This AI-powered application transforms raw Azure DevOps work item data into value-driven, stakeholder-focused release notes. It analyzes technical details and value propositions to craft communications tailored for executives, sales, marketing, and support teams.

• Connects directly to Azure DevOps using a Personal Access Token (PAT).
• Fetches work items based on Release Version and an optional Iteration Path.
• Uses a Large Language Model (LLM) to generate tailored release notes for different audiences.
• Provides a review screen to approve, reject, or edit the AI-generated content.
• Compiles all approved notes into a final Markdown document, ready for distribution.

• React
• TypeScript
• Tailwind CSS
• Google Gemini API

The application is designed to be model-agnostic, allowing you to switch from the default Google Gemini provider to another LLM service like AWS Bedrock. This is managed through a simple abstraction layer.

• services/geminiService.ts: This file contains the specific implementation for calling the Google Gemini API.
• services/llmService.ts: This is the central abstraction layer or "router". It determines which underlying LLM service to use based on a configuration constant. Your application components should only call functions from this file, not from the specific service files.

Follow these steps to integrate a new LLM provider:

Step 1: Create a New Service File

Create a new file in the services directory for your provider. For example, services/bedrockService.ts.

Step 2: Implement the Provider Logic

Inside your new file (services/bedrockService.ts), you must implement and export a function named generateSingleItemNote. This function must have the same signature as the one in geminiService.ts:

// Example: services/bedrockService.ts

import { WorkItem, StakeholderNotes } from "../types";

// You will need to install and import the AWS SDK
// import { BedrockClient, ... } from "@aws-sdk/client-bedrock"; 

export async function generateSingleItemNote(
    item: WorkItem, 
    model: string // e.g., 'anthropic.claude-v2'
): Promise<StakeholderNotes> {
    
    // 1. Format the work item data into a prompt for your chosen model.
    const prompt = `...`; // Construct your prompt here

    // 2. Initialize and configure your AWS Bedrock client.
    // const client = new BedrockClient(...);

    // 3. Make the API call to AWS Bedrock.
    // const response = await client.send(...);

    // 4. Parse the response from Bedrock. It must be parsed into
    //    a JSON object that matches the `StakeholderNotes` interface.
    const parsedNotes: StakeholderNotes = JSON.parse(response.body);

    // 5. Return the parsed notes.
    return parsedNotes;
}

Step 3: Register the New Provider in the Abstraction Layer

Open services/llmService.ts.

1. Import your new function at the top of the file. It's best to rename it on import to avoid naming conflicts.

   // services/llmService.ts
   import { generateSingleItemNote as generateWithGemini } from "./geminiService";
   import { generateSingleItemNote as generateWithBedrock } from "./bedrockService"; // Add this line

2. Update the case 'BEDROCK' inside the generateSingleItemNote function to call your new implementation.

   // services/llmService.ts
   
   // ... inside the generateSingleItemNote function
   switch (LLM_PROVIDER) {
       case 'GEMINI':
           return generateWithGemini(item, model);
       case 'BEDROCK':
           // Replace the placeholder with your new function
           return generateWithBedrock(item, model); 
       // ...
   }

Step 4: Switch the Active Provider

Finally, still in services/llmService.ts, change the LLM_PROVIDER constant to activate your new service.

// services/llmService.ts

const LLM_PROVIDER: 'GEMINI' | 'BEDROCK' = 'BEDROCK'; // Change 'GEMINI' to 'BEDROCK'

That's it. The application will now route all AI generation requests through your new AWS Bedrock service without any changes needed in the UI components.