d+
devoteam
Snowflake Partner
Training pack · Devoteam Snowflake Practice · 2026

I want to run the example apps

User Guide · Tasks
← Back to overview

For anyone evaluating or learning the library · 20 min · a Snowflake connection for the CoCo-backed demos

You end up with each of the repo's example apps running locally, and know which one demonstrates which part of the API.

Before you start#

  • streamlit-coco[sdk] installed and check_environment().ready is True

    (see Getting started).

  • A checkout of the repo (the examples run from make targets, not from a

    pip install alone).

  • headless_pipeline.py and the BI → Semantic estate-map/KPI screens run

    without a live Snowflake connection; the rest need one.

Steps#

  1. Run the preferred-pattern chat demo:
    bash
    make chat

    This is examples/chat_app.py — the smallest working app built on check_environment → render_start_gate → get_or_create_session → panel() + chat_input_bar(). Its sidebar includes a test-prompt runner backed by examples/testdata/prompts.json.

  2. Run the legacy all-in-one demo:
    bash
    make approval

    This is examples/approval_gate.py, built on the legacy CCv2 chat() helper instead of panel().

  3. Run the structured-output demo:
    bash
    make structured

    This is examples/structured_output.py — it sets output_schema and routes the agent's JSON into a widget via on_structured_output instead of chat text.

  4. Run the headless demo from a terminal, no browser:
    bash
    make headless

    This is examples/headless_pipeline.py — pure async Python with no Streamlit import, driving query() and CocoSession directly.

  5. Run the file-upload demo:
    bash
    make cwd-upload

    This is examples/cwd_upload_chat.py — chat_input_bar(accept_file=...) and cwd_uploader() wiring browser uploads into the agent's cwd.

  6. Run the two multipage, right-rail demos:
    bash
    make backlog        # Product Backlog Desk — file/JSON-backed, no SQL
make bi-semantic     # BI -> Semantic (alias: make tableau-semantic)
  7. Run the guided app-generation demo:
    bash
    make app-builder

Each make target and its variables are listed in Command reference; what each example demonstrates and what it needs is in Example apps reference.

Result#

Each make target opens (or prints, for make headless) a running app on its own port. Nothing outside examples/ is modified by running a demo; generated artifacts land under each example's own out/ or data/ directory — see Outputs.

Variations#

  • make run runs examples/chat_app.py by default; set EXAMPLE= to point

    it at a different example file.

  • make tableau-semantic is a deprecated alias for make bi-semantic — use

    the latter.

If it fails#