Spark Pricing Agent

A small Spark-ready pricing app for sales reps. The deployed runtime is Python/FastAPI so the Zapier MCP call path matches AP-Sparkle's known-working Snowflake bridge.

Current State

• Working FastAPI app with a sales rep pricing form.
• POST /api/price validates requests with Pydantic.
• PRICING_PROVIDER=mock works locally with deterministic sample guidance.
• PRICING_PROVIDER=zapier-mcp calls Zapier MCP with the official Python MCP client.
• /api/health is wired for Spark health checks.
• Dockerfile forces Spark/Kaniko to deploy the Python app instead of the older Next.js prototype.

Local Setup

pip install .
uvicorn app.main:app --host 0.0.0.0 --port 3000

Open http://localhost:3000.

Environment Variables

Name	Required	Description
PRICING_PROVIDER	Yes	mock or zapier-mcp. Defaults to mock if omitted.
ZAPIER_MCP_SERVER_URL	For Zapier MCP	Private Zapier MCP server URL. This usually includes auth in the URL and should be treated like a secret.
ZAPIER_MCP_URL	For Zapier MCP	Alias used by AP-Sparkle for the Zapier MCP server URL.
ZAPIER_MCP_KEY	For Zapier MCP	Bearer token for Zapier MCP when auth is stored separately from the URL.
ZAPIER_MCP_TOKEN	For Zapier MCP	Alias used by AP-Sparkle for the Zapier MCP bearer token.
ZAPIER_MCP_API	For Zapier MCP	Bearer token for Zapier MCP when auth is stored separately from the URL.
ZAPIER_MCP_BEARER_TOKEN	Optional	Bearer token if your Zapier MCP setup gives auth separately from the URL.
ZAPIER_MCP_TOOL_NAME	Optional	Exact Zapier MCP tool name to call. Defaults to snowflake_execute_sql.
PRICING_MODEL_VERSION	Optional	Included in mock/API responses for audit and debugging.

Pricing Request Shape

POST /api/price

All fields are optional. Blank form fields are omitted from the request.

{
  "accountId": "123456",
  "planTier": "Enterprise",
  "region": "NA",
  "productLine": "Marketing",
  "resellerId": "RSID-123",
  "contactLimit": 50000,
  "listPrice": 50000,
  "discountRate": 10,
  "smsFlag": true,
  "smsCredits": 1000,
  "whatsapp": false,
  "termLength": 12,
  "arr": 42000,
  "priceRealization": 0.84
}

Normalized Pricing Result

The UI expects /api/price to return this object under a result key:

{
  "quoteId": "Q-123",
  "recommendedDiscount": 12,
  "maxDiscount": 18,
  "floorPrice": 41000,
  "recommendedPrice": 44000,
  "approvalRequired": false,
  "approvalLevel": "None",
  "reasonCodes": ["Segment: mid market", "Term: 12 months"],
  "modelVersion": "snowflake-2026-05",
  "provider": "zapier-mcp",
  "calculatedAt": "2026-05-26T12:00:00.000Z"
}

Zapier MCP Bridge

Use this to call the same Snowflake-authed Zapier MCP path AP-Sparkle uses.

Spark secrets:

PRICING_PROVIDER=zapier-mcp
ZAPIER_MCP_SERVER_URL=<private Zapier MCP server URL>
ZAPIER_MCP_KEY=<Zapier MCP bearer token>
ZAPIER_MCP_TOOL_NAME=snowflake_execute_sql

The provider calls snowflake_execute_sql with statement, output_hint, and instructions. The statement is:

CALL AC.SANDBOX.CALCULATE_PRICING_GUIDANCE(
  PARSE_JSON($${...pricing request json...}$$)
);

Spark Notes

• Health check path: /api/health.
• Set PRICING_PROVIDER=zapier-mcp in Spark to use the Zapier MCP Snowflake connection.
• For the MCP bridge, set PRICING_PROVIDER=zapier-mcp and add the ZAPIER_MCP_* secrets above.
• Direct Snowflake can be added later behind the same provider interface in app/pricing.py.