Cortex Code Quiz App - Snowsight edition

A browser-only asset for building a Streamlit-in-Snowflake certification quiz app with Cortex Code in Snowsight (GA since 2026-03-09). Load a study guide PDF, and the agent creates the schema, extracts exam domains, generates questions, builds quiz.py, and deploys. All from a single chat session, zero local tooling.

Baseline exam: SnowPro Core COF-C03. Swappable to any Snowflake certification by re-running /setup-exam with a different study guide PDF.

---

what you get

• A two-page Streamlit app:
  • Quiz - configure round, answer single-/multi-select questions, see score + pass/fail (75% threshold) with wrong-answer cards.
  • Review - filterable history of every mistake + learning dashboard (score trend, per-domain error distribution, session history).
• AI-grounded content: domains and key facts come from the official study guide PDF (AI_PARSE_DOCUMENT); questions are either AI-generated at runtime (AI_COMPLETE), pre-generated into a question bank, or imported from CSV/json file.
• Persistent progress across sessions via QUIZ_SESSION_LOG + QUIZ_REVIEW_LOG - the app remembers your weak domains between logins.
• Schema-per-exam isolation (QUIZ_<EXAM_CODE>) so multiple certifications coexist in one database.
• custom Cortex Code skills* in .snowflake/cortex/skills/ drive the whole pipeline end to end.
• Optional add-ons: spaced repetition, flashcards, exam-simulation mode, achievement badges, AI study recommendations ($quiz/features).

what is different from the CLI version

	CLI variant	Snowsight variant (this branch)
environment	terminal + snow CLI + bash	browser workspace, zero local tooling
file uploads	snow stage copy	manual upload via Snowsight UI
isolation	git branch per exam + schema (agent-automated)	schema per exam always; optional branch per exam if workspace is Git-backed (user creates the branch manually)
custom skills path	.cortex/skills/	.snowflake/cortex/skills/
global skills	~/.snowflake/cortex/skills/	built into Cortex Code
quiz.py, environment.yml	tracked in repo	generated by agent, not committed
input data (data/)	tracked in repo	user uploads PDF/CSV directly to stage

---

prerequisites

• Snowflake account with a role that has USAGE+CREATE SCHEMA on a database, USAGE on a warehouse, and access to Cortex AI functions.
• one-time as ACCOUNTADMIN (EU accounts only): ALTER ACCOUNT SET CORTEX_ENABLED_CROSS_REGION = 'AWS_US';
• Snowsight > AI & ML > Agents > Settings > Tools and connectors > Web search > enable.
• a study guide PDF for your target exam - Snowflake certifications catalog.

---

disclaimer

AI-generated questions and explanations may be inaccurate - LLMs hallucinate, especially on niche Snowflake features. Verify against official Snowflake docs when in doubt.

When AI question generation or AI explanations are enabled, the app runs slower a little bit: each question is a live AI_COMPLETE round-trip. You can switch to a faster model (see docs/customization.md if present, or edit cortex llm in AGENTS.md) at the cost of quality and more hallucinations.

---

getting started

first time in Snowsight or with Cortex Code? → docs/instructions-detailed.md

already comfortable? → docs/instructions.md

then paste the prompt from docs/prompts.md into the Cortex Code chat.

two ways to load this asset into a workspace

• via Git integration (recommended): fork this repo on GitHub, register it in Snowflake (CREATE API INTEGRATION + CREATE GIT REPOSITORY), then Projects > Workspaces > + Workspace > From Git repository. Workspace opens with the full tree in place - skills, AGENTS.md, and docs/ all accessible from the Snowsight file browser. You also get version control: edits to skills / AGENTS.md / generated quiz.py can be committed back to your fork from inside Snowsight.
• via manual upload (fallback, fully offline-capable): clone the repo locally, then in Snowsight create an empty workspace and drag-drop .snowflake/cortex/skills/ and AGENTS.md into it via the Cortex Code UI. docs/ stays on your local clone - no in-browser access.

See docs/instructions-detailed.md step 2 for the DDL snippets and walk-through. Canonical Snowflake docs: Integrate workspaces with a Git repository and Using a Git repository in Snowflake.

---

repo structure

coco-quiz-app
├── README.md                         
├── AGENTS.md                         -- Cortex Code context (read, but do not restructure)
├── .snowflake/cortex/skills/         -- custom project skills (upload to workspace)
│   ├── setup-exam                    -- 10-step end-to-end pipeline
│   ├── adapt-questions               -- optional CSV/JSON question-bank import
│   ├── cortex/                       -- Cortex AI patterns + prompt audit
│   ├── sis/                          -- Streamlit-in-Snowflake patterns + 22-item pre-deploy scan
│   └── quiz/                         -- screens, questions, style, optional features
└── docs/
    ├── instructions.md               -- concise step-by-step
    ├── instructions-detailed.md      -- detailed walk-through with Snowsight UI paths
    ├── prompts.md                    -- the prompt to paste into Cortex Code chat
    ├── skills.md                     -- one-paragraph description per skill
    ├── customization.md              -- model, UI, features, cross-provider exam swap
    ├── architecture.md               -- data flow, table relationships, skill graph
    └── troubleshooting.md            -- common issues and fixes

Outputs (quiz.py, environment.yml) live in your workspace once the agent generates them. They are intentionally not tracked in this repo.