@caesura-io/core

Framework-agnostic core for Caesura — shared analyze, inject, and credit-metering logic used by all SDK integrations.

This package is not meant to be used directly. It is the shared engine consumed by:

• @caesura-io/ai-sdk — Vercel AI SDK middleware
• @caesura-io/openai — OpenAI Node SDK wrapper

What's inside

| Module | Purpose | |--------|---------| | CaesuraClient | HTTP client that calls the /api/analyze endpoint | | MemoryCaesuraStore | In-memory conversation state with LRU + idle-time eviction | | createCaesuraEngine | Orchestrator: cadence checks, observe/analyze cycle, buffering, event emission | | createCreditMeter | Accumulates and queries credit-usage metrics | | createDebugLogger | Structured onEvent logger for debugging | | Helpers | hashMessage, selectActive, renderAnalysis, renderBlock, buildAnalyzeMessages | | Types | CaesuraConfig, CaesuraEvent, InjectConfig, SendConfig, etc. |

Install

npm i @caesura-io/core

Usage

Most consumers should use the framework-specific adapters. If you're building your own integration:

import { createCaesuraEngine, selectActive, renderBlock } from '@caesura-io/core';

const engine = createCaesuraEngine({
  baseUrl: 'https://dev.caesura.io',
  apiKey: process.env.CAESURA_API_KEY,
});

// 1. Observe a conversation turn
await engine.observe('conversation-id', [
  { speakerRole: 'user', speakerName: 'Customer', text: 'I need help preparing for the next meeting' },
]);

// 2. Retrieve buffered recommendations
const state = engine.store.get('conversation-id');
const active = selectActive(state, engine.config.inject, Date.now());
const blocks = renderBlock(active, engine.config.inject);
// → blocks contains rendered recommendation text ready for injection

License

Apache-2.0