@upendra.manike/agent-orchestrator

Autonomous agent orchestrator: chained LLM calls, tool-use, memory, streaming telemetry.

Install

npm i @upendra.manike/agent-orchestrator

Usage

import { runAgent, streamAgent, defineTool, makeMemory, makeLLMMock } from '@upendra.manike/agent-orchestrator';

const add = defineTool({
  name: 'math.add',
  schema: { type: 'object', properties: { a: { type: 'number' }, b: { type: 'number' } }, required: ['a','b'] },
  async run({ a, b }) { return { text: String(a + b) }; }
});

const memory = makeMemory();
const llm = makeLLMMock({ responses: [{ role: 'assistant', content: 'Hello from agent.' }] });

const result = await runAgent({ goal: 'Say hello', tools: [add], memory, llm });
console.log(result.finalAnswer);

API

• runAgent(input) – plan, run, and return final answer
• streamAgent(input, onChunk) – like runAgent but streams partial output
• defineTool({ name, schema, run }) – register a tool
• makeMemory() – in-memory vector-less memory with simple search
• makeLLMMock(opts) – deterministic LLM mock for tests

License

MIT

🔗 Explore All JSLib Libraries

• JSLib Usage Guide