Skip to content

Latest commit

 

History

History
49 lines (31 loc) · 1.67 KB

README.md

File metadata and controls

49 lines (31 loc) · 1.67 KB
Raw

Ask data

Ask data is a project for asking structured data in a natural language.

Quickstart

The easiest way to give it a try is to open in GitHub Codespaces

There is a free quota, open the codespace, once loaded do this in the terminal:

cd ask-data-pf-dag
pip install -r requirements.txt
pf connection create -f openai.yaml --set api_key=YOUR_OPENAI_KEY
pf flow serve --source . --port 8080 --host localhost

This will start the simple chat where you could ask the data in a natural language. Please remember that you pay for the LLM API.

Data source

The project uses ClickHouse playground as a data source, because it's free, fast and easy to use, plus it has variety of datasets from different areas.

Here are some examples of the queries you may ask:

  • What's the top five most expensive apartments in the UK?
  • What are the months when COVID is most dangerous?
  • What is the average length of the taxi trip?
  • Who made the most of the contributions to ClickHouse/ClickHouse in the summer 2021
  • ...

Ask data with promptflow

DAG

DAG (Directed Acyclic Graph) is a simple and natural way to describe a flow for solving a problem. Promptflow gives us an easy way to build/debug/run/evaluate the flow.

Follow the installation instructions.

The flow definition looks like this. dag.png

Ask questions in the chat. chat-example.png