Configuration

The CRGConfig object is passed to CRGProvider and controls all ConCRG behavior.

Full Configuration

interface CRGConfig {
  // Required
  appId: string;

  // Service URLs
  serviceUrl?: string;
  graphServiceUrl?: string;

  // Training inputs
  sourceCodePath?: string;
  docsUrl?: string;

  // User context
  userRole?: string;
  userId?: string;

  // Background capture
  backgroundCapture?: BackgroundCaptureConfig;

  // Feature flags
  useAgents?: boolean;
  useAgentsV2?: boolean;
  crawlerMode?: 'legacy' | 'stategraph';
  crawler?: CrawlerConfig;
}

---

Core Options

appId (required)

A unique identifier for your application. This namespaces all knowledge data — use a stable slug.

appId: 'my-crm-app'

serviceUrl

The URL of the running Train Service.

serviceUrl: 'http://localhost:3001'  // default

In production, point this at your deployed Train Service instance.

graphServiceUrl

URL of the optional Graph Service (Neo4j-backed).

graphServiceUrl: 'http://localhost:3002'  // default

---

Training Inputs

sourceCodePath

Absolute path to your application's TypeScript source directory. Used by the Code training source.

sourceCodePath: '/Users/me/projects/my-app/src'

docsUrl

The base URL of your product documentation site. Used by the Docs training source.

docsUrl: 'https://docs.myproduct.com'

---

User Context

userRole

The current user's role. ConCRG uses this to filter visible pages and accessible features during assist mode responses.

userRole: 'admin'         // sees all pages
userRole: 'viewer'        // restricted to read-only flows
userRole: 'sales_rep'     // CRM-specific workflow access

userId

A stable identifier for the current user. Used for per-user journey tracking (REMEMBER mode).

userId: 'user_abc123'

---

Background Capture

Continuously monitors the running app for route changes and DOM mutations, keeping the knowledge graph current without a full re-probe.

backgroundCapture: {
  enabled: true,
  debounceMs: 500,
  mutationThreshold: 10,
  includeScreenshots: true,
  syncStrategy: 'immediate',    // or 'periodic'
  minRecaptureIntervalMs: 5000,
  enableContentHashing: true,   // skip unchanged pages
}

Option	Default	Description
enabled	false	Turn on passive monitoring
debounceMs	500	Wait time after last mutation before capturing
mutationThreshold	10	Minimum DOM mutations to trigger capture
includeScreenshots	true	Capture screenshots with DOM snapshots
syncStrategy	'immediate'	'immediate' or 'periodic' (batches syncs)
minRecaptureIntervalMs	5000	Minimum time between captures of the same route
enableContentHashing	false	Skip unchanged pages using content hash

---

Crawler Options

crawlerMode

Controls the Probe's route exploration strategy.

crawlerMode: 'stategraph'  // default for new apps
crawlerMode: 'legacy'      // breadth-first queue walker

stategraph mode is more thorough — it models the app as a state machine and discovers transitions, not just routes.

crawler

Fine-tune the stategraph crawler:

crawler: {
  maxNodes: 50,                   // max unique routes to explore
  maxActionsPerNode: 10,          // max interactions per page
  maxInstancesPerRoutePattern: 3, // e.g., max 3 /deals/:id pages
}

---

Feature Flags

Flag	Default	Description
useAgents	false	Enable hybrid AST + LLM code analysis
useAgentsV2	false	V2 agent pipeline (more accurate, slower)

---

Example: Full Config

const crgConfig: CRGConfig = {
  appId: 'acme-crm',
  serviceUrl: 'http://localhost:3001',
  graphServiceUrl: 'http://localhost:3002',
  sourceCodePath: '/workspace/acme-crm/src',
  docsUrl: 'https://help.acme.com',
  userRole: 'admin',
  userId: 'user_dasun',
  backgroundCapture: {
    enabled: true,
    debounceMs: 500,
    mutationThreshold: 5,
    includeScreenshots: false,
    syncStrategy: 'periodic',
    minRecaptureIntervalMs: 10000,
    enableContentHashing: true,
  },
  crawlerMode: 'stategraph',
  crawler: {
    maxNodes: 80,
    maxActionsPerNode: 15,
  },
  useAgentsV2: true,
};