Skip to content

configuration.md

Latest commit

 

History

History
301 lines (239 loc) · 5.92 KB

configuration.md

File metadata and controls

301 lines (239 loc) · 5.92 KB

Configuration Reference

Complete reference for config.yaml options.

Basic Structure

# Requirements - validated before workflow runs
requirements:
  env: []           # Required environment variables
  files: []         # Required files

# Workflow metadata
workflow:
  name: "My Workflow"
  description: "What this workflow does"
  version: "0.1.0"

# Default settings for all phases
defaults:
  parallel: false
  max_workers: 4
  timeout: 120
  critical: true
  min_success: 0

# Phases - executed in order
phases:
  - name: collect
    tools:
      - fetch_data.py

# AI model configuration
models:
  default:
    model: claude-sonnet-4-20250514
    temperature: 0.3
    max_tokens: 8000

# Workflow-specific settings
settings:
  example_setting: "value"

# Deployment (optional)
deployment:
  platform: modal
  schedule: weekly
  schedule_day: sunday
  schedule_time: "08:00"

Requirements

Define environment variables and files that must exist before the workflow runs.

requirements:
  env:
    - ANTHROPIC_API_KEY
    - GITHUB_TOKEN
  files:
    - credentials.json
    - token.json

The workflow fails fast with clear error messages if any requirements are missing.

Workflow Metadata

workflow:
  name: "My Workflow"           # Required - display name
  description: "Description"    # What the workflow does
  version: "0.1.0"              # Semantic version

Defaults

Default settings applied to all phases. Individual phases can override these.

defaults:
  parallel: false       # Run tools sequentially by default
  max_workers: 4        # Max parallel workers when parallel: true
  timeout: 120          # Tool timeout in seconds
  critical: true        # Stop workflow on failure
  min_success: 0        # Minimum tools that must succeed
  retry: 0              # Retry count on failure

Phases

Phases are executed in order. Each phase runs a list of tools.

phases:
  - name: collect
    parallel: true        # Override: run tools in parallel
    max_workers: 5
    timeout: 300
    critical: false       # Continue if some tools fail
    min_success: 1        # At least 1 tool must succeed
    tools:
      - fetch_github.py
      - fetch_reddit.py
      - fetch_hackernews.py

  - name: process
    parallel: false       # Sequential execution
    timeout: 120
    tools:
      - validate_data.py
      - transform_data.py

  - name: analyze
    timeout: 180
    tools:
      - extract_signals.py

  - name: deliver
    parallel: true
    max_workers: 3
    tools:
      - export_results.py
      - send_email.py

Phase Settings

Setting Type Default Description
name string required Phase identifier
tools list required Python scripts to execute
parallel bool false Run tools in parallel
max_workers int 4 Max concurrent tools
timeout int 120 Timeout per tool (seconds)
critical bool true Stop workflow on failure
min_success int 0 Minimum tools that must succeed
retry int 0 Retry count on failure

Models

Configure Claude AI models for different stages.

models:
  default:
    model: claude-sonnet-4-20250514
    temperature: 0.3
    max_tokens: 8000

  # Override for specific stages
  analyze:
    model: claude-opus-4-5-20251101
    temperature: 0.2
    max_tokens: 16000

The framework looks for a stage-specific config first, then falls back to default.

Settings

Workflow-specific configuration. Access these in your tools.

settings:
  max_results: 100
  output_format: json
  deduplicate: true
  custom_option: "value"

Deployment

Configure Modal deployment for scheduled execution.

deployment:
  platform: modal              # Currently only 'modal' supported

  # Schedule (optional - omit for on-demand only)
  schedule: weekly             # Options: daily, weekly, monthly
  schedule_day: sunday         # For weekly: monday-sunday
  schedule_time: "08:00"       # Time in UTC (24-hour format)

  # Optional overrides
  timeout: 1800                # Seconds (default: 1800 = 30 minutes)
  python_version: "3.11"       # Python version for container

Schedule Options

Schedule Additional Options
daily schedule_time
weekly schedule_day, schedule_time
monthly schedule_day_of_month, schedule_time

Day Options (for weekly)

monday, tuesday, wednesday, thursday, friday, saturday, sunday

Complete Example

# Requirements
requirements:
  env:
    - ANTHROPIC_API_KEY
    - GITHUB_TOKEN
    - REDDIT_CLIENT_ID
    - REDDIT_CLIENT_SECRET
  files:
    - credentials.json
    - token.json

# Metadata
workflow:
  name: "SaaS Hunter"
  description: "Discover trending SaaS ideas from multiple sources"
  version: "2.0.0"

# Default phase settings
defaults:
  parallel: false
  max_workers: 4
  timeout: 120
  critical: true

# Execution phases
phases:
  - name: collect
    parallel: true
    max_workers: 5
    timeout: 300
    critical: false
    min_success: 1
    tools:
      - fetch_github_trending.py
      - fetch_hackernews.py
      - fetch_reddit.py
      - fetch_product_hunt.py
      - fetch_google_trends.py

  - name: process
    timeout: 120
    tools:
      - validate_data.py

  - name: analyze
    timeout: 300
    tools:
      - extract_signals.py

  - name: deliver
    parallel: true
    tools:
      - export_results.py
      - send_email.py

# AI models
models:
  default:
    model: claude-sonnet-4-20250514
    temperature: 0.3
    max_tokens: 8000
  analyze:
    model: claude-opus-4-5-20251101
    temperature: 0.2
    max_tokens: 16000

# Custom settings
settings:
  max_results_per_source: 50
  output_format: json
  deduplicate: true

# Deployment
deployment:
  platform: modal
  schedule: weekly
  schedule_day: sunday
  schedule_time: "08:00"
  timeout: 1800