Training pack · Devoteam Snowflake Practice · 2026

Setup guide

Get the development repository running locally and verify you are ready for the guided tour and the hands-on exercises. Full reference: doc/deployment/local.md — this page is the condensed, training-specific path through it.

Required accounts and permissions#

A Snowflake account with Cortex Code access. Ask your asset owner if
you do not have one for training — a shared/sandbox connection may be available.
Git access to DevoteamSP/streamlit-coco-dev (the development repository).
If you only need the published package, pip install "streamlit-coco[sdk]" from PyPI is enough and you can skip step 1 below.

1. Clone and install#

bash

git clone https://github.com/DevoteamSP/streamlit-coco-dev.git
cd streamlit-coco-dev
make install     # uv sync --extra dev

Verify the package imports and check the version:

bash

uv run python -c "import streamlit_coco as c; print(c.__version__)"

2. Install and verify the CoCo CLI#

bash

cortex --version

If it is not on PATH, either install it or point at the binary:

bash

export CORTEX_CODE_CLI_PATH=/path/to/cortex

CocoOptions(cli_path=...) also works, in code, as an alternative to the environment variable.

3. Set up a Snowflake connection#

Create or confirm ~/.snowflake/connections.toml:

toml

[connections.analytics]
account = "xy12345"
user = "you@company.com"
authenticator = "externalbrowser"

Legacy ~/.snowflake/config.toml is also detected. When connection is omitted in CocoOptions, the SDK uses the Snowflake CLI's default connection.

4. Environment check (no agent started yet)#

bash

uv run python -c "
import streamlit_coco as st_coco
env = st_coco.check_environment(connection='analytics')
print(env.ready, env.sdk_version, env.cli_version, env.snowflake_config_display)
"

env.ready should print True. If not, see the troubleshooting table in doc/deployment/local.md §7 before continuing.

5. Run the test suite#

bash

make check    # ruff + unit/smoke pytest (excludes tests/e2e)

All tests should pass. This confirms your Python environment matches CI.

6. Run your first demo#

bash

make chat

A browser tab opens with the panel() + chat_input_bar() demo (examples/chat_app.py). Type a prompt — for example *"What tables are available in SNOWFLAKE_SAMPLE_DATA.TPCH_SF1?"* — and confirm you see a streaming response with at least one tool card.

Verification checklist — you're ready when...#

uv run python -c "import streamlit_coco as c; print(c.__version__)" prints 0.1.9
cortex --version succeeds
check_environment(...).ready is True
make check passes with no failures
make chat opens a working app and you get a streamed response with at least one tool card

If every box is checked, move on to guided-tour.md.

Troubleshooting quick reference#

Symptom	Fix
`SDKNotInstalledError`	`make install` again, or `pip install streamlit-coco[sdk]`
`CLINotFoundError`	Install the CLI; set `CORTEX_CODE_CLI_PATH`
`CocoConnectionError`	Fix `connections.toml`; test with `cortex --version` and the Snowflake CLI directly
Session stuck on "Starting CoCo…"	Check terminal logs; set `STREAMLIT_COCO_DEBUG=1`

Full table: doc/deployment/local.md §7. The Technical Guide has the complete reference: help.html.