@f8synapse/nodes-ai

AI/ML workflow nodes — LLM completion, LangChain, agent loops, and vector search.

Installation

npm install @f8synapse/nodes-ai
# or
pnpm add @f8synapse/nodes-ai

Usage

import { createWorkflowEngine } from '@f8synapse/engine';
import { createConfiguredNodeRegistry } from '@f8synapse/nodes-core';
import { registerAiNodes } from '@f8synapse/nodes-ai';

// Start with core nodes, then layer in AI nodes
const registry = createConfiguredNodeRegistry();
registerAiNodes(registry);

const engine = createWorkflowEngine({
  nodeRegistry: registry,
  // Secrets adapter supplies OPENAI_API_KEY, ANTHROPIC_API_KEY, etc.
});

const execution = await engine.executeWorkflow({
  id: 'wf-ai-001',
  name: 'LLM Summarizer',
  nodes: [
    {
      id: 'n1',
      type: 'llm',
      config: {
        provider: 'openai',
        model: 'gpt-4o',
        prompt: 'Summarize: {{input.text}}',
      },
    },
    { id: 'n2', type: 'end', config: {} },
  ],
  edges: [{ from: 'n1', to: 'n2' }],
});

API

Functions

| Export | Description | | -------------------------------- | ---------------------------------------------------------- | | registerAiNodes | Registers all AI nodes into an existing NodeRegistry | | createConfiguredAiNodeRegistry | Creates a new NodeRegistry pre-loaded with AI nodes only |

Node Classes (selective imports)

| Node Class | Type Key | Description | | ------------------ | --------------- | -------------------------------------------------------------- | | LlmNode | llm | Generic LLM completion via engine provider registry | | AiCompletionNode | ai_completion | Direct AI completion with structured output support | | LangChainNode | langchain | Executes LangChain chains and pipelines | | AgentNode | agent | Autonomous agent loop with tool use | | VectorSearchNode | vector_search | Semantic similarity search via Pinecone or other vector stores |

License

MIT