Memory
Agents that actually remember
Vektor Memory v2 provides production-ready storage for long-horizon agent tasks. Handle multi-session context, user preferences, and accumulated knowledge — without managing vector DBs yourself.
Memory stores
Episodic
Time-stamped records of past interactions. Agents recall what happened in specific sessions or with specific users.
“Last week you asked about refund policy for order #8821...”
Semantic
Knowledge about the world, products, and facts. Stored as vector embeddings; retrieved by relevance.
“Product X has a 30-day return window. It is compatible with...”
Procedural
How-to knowledge and learned workflows. Agents internalize successful resolution patterns over time.
“When the customer mentions shipping delay, first check...”
SDK example
import { VektorMemory } from "vektor-sdk";
const memory = new VektorMemory({
agentId: "support-agent",
stores: ["episodic", "semantic"],
compression: "auto",
});
// Write
await memory.add({ type: "episodic", content: interaction });
// Read
const context = await memory.recall({
query: "user's previous orders",
topK: 5,
});