Rivière

Appearance

Step 3: Extract Components ​

Objective ​

Find all component instances using patterns from Step 2 and add them to the graph via CLI.

Prerequisites ​

  • Do not use plan mode. Execute directly.
  • Read .riviere/config/metadata.md for domains and conventions
  • Read .riviere/config/component-definitions.md for extraction rules
  • CLI installed: npm install @living-architecture/riviere-cli

Initialize Graph ​

Create graph with sources and domains:

bash
npx riviere builder init \
  --source "https://github.com/your-org/your-repo" \
  --domain '{"name":"[name]","description":"[desc]","systemType":"domain"}'

Add additional sources if needed:

bash
npx riviere builder add-source --repository "https://github.com/your-org/your-repo"

Add additional domains if needed:

bash
npx riviere builder add-domain --name "[name]" --system-type "[domain|bff|ui|other]" --description "[desc]"

Define custom types (for accepted proposals in component-definitions.md):

bash
npx riviere builder define-custom-type \
  --name "[CustomTypeName]" \
  --description "[description]" \
  --required-property "propertyName:type:description" \
  --optional-property "propertyName:type:description"

Example (BackgroundJob with schedule property):

bash
npx riviere builder define-custom-type \
  --name "BackgroundJob" \
  --description "Scheduled background task running on an interval" \
  --required-property "schedule:string:Cron expression or interval"

Extract Components ​

No planning or counting. Extract directly using patterns from Step 2.

1. Find All Instances ​

For each component type, grep for its code signature:

bash
# Examples - actual patterns come from Step 2
grep -rn "@Controller" src/
grep -rn "extends BaseUseCase" src/
grep -rn "@EventHandler" src/

This gives file paths and line numbers for ALL occurrences.

2. Process Each Match ​

For each grep result:

  1. Read the file at that location
  2. Extract component details (name, domain, HTTP method, etc.)
  3. Add via CLI
bash
npx riviere builder add-component \
  --type "[API|UseCase|DomainOp|Event|EventHandler|UI|Custom]" \
  --domain "[domain]" \
  --module "[module]" \
  --name "[name]" \
  --repository "[repo]" \
  --file-path "[path]" \
  --line-number "[line]"

Type-specific options:

  • API: --api-type (required), --http-method, --http-path
  • DomainOp: --entity, --operation-name
  • Event: --event-name, --event-schema
  • EventHandler: --subscribed-events
  • UI: --route
  • Custom: --custom-type, --custom-property (repeatable, format: key:value)

3. Complete Each Type ​

Finish ALL matches for one component type before starting the next:

text
APIs (@Controller): 12 matches → 18 endpoints extracted
UseCases (extends BaseUseCase): 8 matches → 8 use cases extracted

If grep pattern doesn't work (e.g., complex dynamic patterns), note it and ask user if a script is needed.

Verify Extraction ​

Generate summary:

bash
npx riviere builder component-summary > ".riviere/step-3-summary.json"

Check for:

  • Domains with zero components
  • Component types with zero instances
  • Unexpected counts suggesting missed patterns

Offer to run a sub-agent to scan for components that may have been missed.

Feedback ​

If user reports missing components, update .riviere/config/component-definitions.md with corrected patterns and re-extract.

Output ​

Graph: .riviere/[project-name]-[commit].json

Completion ​

Present extraction summary showing component counts by domain and type.

Step 3 complete. Wait for user feedback before proceeding.