Templates

Templates are curated starter workflows that demonstrate CFFBRW's capabilities. Each template is designed to teach a specific pattern.

Using a Template

1. Go to Templates in the sidebar
2. Browse the available templates — each shows what it demonstrates
3. Click Use Template
4. The markdown is loaded into a new workflow
5. Click Compile and Run to see it in action
6. Edit the markdown to customize it for your needs

Available Templates

HackerNews Trend Analyzer

Fetches live data from HackerNews and uses AI to analyze tech trends.

Demonstrates: Parallel fan-out (2 HTTP calls run simultaneously), http_request, ai_call, transform, conditional with branch skipping

What it does:

1. Fetches top story IDs from HackerNews
2. Fetches details for two stories in parallel
3. AI analyzes trends across both stories
4. Calculates engagement score
5. Conditionally marks as "viral" or "normal"

Vendor Invoice Processor

Looks up a vendor, creates an invoice, and searches for related products using MCP tools.

Demonstrates: Parallel MCP calls, ai_call, conditional, human_approval with approve/reject UI

What it does:

1. Looks up vendor details via MCP
2. Searches product catalog in parallel
3. AI drafts an invoice summary
4. Budget check — if over $5,000, requires manager approval
5. Creates the invoice

API Schema Mismatch Recovery

Calls an API that returns an unexpected nested format, triggering Tier 2 AI recovery.

Demonstrates: Tier 2 AI-in-the-loop recovery, automatic error diagnosis and correction

What it does:

1. Fetches data from randomuser.me
2. Attempts to extract fields — the API's nested format doesn't match the expected shape
3. Tier 2 AI recovery kicks in: sees the error, reads the actual response, generates a corrected output
4. Workflow continues automatically as if nothing went wrong

Parallel API Race

Fetches data from three public APIs simultaneously and compares results.

Demonstrates: Maximum parallelism (3 independent HTTP calls), transform with array operations, conditional branching

Data Pipeline (4-way Fan-out)

Fetches raw data, processes it through four parallel AI stages, then combines results.

Demonstrates: 4-way parallel fan-out, AI data processing, 4-way fan-in join

Procurement with Approval

Vendor lookup, product search, budget check, and approval gate.

Demonstrates: Sequential MCP chain, conditional budget check, human approval gate

Building Your Own

Templates are just markdown. You can start from any template and modify it to suit your use case:

• Change URLs to your own APIs
• Swap step descriptions to match your business logic
• Add or remove steps
• Introduce conditional branches or approval gates

The AI compiler adapts to whatever you write — there's no special template syntax.