Add research router

[alexechoi]

Add research router

High-level PR Summary

This PR adds infrastructure for a research router by integrating the Google Gemini AI SDK (@google/genai). It includes a health check endpoint at /api/health, initializes a Gemini AI client in lib/gemini.ts, and updates dependencies to include the necessary Google AI library along with its transitive dependencies.

⏱️ Estimated Review Time: 15-30 minutes

💡 Review Order Suggestion

Order	File Path
1	junction-app/app/api/health/index.ts
2	junction-app/lib/gemini.ts
3	junction-app/package.json
4	junction-app/bun.lock

⚠️ Inconsistent Changes Detected

File Path	Warning
junction-app/app/api/health/index.ts	Adding a generic health check endpoint seems unrelated to the main purpose of adding a research router with AI capabilities. This appears to be general infrastructure rather than research-specific functionality.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Comments	Updated (UTC)
aegis	Ready	Preview	Comment	Nov 15, 2025 7:23pm

[recurseml]

Review by RecurseML

🔍 Review performed on 327c8d4..5cfe36c

Severity	Location	Issue	Delete
	junction-app/lib/gemini.ts:4	Missing required API key parameter

✅ Files analyzed, no issues (3)

• junction-app/app/api/health/index.ts
• junction-app/bun.lock
• junction-app/package.json

[recurseml]

The GoogleGenAI client is initialized with an empty configuration object, missing the required 'apiKey' parameter. According to the official Google GenAI SDK documentation (https://docs.cloud.google.com/vertex-ai/generative-ai/docs/sdks/overview), the correct initialization requires explicitly passing the API key:

const GEMINI_API_KEY = process.env.GEMINI_API_KEY;
const ai = new GoogleGenAI({vertexai: false, apiKey: GEMINI_API_KEY});

The current implementation passes an empty object {} without the required apiKey parameter. While the comment suggests "The client gets the API key from the environment variable GEMINI_API_KEY", the @google/genai library does NOT automatically read environment variables without explicit configuration.

When this client is used to make any API calls (e.g., ai.models.generateContent()), it will fail with an authentication error because no API key is configured. This will cause immediate runtime failures in any code that imports and uses this client.

To fix this bug, the code should be updated to:

import { GoogleGenAI } from "@google/genai";

const GEMINI_API_KEY = process.env.GEMINI_API_KEY;
export const ai = new GoogleGenAI({apiKey: GEMINI_API_KEY});

---

React with 👍 to tell me that this comment was useful, or 👎 if not (and I'll stop posting more comments like this in the future)