AI integration

​

Overview

• Ticket generation - Analyze project descriptions and generate actionable tickets
• Project chat - Interactive assistant that asks clarifying questions
• Smart recommendations - Priority, effort, and dependency suggestions

​

Getting a Gemini API key

1. Visit Google AI Studio
2. Sign in with your Google account
3. Click “Get API key” or “Create API key”
4. Copy your API key

​

Configuration

GOOGLE_GENERATIVE_AI_API_KEY="your-api-key-here"

​

AI model

import { google } from "@ai-sdk/google";

const model = google("gemini-2.5-flash");

• Fast response times
• High-quality text generation
• Tool calling and function execution
• Structured output support

​

AI features

​

Ticket generation

type Ticket = {
  title: string;
  type: "Story" | "Task" | "Bug" | "Epic";
  priority: "P0" | "P1" | "P2" | "P3";
  description: string;
  acceptanceCriteria: string[];
  estimatedEffort: string;
  dependencies: string[];
  labels: string[];
};

​

Project chat assistant

1. Clarifying questions - Asks 2-4 targeted questions about:
   • Goals and success criteria
   • Stakeholders and audience
   • Key deliverables and milestones
   • Timeline and constraints
   • Dependencies
2. Context gathering - Remembers conversation history and project details
3. Ticket refinement - Add, update, or remove tickets based on feedback

​

AI tools

• generateTickets - Create new tickets for the project
• listTickets - Retrieve existing tickets
• updateTickets - Modify ticket properties
• removeTickets - Delete tickets by ID
• setClarifyingQuestions - Present structured questions to users

​

Implementation details

​

Vercel AI SDK

import { streamText } from "ai";
import { google } from "@ai-sdk/google";

const result = streamText({
  model: google("gemini-2.5-flash"),
  system: systemPrompt,
  messages: conversationHistory,
  tools: aiTools,
});

• Streaming responses - Real-time token streaming
• Tool calling - Execute functions from AI responses
• Message history - Maintain conversation context
• UI components - React hooks for chat interfaces

​

API endpoint

export async function POST(req: Request) {
  if (!process.env.GOOGLE_GENERATIVE_AI_API_KEY) {
    return new Response(
      JSON.stringify({
        error: "Google Generative AI API key is missing."
      }),
      { status: 503 }
    );
  }

  const { messages, project } = await req.json();
  
  const result = streamText({
    model: google("gemini-2.5-flash"),
    system: systemPrompt,
    messages: messages,
    tools: createTools(project?.id),
  });

  return createUIMessageStreamResponse({ stream: result });
}

​

System prompts

• Analyzes new project ideas
• Asks clarifying questions
• Generates initial tickets

• References existing project context
• Avoids asking user to re-describe the project
• Focuses on refinement and iteration

​

Usage limits and pricing

• Free tier - 60 requests per minute
• Rate limits - May vary by region and model
• Pricing - Check Google AI pricing for current rates

​

Error handling

​

Missing API key

{
  "error": "Google Generative AI API key is missing. Set GOOGLE_GENERATIVE_AI_API_KEY in your .env file.",
  "status": 503
}

​

API failures

• Error messages are logged to the console
• Users see a friendly error message
• Chat history is preserved

​

Testing AI features

​

Development workflow

1. Set your API key in .env
2. Start the development server: pnpm dev
3. Create a new project or open an existing one
4. Test the chat assistant:
   • Ask questions
   • Request ticket generation
   • Try updating and removing tickets

​

Example prompts

I want to launch a new onboarding flow for our SaaS product

The goal is to reduce time-to-value. Target audience is small business owners. Timeline is 6 weeks.

Add a ticket for user research interviews

Bump the "design mockups" ticket to P0

Remove the analytics dashboard ticket

​

Production considerations

​

API key security

• Store API keys in environment variables
• Use different keys for staging and production
• Rotate keys periodically
• Never log or expose keys in client-side code

​

Rate limiting

• Limit requests per user
• Add request queuing
• Implement exponential backoff for retries

​

Monitoring

• Response times
• Error rates
• Token usage
• User satisfaction

​

Future AI features

• Insight extraction - Analyze customer feedback and interviews
• Evidence-based recommendations - Suggest features backed by data
• Semantic search - Find relevant information using embeddings
• Multi-source synthesis - Combine insights from multiple data sources

​

Troubleshooting

​

API key invalid

1. Verify your API key is correct
2. Check the key hasn’t been revoked
3. Ensure you’re using a valid Google account

​

Slow responses

1. Check your internet connection
2. Verify you’re not hitting rate limits
3. Consider using a different model or parameters

​

Unexpected outputs

1. Review your system prompts
2. Check conversation history for context issues
3. Add more specific instructions or constraints

​

Next steps

1. Start the development server
2. Create your first project and test ticket generation
3. Explore the project chat assistant
4. Review generated tickets on the Kanban board