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

I want to build a minimal CoCo app

User Guide · Tasks
← Back to overview

For a Python developer building a Streamlit app · 10 min · a working Python environment

You end up with a single-page Streamlit app that embeds a CoCo panel and input bar, using the preferred integration pattern.

Before you start#

  • streamlit-coco[sdk] installed (see Enablement

    setup.html).

  • A Snowflake connection configured and reachable —

    check_environment().ready should already be True; see Getting started.

  • No prior CoCo app is required; this is the starting point for every other

    task page in this guide.

Steps#

  1. Create streamlit_app.py and construct a CocoOptions with at least a

    connection name:

    python
    import streamlit as st
import streamlit_coco as st_coco

st.set_page_config(page_title="Minimal CoCo app")

options = st_coco.CocoOptions(
    connection="my_connection",
    cwd=".",
)
  2. Call check_environment() through the landing gate, and stop the script

    if it is not ready:

    python
    if not st_coco.render_start_gate(options):
    st.stop()

    render_start_gate() renders the landing screen itself (SDK/CLI/Snowflake status, a Start action) and returns False while the environment or the user has not confirmed — your app must call st.stop() in that case, render_start_gate() does not stop the script for you.

  3. Get or create the session for this page, and mount the panel and its

    input bar:

    python
    session = st_coco.get_or_create_session(options)
st_coco.panel(session)
st_coco.chat_input_bar(session)

    get_or_create_session() stores the session in st.session_state["coco"] by default (override with key=) and reuses it across reruns.

  4. Run the app:
    bash
    streamlit run streamlit_app.py
  5. Type a prompt in the input bar at the bottom of the page and confirm the

    transcript streams a response above it.

Result#

A browser tab showing the CoCo panel: a status header, an empty transcript, and the chat input bar. After a prompt, the transcript fills with the agent's streamed response and any tool cards it used. No file is written by this minimal app on its own — file output starts once you add uploads (I want to add file uploads) or the agent writes to cwd.

Variations#

  • Use copilot_rail() instead of panel() + chat_input_bar() to embed CoCo

    as a right-hand rail inside a larger, multipage app rather than a single-page chat — see the Backlog Desk and BI → Semantic examples in Example apps reference.

  • Set panel(session, output_mode="field") to render only the latest

    assistant output instead of the full transcript.

  • Set warm_up=True on render_start_gate() (the default) to start the

    session as soon as the gate passes, instead of waiting for the first prompt.

If it fails#