Build a RAG-Powered Chatbot with GitHub Docs

Step 1: Create the Flow

1. Go to Lamatic Studio
2. Create a new Flow from Scratch
3. Update the Flow name: GitHub Indexing Flow

Step 2: Configure Trigger Node

1. Click on the Trigger Node and use Webhook as the trigger Node
2. In the following sequence, add:
   • Chunking Node
   • Vectorize Node
   • Code Node
   • Index Node

Step 3: Configure Chunking Node

Select the Chunking Node and configure:

• To add text to a chunk, click the + icon and add: {{triggerNode_1.output.content}}
• Number of Characters: 1000
• Characters that Overlap: 100
• Chunking Method: Recursive Character Text Splitter
• Separators: \n\n \n

Step 4: Configure Vectorize Node

• Texts to vectorize: {{chunkNode_860.output.chunks[:].pageContent}}
• Embedding Model Name: Gemini
• Model: gemini-embedding-exp-03-07

Step 5: Set Up Code Node

1. Change the name to: Prepare Metadata
2. Include the following JavaScript code:

let vectors = {{vectorizeNode_814.output.vectors}};
let metadataProps = [];
let texts = {{chunkNode_860.output.chunks}};
 
for (const idx in vectors) {
    let metadata = {}
    metadata["content"] = texts[idx].pageContent;
    metadata["file_name"] = {{triggerNode_1.output.filename}};
    metadataProps.push(metadata)
}
 
output = {"metadata": metadataProps, "vectors": vectors};

Step 6: Set Up Index Node

Configure the Index Node with:

• Vector DB: Choose your preferred vector database
• Input Vectors: {{codeNode_443.output.vectors}}
• Input Meta Data: {{codeNode_443.output.metadata}}
• Input Primary Key: file_name

Step 7: Deploy and Get Credentials

1. Save the Flow
2. Deploy the Flow and get the Webhook URL from the Setup Guide
3. Acquire the API Key from Settings > API Keys
4. Use the Webhook URL and API Key for GitHub Actions configuration

Step 1: Create GitHub Repository

Initialize a new Git repository and populate it with Markdown files, or fork an existing repository with documentation.

Step 2: Configure Repository Secrets

1. Go to Repo Settings
2. Navigate to Secrets and variables in the Security section (left sidebar)
3. Select Actions
4. Establish two variables within Repository secrets:
   • WEBHOOK_KEY
   • WEBHOOK_URL

Step 3: Create GitHub Action

1. Go to Actions, then select "Create new Flow"
2. Choose the "Simple workflow by GitHub" option
3. Click the Configure button
4. Update the file name: send-to-lamatic.yml

Step 4: Add YAML Configuration

name: Send File Changes
 
on:
  push:
    branches:
      - main  # Adjust to your branch as necessary
    paths: 
      - '**.md'  # Only trigger for changes in .md files
 
jobs:
  send-changes:
    runs-on: ubuntu-latest
    
    steps:
    - name: Send File Changes to Webhook
      uses: Lamatic/[email protected]
      with:
        webhook_url: ${{ secrets.WEBHOOK_URL }}
        webhook_key: ${{ secrets.WEBHOOK_KEY }}
        github_ref: ${{ github.ref }}
        file_type: "md"
        mode: "incremental"  # or "full-refresh"
        verbose: "true"  # or "false"

Step 5: Commit Changes

Commit the changes to activate the workflow.

Step 1: Create Chat Flow

1. Create a Flow from Scratch
2. Rename as Github AI Chatbot

Step 2: Configure Chat Widget

Configure the Chat Widget properties:

• Allow domain settings
• Interface properties (default message, color theme, etc.)

Step 3: Add and Configure RAG Node

1. Add RAG Node
2. Configure the node:
   • Use gemini-embedding-exp-03-07 embedding model (same as before)
   • Use the Gemini Model
   • Click + symbol to include a Search Query, then use chatMessage

Step 4: Update Chat Response Node

Select the modelResponse (Click the + symbol to select the response from the RAG Node)

Step 1: Test the Chat Widget

Click the Test button to test the Node functionality.

Step 2: Deploy Chat Flow

Deploy your flow to make it accessible for chatbot integration. Monitor the logs in the logs section.

Step 3: Integrate Chat Widget

1. Click the Setup button to access the integration code
2. Copy the provided code snippet
3. Paste it into your website