@foxruv/iris-core
Advanced tools
Intelligent AI orchestration with multi-provider LM management, drift detection, auto-retraining, and performance tracking for production AI systems
Core orchestration and multi-provider LM management for Iris AI
Features • Installation • Quick Start • API • Migration
@iris/core is the modular core package extracted from the Iris monolith, providing:
Supported Providers:
Provider Features:
npm install @iris/core
Requirements:
import { createIrisOrchestrator } from '@foxruv/iris-core'
// Create orchestrator instance
const iris = createIrisOrchestrator({
dbBasePath: './data/iris',
defaultAutoRetrain: true,
defaultAutoPromote: true
})
// Configure project
iris.configureProject({
projectId: 'my-project',
autoRetrain: true,
autoPromote: true,
retrainingThreshold: 0.1, // 10% accuracy drop
promotionThreshold: 0.1, // 10% improvement
minEvaluations: 10
})
// Evaluate project health
const report = await iris.evaluateProject('my-project')
console.log(`Health Score: ${report.healthScore}/100`)
console.log(`Status: ${report.overallHealth}`)
console.log(`Drift Alerts: ${report.driftAlerts.length}`)
console.log(`Recommendations: ${report.recommendedActions.length}`)
// Auto-promote better prompts
const promoted = await iris.autoPromotePrompts('my-project')
console.log(`Promoted ${promoted.length} experts`)
// Auto-retrain drifting experts
const retrained = await iris.autoRetrainExperts('my-project')
console.log(`Retrained ${retrained.length} experts`)
// Clean up
iris.close()
import { ClaudeProvider } from '@foxruv/iris-core/providers'
const provider = new ClaudeProvider(
process.env.ANTHROPIC_API_KEY,
'claude-sonnet-4-5-20250929'
)
const signature = {
instructions: 'Classify sentiment as positive, negative, or neutral',
input: { text: 'Input text to classify' },
output: { sentiment: 'Sentiment classification' }
}
const result = await provider.predict(
signature,
{ text: 'I love this product!' },
undefined, // custom instructions (optional)
0.0, // temperature
1024 // max tokens
)
console.log(result) // { sentiment: "positive" }
import { Qwen3Provider } from '@foxruv/iris-core/providers'
const provider = new Qwen3Provider(
'http://localhost:1234', // LM Studio endpoint
'qwen2.5-32b-instruct', // model name
5 // max concurrent requests
)
// Single prediction
const result = await provider.predict(signature, input)
// Batch predictions (5x throughput)
const results = await provider.batchPredict(
signature,
[input1, input2, input3, ...],
undefined, // custom instructions
0.3, // temperature
2048 // max tokens
)
// Batch with retry (auto-recover from failures)
const results = await provider.batchPredictWithRetry(
signature,
inputs,
undefined,
0.3,
2048,
2 // max retries per prediction
)
// Health check
const isHealthy = await provider.healthCheck()
import { getLMProvider } from '@foxruv/iris-core/providers'
// Auto-detect provider from environment
const manager = getLMProvider()
const provider = manager.getProvider()
// Get performance metrics
const metrics = manager.getPerformanceMetrics('anthropic')
console.log(`Latency: ${metrics.averageLatencyMs}ms`)
console.log(`Success Rate: ${metrics.successRate * 100}%`)
// Compare all providers
const comparison = manager.compareProviders()
console.log(`Fastest: ${comparison.fastest}`)
console.log(`Most Reliable: ${comparison.mostReliable}`)
// Switch providers
manager.switchProvider('lmstudio')
import { calculateHealthScore, getHealthLevel, incrementVersion } from '@foxruv/iris-core'
// Calculate health score
const score = calculateHealthScore({
driftAlerts: 2,
staleReflexions: 5,
avgValidity: 0.8,
highPriorityRotations: 1
})
console.log(score) // 47 (0-100)
// Get health level
const level = getHealthLevel(score)
console.log(level) // "poor"
// Increment version
const newVersion = incrementVersion('v1.0.5')
console.log(newVersion) // "v1.0.6"
createIrisOrchestrator(config?): IrisOrchestratorCreates a new Iris orchestrator instance.
Config:
interface IrisPrimeConfig {
dbBasePath?: string // Default: './data/iris'
defaultAutoRetrain?: boolean // Default: false
defaultAutoPromote?: boolean // Default: false
scheduleIntervalMs?: number // Default: 86400000 (24h)
logPath?: string // Default: './logs'
notifiers?: IrisNotifier[] // Default: []
}
IrisOrchestrator.evaluateProject(projectId): Promise<IrisReport>Evaluates a project's health and returns a comprehensive report.
Returns:
interface IrisReport {
projectId: string
timestamp: Date
overallHealth: 'excellent' | 'good' | 'fair' | 'poor' | 'critical'
healthScore: number // 0-100
driftAlerts: DriftAlert[]
promptRecommendations: PromptRecommendation[]
reflexionStatus: ReflexionStatus
rotationRecommendations: RotationRecommendation[]
transferablePatterns: TransferablePattern[]
recommendedActions: RecommendedAction[]
}
ClaudeProviderAnthropic Claude API wrapper with fetch-based implementation.
Methods:
predict(signature, input, customInstructions?, temperature?, maxTokens?): Promise<Record<string, any>>Qwen3ProviderOpenAI-compatible local model provider with concurrency control.
Methods:
predict(signature, input, customInstructions?, temperature?, maxTokens?, schema?): Promise<Record<string, any>>batchPredict(signature, inputs, customInstructions?, temperature?, maxTokens?): Promise<Array<Record<string, any>>>batchPredictWithRetry(signature, inputs, customInstructions?, temperature?, maxTokens?, maxRetries?): Promise<Array<Record<string, any>>>healthCheck(): Promise<boolean>LMProviderManagerMulti-provider orchestration with performance tracking.
Methods:
getProvider(): anygetProviderByName(name): any | undefinedgetAvailableProviders(): ModelProvider[]switchProvider(provider): voidrecordPerformance(provider, latencyMs, success, qualityScore?): voidgetPerformanceMetrics(provider?): PerformanceMetrics | PerformanceMetrics[]compareProviders(): ProviderComparisoncalculateHealthScore(factors): number - Calculate health score (0-100)getHealthLevel(score): HealthStatus - Get health level from scoreincrementVersion(version): string - Increment semver versionvalidateProjectId(id): boolean - Validate project ID formatvalidateSemver(version): boolean - Validate semver stringBefore (Monolith):
import { IrisPrime } from './orchestrators/iris-prime.js'
const iris = new IrisPrime({ dbBasePath: './data' })
After (@iris/core):
import { createIrisOrchestrator } from '@foxruv/iris-core'
const iris = createIrisOrchestrator({ dbBasePath: './data' })
IrisPrime → IrisOrchestratorcreateIrisOrchestrator() instead of new IrisPrime()@iris/core instead of local pathThe monolith provides a backward compatibility shim at src/orchestrators/iris-prime.ts that re-exports from @iris/core:
// Still works (with deprecation warning)
import { IrisPrime } from './orchestrators/iris-prime.js'
Deprecation Timeline:
# Run tests
npm test
# Watch mode
npm run test:watch
# Coverage report
npm run test:coverage
Test Results:
# Build TypeScript
npm run build
# Type checking only
npm run typecheck
Build Output:
dist/.d.ts)| Metric | Target | Actual | Status |
|---|---|---|---|
| Bundle Size | <500KB | 154KB | ✅ |
| Build Time | <2s | ~1.5s | ✅ |
| Test Execution | <5s | 1.93s | ✅ |
| Coverage | >95% | 99% | ✅ |
| Zero Dependencies | Yes | Yes | ✅ |
@iris/core/
├── src/
│ ├── index.ts # Main exports
│ ├── orchestrator.ts # IrisOrchestrator class
│ ├── providers.ts # Provider implementations
│ ├── types.ts # Type definitions
│ ├── utils.ts # Utility functions
│ ├── providers/ # Provider modules
│ │ ├── claude.ts
│ │ ├── qwen.ts
│ │ ├── manager.ts
│ │ └── index.ts
│ ├── types/ # Type modules
│ │ ├── config.ts
│ │ ├── metrics.ts
│ │ ├── reports.ts
│ │ └── index.ts
│ └── utils/ # Utility modules
│ ├── health.ts
│ └── validation.ts
├── tests/ # Comprehensive test suite
├── dist/ # Build output (ESM + CJS)
├── package.json
├── tsconfig.json
├── vitest.config.ts
└── README.md
Runtime: Zero external dependencies
Development:
typescript - Type checking and compilationvitest - Fast unit testing@vitest/coverage-v8 - Code coverage reporting@types/node - Node.js type definitionsMIT © [Iris Team]
Contributions are welcome! Please:
npm test)@iris/council - Expert council coordination@iris/voice - Voice interface integration@foxruv/iris - Federated learning componentsBuilt with ❤️ by the Iris Team
FAQs
Intelligent AI orchestration with multi-provider LM management, drift detection, auto-retraining, and performance tracking for production AI systems
We found that @foxruv/iris-core demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 2 open source maintainers collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Research
/Security News
Campaign of 108 extensions harvests identities, steals sessions, and adds backdoors to browsers, all tied to the same C2 infrastructure.
Security News
OpenAI rotated macOS signing certificates after a malicious Axios package reached its CI pipeline in a broader software supply chain attack.
Security News
Open source is under attack because of how much value it creates. It has been the foundation of every major software innovation for the last three decades. This is not the time to walk away from it.