❄️ Snow Instructor ❄️

Snow instructor teaches you about Snowflake's capabilities by playfully answering quiz questions. Check out this video!

💫 Features

With this full-fledged AI-based Streamlit Application on Snowflake, you'll learn about:

• Snowpark's Python API
• Snowflake's Cortex AI capabilities and especially the Snowflake Arctic LLM
• Multi-page Streamlit application
• Streamlit Deployment on Snowflake
• Snowflake CLI
• Threading in conjunction with Streamlit for long-running background tasks

⛷️ Getting Started

1. If you are a new Snowflake user, register a 30-day free trial Snowflake account. Choose the enterprise edition, AWS as cloud provider and region AWS US West 2 (Oregon), as Snowflake Arctic is currently only available in this region.

2. Create a Snowflake configuration file under ~/.snowflake/connections.toml with:

   [default]
   account = "YOUR_ACCOUNT"
   user = "YOUR_USER_NAME"
   password = "YOUR_PASSWORD"
   role = "accountadmin"
   warehouse = "COMPUTE_WH"
   database = "SNOWINSTRUCTOR"
   schema = "public"

3. Clone this repository into a directory snowflake-instuctor.

4. Install hatch globally, e.g. with pipx, i.e. pipx install hatch.

5. Let our spider crawl the Snowflake documentation and upload it into a Snowflake table with:

   hatch run crawl-snow-docs

   This needs to be done exactly once and will also make sure that your connection works.

6. To check if our Snowflake Instructor works on the command-line, try:

   hatch run cli-quiz

7. To start the Streamlit client locally on your machine, run:

   hatch run snow-instructor

8. To deploy everything on Snowflake, run within the root of this repo:

   hatch run prep-deployment  # to setup the warehouse, etc.
   hatch run deploy-streamlit

   And open up the URL shown in the output of the last command.

🛠️ Development

To set up hatch and pre-commit for the first time:

1. install hatch globally, e.g. with pipx, i.e. pipx install hatch,
2. make sure pre-commit is installed globally, e.g. with pipx install pre-commit.

A special feature that makes hatch very different from other familiar tools is that you almost never activate, or enter, an environment. Instead, you use hatch run env_name:command and the default environment is assumed for a command if there is no colon found. Thus you must always define your environment in a declarative way and hatch makes sure that the environment reflects your declaration by updating it whenever you issue a hatch run .... This helps with reproducability and avoids forgetting to specify dependencies since the hatch workflow is to specify everything directly in pyproject.toml. Only in rare cases, you will use hatch shell to enter the default environment, which is similar to what you may know from other tools.

To get you started, use hatch run test:cov or hatch run test:no-cov to run the unitest with or without coverage reports, respectively. Use hatch run lint:all to run all kinds of typing and linting checks. Try to automatically fix linting problems with hatch run lint:fix and use hatch run docs:serve to build and serve your documentation. You can also easily define your own environments and commands. Check out the environment setup of hatch in pyproject.toml for more commands as well as the package, build and tool configuration.

The environments defined by hatch are configured to generate lock files using hatch-pip-compile under locks. To upgrade all packages in an environment like test, just run hatch run test:upgrade-all. To upgrade specific packages, type hatch run test:upgrade-pkg pkg1,pkg2.

🙏 Credits

This package was created with The Hatchlor project template and initiated by the Arctic Streamlit Hackathon.