Skip to Content
Getting Started

Getting Started

Get your AI-powered documentation chatbot running in minutes. This guide covers installation, configuration, and integration with your Next.js application.

Prerequisites: Node.js 20+, a Next.js 15+ project (App Router), and API keys for Gemini/OpenAI and Qdrant.

Install Docsy

Install the CLI globally and add the core package to your project.

pnpm add -g @gaureshart/docsy-cli
pnpm add @gaureshart/docsy-core ai

The ai package is Vercel’s AI SDK, required for streaming responses via useChat().

Initialize Configuration

Run the interactive setup wizard from your project root.

npx @gaureshart/docsy-cli init

This creates two files:

  • docsy.config.ts
  • .env.example

What happens:

  • Interactive prompts ask for your GitHub repo, embedding provider, vector DB settings.
  • Generates a type-safe configuration file.
  • Creates an environment variables template.

Configure Environment Variables

Copy .env.example to .env.local and add your API keys.

cp .env.example .env.local
.env.local
# GitHub Token (optional for public repos)
GITHUB_TOKEN=ghp_your_github_personal_access_token

# Embedding Provider API Key
GOOGLE_GENERATIVE_AI_API_KEY=your_gemini_api_key
# or
# OPENAI_API_KEY=your_openai_api_key

# Vector Database
QDRANT_URL=[https://your-cluster.qdrant.io](https://your-cluster.qdrant.io)
QDRANT_API_KEY=your_qdrant_api_key

Security: Never commit .env to version control. Add it to .gitignore.

Review Your Configuration

The CLI generated docsy.config.ts. Review and adjust as needed.

docsy.config.ts
import { defineConfig } from '@gaureshart/docsy-core'
 
export default defineConfig({
  source: {
    type: 'github',
    owner: 'facebook', // Your target repo owner
    repo: 'react', // Your target repo name
    branch: 'main',
  },
 
  processing: {
    maxFiles: 100, // Max files to index
    chunkSize: 1000, // Characters per chunk
    chunkOverlap: 200, // Overlap between chunks
  },
 
  embeddings: {
    provider: 'google', // 'google' or 'openai'
    model: 'gemini-embedding-001',
    taskType: 'QUESTION_ANSWERING',
  },
 
  vectorDatabase: {
    provider: 'qdrant',
    collection: 'react-docs',
  },
})

Type safety: defineConfig() provides full autocomplete and type checking in your editor.

Run Ingestion

Process your documentation and store it in the vector database.

npx @gaureshart/docsy-cli ingest

What happens:

  1. Fetches markdown files from your GitHub repo.
  2. Filters and cleans the content.
  3. Chunks documents intelligently (preserves code blocks, headers).
  4. Generates embeddings via Gemini/OpenAI.
  5. Stores vectors in Qdrant.

Time: Typically takes 1-3 minutes for ~100 docs. Progress is shown in your terminal.

Create the API Route

Set up the backend endpoint to handle chat requests.

        • route.ts
app/api/docsy/route.ts
import { createDocsy } from '@gaureshart/docsy-core'
 
export const maxDuration = 30 // Vercel function timeout
 
export async function POST(req: Request) {
  const { messages } = await req.json()
 
  // Extract the latest user query
  const query = messages.at(-1)?.content || messages.at(-1)?.parts?.at(-1)?.text
 
  // Run the RAG pipeline
  const result = await createDocsy({
    pattern: 'naive',
    query,
    messages,
 
    vectorDatabase: {
      provider: 'qdrant',
      collection: 'react-docs', // Must match your config
    },
 
    llmConfig: {
      provider: 'google',
      model: 'gemini-2.5-flash',
    },
 
    embeddings: {
      provider: 'google',
      model: 'gemini-embedding-001',
      taskType: 'QUESTION_ANSWERING',
    },
  })
 
  // Stream the response
  return result
}

Streaming: The response streams token-by-token for a better user experience.

Add the Chat UI

Create a chat component using Vercel AI SDK’s useChat hook. Docsy handles the backend, leaving you free to design the UI using Shadcn or any custom components you prefer.

components/DocsChat.tsx
'use client'
 
import { useChat } from '@ai-sdk/react'
import { DefaultChatTransport } from 'ai'
import { useState } from 'react'
 
export default function Page() {
  const { messages, sendMessage, status } = useChat({
    transport: new DefaultChatTransport({
      api: '/api/docsy',
    }),
  })
  const [input, setInput] = useState('')
 
  return (
    <>
      {messages.map((message) => (
        <div key={message.id}>
          {message.role === 'user' ? 'User: ' : 'AI: '}
          {message.parts.map((part, index) =>
            part.type === 'text' ? <span key={index}>{part.text}</span> : null,
          )}
        </div>
      ))}
 
      <form
        onSubmit={(e) => {
          e.preventDefault()
          if (input.trim()) {
            sendMessage({ text: input })
            setInput('')
          }
        }}
      >
        <input
          value={input}
          onChange={(e) => setInput(e.target.value)}
          disabled={status !== 'ready'}
          placeholder="Say something..."
        />
        <button type="submit" disabled={status !== 'ready'}>
          Submit
        </button>
      </form>
    </>
  )
}

Done! Your AI-powered documentation chatbot is ready. Users can now ask questions and get intelligent answers directly from your repository.

Need help? Open an issue on GitHub  →

Last updated on
Welcome