-
Locate the sf-hcls-solutions repository:
- Check
~/project/sf-hcls-solutions/
- Check current working directory
- If not found:
git clone https://github.com/Snowflake-Labs/sf-hcls-solutions.git /tmp/sf-hcls-solutions
-
Read solutions/clinical-quality-agent/manifest.json.
-
Present the installation plan:
Solution: Clinical Quality and Patient Safety Agent v1.0.0
Industry: Healthcare & Life Sciences
Database: SF_SOLUTIONS
Schema: CLINICAL_QUALITY_SAFETY
Role: ACCOUNTADMIN
What will be created:
- 8 clinical quality tables (75K patients with realistic data)
- Semantic model for Cortex Analyst (text-to-SQL)
- Cortex Agent with Analyst + Email tools
- PubMed Search tool can be added after install (see NEXT_ACTIONS.md)
- Accessible via Snowflake Intelligence UI
Optional prerequisite:
- PubMed Biomedical Research Corpus (free Marketplace dataset)
- Enables medical literature search (38M+ articles)
- Without it, the Agent still works for analytics but cannot search research literature
Proceed with installation?
-
Wait for user confirmation.
-
Read solutions/clinical-quality-agent/scripts/setup.sql and execute it.
CRITICAL SESSION CONSTRAINT: This script uses SET session variables (e.g., $num_patients, $mortality_multiplier) that the data generation INSERT statements depend on. These variables MUST persist within the same session.
Execution strategy:
- Lines 1-208 (schema/table creation): Execute statement by statement — these are independent DDL.
- Lines 209-1963 (SET variables + data generation): Execute as a SINGLE call using
snowflake_sql_execute with all statements concatenated. The SET variables, TRUNCATE, and INSERT...GENERATOR statements MUST run in one session, otherwise the variables are lost and tables will be empty.
- Use
timeout_seconds: 1200 (data generation takes ~2-5 min on LARGE warehouse)
- Lines 1964+ (stage creation only): Execute statement by statement. This creates the SEMANTIC_MODEL_STAGE.
Log progress after each major section.
-
Upload semantic model YAML (CRITICAL — agent won't work without this):
Step 6a — Locate the YAML file:
solutions/clinical-quality-agent/scripts/semantic_model.yaml
Step 6b — Upload to stage via PUT:
PUT file://<repo_path>/solutions/clinical-quality-agent/scripts/semantic_model.yaml
@SF_SOLUTIONS.CLINICAL_QUALITY_SAFETY.SEMANTIC_MODEL_STAGE
AUTO_COMPRESS=FALSE OVERWRITE=TRUE;
Replace <repo_path> with the actual absolute path to the repository on disk.
If PUT fails, write the file contents to /tmp/semantic_model.yaml first, then PUT that:
PUT file:///tmp/semantic_model.yaml @SF_SOLUTIONS.CLINICAL_QUALITY_SAFETY.SEMANTIC_MODEL_STAGE AUTO_COMPRESS=FALSE OVERWRITE=TRUE;
Step 6c — Verify the file is on stage:
LIST @SF_SOLUTIONS.CLINICAL_QUALITY_SAFETY.SEMANTIC_MODEL_STAGE;
You MUST see semantic_model.yaml. If not, retry PUT before proceeding.
Step 6d — Execute solutions/clinical-quality-agent/scripts/deploy_agent.sql:
This creates the Cortex Agent. Execute it as a SINGLE call (the agent spec is one statement):
Step 6e — Verify agent was created:
SHOW AGENTS SCHEMA SF_SOLUTIONS.CLINICAL_QUALITY_SAFETY;
-
Verify table data:
SELECT TABLE_SCHEMA, TABLE_NAME, ROW_COUNT
FROM SF_SOLUTIONS.INFORMATION_SCHEMA.TABLES
WHERE TABLE_SCHEMA = 'CLINICAL_QUALITY_SAFETY'
ORDER BY TABLE_NAME;
-
[MANDATORY — DO NOT SKIP] Show the Snowflake Intelligence Agent URL.
Execute this query:
SELECT 'https://app.snowflake.com/'
|| LOWER(CURRENT_ORGANIZATION_NAME()) || '/' || LOWER(CURRENT_ACCOUNT_NAME())
|| '/#/agents/database/SF_SOLUTIONS/schema/CLINICAL_QUALITY_SAFETY/agent/CLINICAL_QUALITY_SAFETY_AGENT/details' AS AGENT_URL;
Display it to the user:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Cortex Agent:
<paste the full URL here>
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
This step is NON-OPTIONAL. The user must always see the Agent URL after install.
-
Show final summary:
Installation complete: Clinical Quality and Patient Safety Agent v1.0.0
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Cortex Agent:
<the URL from step 9>
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Next Actions:
1. Open the Cortex Agent URL above in Snowflake CoWork
2. Try: "How many patients died in the last year?"
3. Try: "Show me catheter infections that resulted in death"
4. Add PubMed medical literature search (see NEXT_ACTIONS.md)
Teardown: $sf-hcls-solutions:clinical-quality-agent teardown
