@meetdewey/vercel-ai

Dewey integration for the Vercel AI SDK. Adds two ready-to-use AI SDK tools that connect your application to a Dewey document collection.

Installation

npm install @meetdewey/vercel-ai
# or
pnpm add @meetdewey/vercel-ai

Tools

deweyRetrievalTool

Calls Dewey's hybrid semantic + BM25 search and returns chunks with citation metadata. Use this when you want the model to decide when and how to search — it receives query results as tool output and can call it multiple times.

import { streamText } from 'ai'
import { anthropic } from '@ai-sdk/anthropic'
import { deweyRetrievalTool } from '@meetdewey/vercel-ai'

const result = streamText({
  model: anthropic('claude-haiku-4-5-20251001'),
  system: 'Answer questions using the search tool. Always cite your sources by filename and section.',
  messages,
  tools: {
    search: deweyRetrievalTool({
      apiKey: process.env.DEWEY_API_KEY!,
      collectionId: process.env.DEWEY_COLLECTION_ID!,
      limit: 8,
    }),
  },
  maxSteps: 5,
})

Tool output per result:

{
  content: string        // chunk text
  score: number          // relevance score (RRF)
  filename: string       // source document filename
  sectionTitle: string   // section heading
  sectionLevel: number   // heading depth (1 = top-level)
  documentId: string     // document UUID
  sectionId: string      // section UUID
}

deweyResearchTool

Delegates a question to Dewey's agentic research endpoint. Dewey runs a multi-step tool-call loop internally — searching, reading sections, reasoning across documents — and returns a cited answer and source list. Use this for complex multi-document questions.

import { generateText } from 'ai'
import { openai } from '@ai-sdk/openai'
import { deweyResearchTool } from '@meetdewey/vercel-ai'

const { text } = await generateText({
  model: openai('gpt-4o-mini'),
  system: 'You are an assistant. Use the research tool for document questions.',
  messages,
  tools: {
    research: deweyResearchTool({
      apiKey: process.env.DEWEY_API_KEY!,
      collectionId: process.env.DEWEY_COLLECTION_ID!,
      depth: 'balanced',
    }),
  },
  maxSteps: 3,
})

Tool output:

{
  answer: string   // full cited answer text
  sources: Array<{
    filename: string
    sectionTitle: string
    sectionId: string
    documentId: string
  }>
}

Next.js example

// app/api/chat/route.ts
import { streamText } from 'ai'
import { anthropic } from '@ai-sdk/anthropic'
import { deweyRetrievalTool } from '@meetdewey/vercel-ai'

export async function POST(req: Request) {
  const { messages } = await req.json()

  const result = streamText({
    model: anthropic('claude-haiku-4-5-20251001'),
    system:
      'You are a helpful research assistant. Search the document collection ' +
      'to answer questions. Always cite the source filename and section.',
    messages,
    tools: {
      search: deweyRetrievalTool({
        apiKey: process.env.DEWEY_API_KEY!,
        collectionId: process.env.DEWEY_COLLECTION_ID!,
      }),
    },
    maxSteps: 5,
  })

  return result.toDataStreamResponse()
}

Options

deweyRetrievalTool(options)

| Option | Type | Default | Description | |---|---|---|---| | apiKey | string | required | Dewey project API key | | collectionId | string | required | Collection UUID to search | | limit | number | 10 | Default max chunks (1–50); model can override per call | | baseUrl | string | — | Override API base URL for local dev |

deweyResearchTool(options)

| Option | Type | Default | Description | |---|---|---|---| | apiKey | string | required | Dewey project API key | | collectionId | string | required | Collection UUID to research | | depth | 'quick' \| 'balanced' \| 'deep' \| 'exhaustive' | 'balanced' | Research thoroughness | | model | string | — | Model ID for Dewey's research loop (requires matching BYOK key) | | baseUrl | string | — | Override API base URL for local dev |

Resources

• Dewey documentation
• Free tier signup — no credit card required
• Vercel AI SDK docs