@pipeline-builder/ai-core

📖 View documentation

Shared AI provider registry for Pipeline Builder: lazily initializes SDK wrappers from environment variables and exposes model resolution helpers used by AI-assisted pipeline and plugin generation.

Supported Providers

| Provider | Env Variable | SDK | |---|---|---| | Anthropic | ANTHROPIC_API_KEY | @ai-sdk/anthropic | | OpenAI | OPENAI_API_KEY | @ai-sdk/openai | | Google | GOOGLE_GENERATIVE_AI_API_KEY | @ai-sdk/google | | xAI (Grok) | XAI_API_KEY | @ai-sdk/xai | | Amazon Bedrock | AWS_ACCESS_KEY_ID | @ai-sdk/amazon-bedrock |

Usage

import {
  getAvailableProviders,
  getProviderModels,
  resolveModel,
  createModelWithKey,
} from '@pipeline-builder/ai-core';

// List providers with configured API keys
const providers = getAvailableProviders();

// Get models for a provider (static catalog, no env vars needed)
const models = getProviderModels('anthropic');

// Resolve a model from the registry (requires env var)
const model = resolveModel('anthropic', 'claude-sonnet-4-20250514');

// Create a one-off model with a custom API key
const custom = createModelWithKey('openai', 'gpt-4o', 'sk-...');

API

• getAvailableProviders() — Returns providers with configured env vars
• getProviderModels(providerId) — Returns model list from static catalog
• resolveModel(providerId, modelId) — Returns a LanguageModel from the registry
• createModelWithKey(providerId, modelId, apiKey) — Creates a temporary LanguageModel (not cached)

License

Apache-2.0. See LICENSE.