Weavlet

^[WHY] AI products should be easy to build and trustworthy to operate. ^[HOW] That requires structuring how intelligence behaves, not just how it responds. ^[WHAT] So we build foundational systems for expressive and reliable AI.

Agent Craft Studio

Visually configure persona, memory, and logic states in a unified studio. Spend more time crafting, less time wrestling with prompt files.

View Details

Custom Model Training

We will train any specialized model for your use case so you can cut costs and own your data. Purpose-built models, fully yours.

3 slots left for April

View Details

Primitives

Open Source building blocks that make your development with AI so much easier.

Factsheet
Open sourcing soon
Turn Definition for Speech
Scoring with LLM
Context Scoping
Ads

FactSheet maintains a live, structured profile for each user by extracting facts from conversations and resolving conflicts deterministically. Instead of searching through chat history, your AI reads current truths from a single JSON object.

Implementation

1Install the package

npm install @weavlet/factsheet zod

2Define your schema and initialize FactSheet

import { FactSheet, MemoryAdapter } from '@weavlet/factsheet'
import { z } from 'zod'

const factSheet = new FactSheet({
  storage: new MemoryAdapter(),
  schema: z.object({
    name: z.string().optional(),
    role: z.enum(['founder', 'engineer', 'designer']).optional(),
    intent: z.enum(['learn', 'buy', 'support']).optional(),
  }),
  extractor: {
    model: 'gpt-4o-mini',
  },
})

3Extract facts and use them in your app

// Extract facts from conversation
await factSheet.observe({
  userId: 'user_123',
  input: "I'm Alice, CTO at Acme. Looking to migrate from Heroku.",
})

// Get current profile
const profile = await factSheet.get('user_123')
// → { name: "Alice", role: "founder", intent: "migrate" }

Read FactSheet docs

Open sourcing soon

Turn Definition for Speech
Scoring with LLM
Context Scoping
Ads

Let's collaborate