Visit Preview	💬 Add feedback	Jul 2, 2025 0:14am

[Mephistic]

For reference, here's the the issue for actually implementing the pipeline to generate these summaries on an ongoing basis: #1840

Feel free to use this PR for just the initial work to support a function that can take a Hearing + Transcription Text and generate a summary - we can then wrap it and backfill old transcriptions once the blockers listed in that issue are removed.

[nesanders]

Note: current state of these scripts is a local exploration, not yet written to be deployed as a firebase function.

[nesanders]

@Aaquila thank you for your willingness to work on this! Here are some suggestions for doing a refactor of Barry's code.

1. Focus on his script llm/hearing_overview_pipeline.py
2. Add a toplevel comment on the script and function docstrings
3. Surface PROMPT_BILL_SUMMARIES as a global variable. I think you can just integrate this with PROMPT_INSTRUCTIONS instead of having separate vars.
4. Make receive_hearing more modular so it calls separate functions to 1) retrieve bill summaries and 2) retrieve hearing metadata (not all in one function)

Nice to have features,

1. Repoint receive_hearing internals to pull hearing metadata from our firebase internally rather than the legislature API.
2. Add error handling to receive_hearing, e.g. to handle the case where the bill summary is missing.
3. Instead of directly using the openai package, query via LangChain using the ChatOpenAI class as we do in maple/llm/llm_functions.py