OpenBook Next

Inspiration

We have city's treasury data publicly available, but it is still difficult to access / understand by an average citizen. OpenBook was an effort to bridge this gap, but it is not usable enough even for the technically-savvy users. There is a need to build a better alternative.

What it does

The goal is to SF's financial information accessible for an average citizen! You can ask your questions about the city's budget and spending using natural language. Then, our app will intelligently generate a SQL query against the source of truth dataset hosted on DataSF.

We show the generated SQL query for transparency, the result of the query in a table, and summarize the result of the SQL in a human-readable format.

Next steps

Although we have open datasets that we can use GPT to query, these data can be fragmented, sometimes corrupted, or inconsistent — for example, the Budget dataset only contains data up to 2018, and some contract data does not contain a contract title or compartment. It is also difficult to link all these data together — it is hard to join all the data together to answer questions like "what percentage of the 5MM we allocated for Homelessness service have been spent", because the code to identify a particular issue can be inconsistent across the "budget", "spending", "voucher" datasets.

We believe that to obtain full transparency, and accuracy, our city needs a fully integrated "Ledger" that tracks all its funds flow from end to end — from revenue intake, budget allocation, grants to departments , contract awarding, to vendor payments.

Built With

• fastapi
• google-bigquery
• gpt4
• llm
• nextjs
• python
• typescript