Memory Example

This example shows a simple memory workflow: create a store, write scoped memories, search them, and build a recall block for downstream chat.

Run the Example

cd examples
export COGNIPEER_API_KEY=your-api-key
export COGNIPEER_BASE_URL=https://your-console.example.com
export COGNIPEER_VECTOR_PROVIDER_KEY=pinecone-main
export COGNIPEER_EMBEDDING_MODEL_KEY=text-embedding-3-small
npm run example:memory

Use the console host or origin for COGNIPEER_BASE_URL. The SDK adds /api/client/v1 internally.

Example Code

import { ConsoleClient } from '@cognipeer/console-sdk';

const client = new ConsoleClient({
  apiKey: process.env.COGNIPEER_API_KEY!,
  baseURL: process.env.COGNIPEER_BASE_URL,
});

const store = await client.memory.stores.create({
  name: 'Support Memory',
  vectorProviderKey: process.env.COGNIPEER_VECTOR_PROVIDER_KEY!,
  embeddingModelKey: process.env.COGNIPEER_EMBEDDING_MODEL_KEY!,
});

await client.memory.add(store.key, {
  content: 'User prefers concise responses.',
  scope: 'user',
  scopeId: 'user_123',
  source: 'manual',
});

const recall = await client.memory.recall(store.key, {
  query: 'What should I remember before answering?',
  scope: 'user',
  scopeId: 'user_123',
});

console.log(recall.context);

When to Use Which Scope

• user: long-lived user preferences and facts
• agent: reusable working knowledge for a single agent
• session: conversation-local state
• global: shared information not tied to a single actor