Never losecontext.Never again.

// Works with any model — memory around your own model call.
const context = await cencori.memory.recall(userId, message);

const reply = await openai.chat.completions.create({
    model: 'gpt-4o',
    messages: [
        { role: 'system', content: context },
        { role: 'user', content: message },
    ],
});

// Extract the durable facts from the exchange and store them.
await cencori.memory.remember(userId, {
    user: message,
    assistant: reply.choices[0].message.content,
});

// Without memory: every new chat starts from zero.
await cencori.chat.completions.create({
    model: 'gpt-4o',
    messages,
});

// On the gateway: recall + remember collapse into one flag.
await cencori.chat.completions.create({
    model: 'gpt-4o',
    messages,
    memory: { userId: session.user.id },
});

session #47 — earlier this week
─────────────────────────────────────────────
user   Help me redo the pricing tiers on my app.
model  [200 messages of back-and-forth. You land
        on tiered Stripe billing, 14-day trial,
        metered overages at $0.01/unit.]


session #48 — new chat, moments ago
─────────────────────────────────────────────
user   Let's keep going.
model  Picking up from where we left off with the
        tiered Stripe billing — the metered overage
        rate you set was $0.01/unit. Want to add
        annual discounting next?

import { Chat, useMemory } from 'cencori/react';

// One flag, memory-aware chat.
<Chat model="gpt-4o" memory={{ userId }} />

// Build a "what do you remember about me" panel from the hook:
// list · search · forget(id) · exportAll (GDPR export).
const { memories, forget, exportAll } = useMemory({ userId });

// Hard-delete a memory by id. Real deletion, audit-logged.
await cencori.memory.forget(memoryId);

// Surface stale, low-value memories to prune (candidates only).
const { suggestions } = await cencori.memory.forgetSuggestions({ userId });