M
Mastra•2mo ago
Saulved

Can we parallelize semantic recall at all

Hey Team, We noticed that when we turn on semantic recall, the time it takes for a response to come in client side increases ~80%. Can we parallelize this in anyway to speed this up? Here is the setup: 
const memory = new Memory({
  storage: new PostgresStore({
    connectionString: process.env.DATABASE_URL!,
    schemaName: "mastra",
  }),
  vector: new PgVector({
    connectionString: process.env.DATABASE_URL!,
    schemaName: "mastra",
  }),
  embedder: 'openai/text-embedding-3-small',
  options: {
    lastMessages: 10,
    workingMemory: {
      enabled: true,
      scope: "resource",
      schema: userContextSchema,
    },
    semanticRecall: {
      topK: 3,
      messageRange: 2,
      indexConfig: {
        type: "hnsw",
        metric: "dotproduct",
      },
    },
  },
});
const memory = new Memory({
  storage: new PostgresStore({
    connectionString: process.env.DATABASE_URL!,
    schemaName: "mastra",
  }),
  vector: new PgVector({
    connectionString: process.env.DATABASE_URL!,
    schemaName: "mastra",
  }),
  embedder: 'openai/text-embedding-3-small',
  options: {
    lastMessages: 10,
    workingMemory: {
      enabled: true,
      scope: "resource",
      schema: userContextSchema,
    },
    semanticRecall: {
      topK: 3,
      messageRange: 2,
      indexConfig: {
        type: "hnsw",
        metric: "dotproduct",
      },
    },
  },
});
2 Replies
Mastra Triager
Mastra Triager•2mo ago
📝 Created GitHub issue: https://github.com/mastra-ai/mastra/issues/9719
GitHub
Can we parallelize semantic recall at all · Issue #9719 · mastra-...
This issue was created from Discord post 1435368650197434378: Hey Team, We noticed that when we turn on semantic recall, the time it takes for a response to come in client side increases ~80%. Can ...
_roamin_
_roamin_•2mo ago
Hey @Saulved ! I'm not really seeing this difference on my end, could you share a reproducible example with us? Thanks 🙏

Did you find this page helpful?