A lightweight proof-of-concept Cloud Function that pulls live legislative data from the OpenStates API v3, feeds the full bill text to Gemini 2.0 on Vertex AI, and returns an HTML page showing:

• a list of recent Virginia bills (or the bill you request by ID)
• the full bill text & token count
• a plain-language summary generated by Gemini

The goal is to demonstrate how you can combine public-sector open data with generative AI to build rapid legislative-tracking tools.

┌────────┐    HTTP request       ┌──────────────────────┐
│Browser │  ───────────────────▶ │Cloud Function        │
└────────┘                       │ 1. Fetch bills from  │
                                 │    OpenStates v3 API │
                                 │ 2. Download latest   │
                                 │    bill version HTML │
                                 │ 3. Count tokens &    │
                                 │    summarise via     │
                                 │    Gemini-2.0 Flash  │
                                 │ 4. Render HTML       │
                                 └──────────────────────┘

• Node.js ≥ 18 (tested with v20)
• A Google Cloud project with Vertex AI and Cloud Functions Gen 2 enabled
• An OpenStates API key – get one at https://open.pluralpolicy.com/

git clone https://github.com/mattcollier/openstates-demo.git
cd openstates-demo
npm install

# set env vars
export OPENSTATES_API_KEY=xxxxxxxxxxxxxxxxxxxx
export GOOGLE_CLOUD_PROJECT=$(gcloud config get-value project)

# run locally
npx @google-cloud/functions-framework --target=myHttpFunction --port=8080
# open http://localhost:8080/           → list of bills
# open http://localhost:8080/<billId>   → summary for a specific bill

gcloud functions deploy myHttpFunction   --gen2   --runtime=nodejs20   --region=us-central1   --entry-point=myHttpFunction   --trigger-http   --allow-unauthenticated   --set-env-vars=OPENSTATES_API_KEY=YOUR_KEY

.
├── index.js         # Cloud Function entry point
├── ai.js            # Gemini helper: summarise & count tokens
├── header.html      # Optional page header (empty by default)
├── footer.html      # Optional page footer (empty by default)
├── package.json
└── .gcloudignore

• Different state – replace the VA_ID value with another jurisdiction ID from the OpenStates docs.
• Styling – add CSS/JS to header.html & footer.html.
• Summary prompt – tweak ai.js → generateContent() for longer/shorter abstracts, JSON output, etc.

• No paging — only the first 20 bills are fetched.
• Bills whose HTML exceeds Gemini’s context window will be truncated.
• Minimal error handling & no caching.
• Consider moving to Cloud Run if you need higher concurrency.

• OpenStates / Plural Policy for open legislative data.
• Google Vertex AI for the Gemini model.
• @digitalbazaar/http-client for fetch wrapper.

Apache-2.0 — see LICENSE for details.