🔍 Use Cases

• Academic Research: Quickly gather information and insights for academic papers, theses, and research projects.
• Market Analysis: Research industry trends, competitors, and market dynamics for business decisions.
• Content Creation: Generate comprehensive research for blog posts, articles, and educational content.
• Product Development: Explore user needs, technical solutions, and industry standards for new products.
• Educational Projects: Help students gather information on complex topics for projects and assignments.

🛠️ Technology Stack

• Cloudflare Workers: Core runtime environment.
• Cloudflare Workflows: Manages reliable research execution.
• Cloudflare D1 Database: Serverless SQL for research data.
• Browser Rendering: Web crawling (free tier available).
• Google Gemini 2.5: Advanced language model for reports.
• Cloudflare AI Gateway (Optional): Proxy for AI requests (caching, management).
• Hono: Lightweight web framework for dashboard and API.
• workers-qb: Query builder for Cloudflare D1.

🏗️ Architecture Overview

graph TD
    B{Cloudflare Worker} -- API Calls --> D[Cloudflare Workflows];
    B -- Fetches Status/Data --> F[Cloudflare D1 Database];
    D -- Triggers --> E[Gemini 2.5 API];
    D -- Stores/Retrieves Research --> F;
    D -- If Crawling Needed --> G[Browser Rendering API];
    E -- Report Chunks/Results --> D;

    subgraph "Cloudflare Ecosystem (Backend)"
        B
        D
        F
        G
    end

    subgraph "External Services"
        E
    end

    style B fill:#ccf,stroke:#333,stroke-width:2px;
    style D fill:#ccf,stroke:#333,stroke-width:2px;
    style E fill:#fca,stroke:#333,stroke-width:2px;
    style F fill:#ccf,stroke:#333,stroke-width:2px;
    style G fill:#ccf,stroke:#333,stroke-width:2px;

Research Workflow Diagram

The following diagram illustrates the detailed workflow of the research agent when processing a request:

graph TD
    A["User Input: Query, Parameters"] --> B["System: Prepare Full Query & Setup"];
    B --> C["Start Iterative Research Cycle"];
    C --> C1["Generate Search Tasks/Queries based on current goal & learnings"];
    C1 --> C2["Gather Information (Web Search & Content Extraction)"];
    C2 --> C3["Process Information: Extract Key Learnings & Identify Follow-up Needs"];
    C3 --> C4["Further Research Iterations Needed? (based on depth/breadth)"];
    C4 -- Yes --> C1;
    C4 -- No --> D["Compile All Learnings & Generate Final Report"];
    D --> E["Store Report & Update Status in Database"];
    E --> F["Research Complete / Report Available"];

🚦 Getting Started

Follow these steps to set up and run workers-research on your Cloudflare account:

Setup Steps

1. Clone the Repository:

   git clone git@github.com:G4brym/workers-research.git
   

2. Install Dependencies:

   npm install
   

3. Cloudflare Login: Authenticate with your Cloudflare account using Wrangler:

   wrangler login
   

4. Create a Cloudflare D1 Database: Create a D1 database in your Cloudflare dashboard or using Wrangler:

   wrangler d1 create workers-research
   

   Important: Replace "replace-me" in your wrangler.toml file with the database_id output from the command above.

   Database migrations are applied automatically.

5. Deploy to Cloudflare Workers: Deploy your worker to Cloudflare:

   wrangler deploy
   

   This will deploy your worker to workers-research.{your-user}.workers.dev (or your custom domain if configured).

6. Set up API Keys:

   • Google AI Studio API Key: Obtain from https://aistudio.google.com.
   • Upload API Keys into workers:

     npx wrangler secret put GOOGLE_API_KEY
     

7. (Optional) Configure Cloudflare AI Gateway:

   • Create an AI Gateway in your Cloudflare account dashboard.
   • Configure the gateway with your preferred settings.
   • Set the required environment variables:

     npx wrangler secret put AI_GATEWAY_NAME
     npx wrangler secret put AI_GATEWAY_ACCOUNT_ID
     

   • If your gateway has authentication enabled, also set the API key:

     npx wrangler secret put AI_GATEWAY_API_KEY
     

8. Access the Dashboard: Open your deployed worker URL (e.g., https://workers-research.{your-user}.workers.dev) in your browser to access the research dashboard.

✍️ Usage

1. Create a New Research:

   • On the dashboard homepage, you'll see a "Create New Research" section.
   • Enter your research query, desired depth, and breadth.
   • Click "Continue with creation".

2. Answer Follow-up Questions:

   • The application will generate follow-up questions to clarify your research intent.
   • Answer these questions to refine the research direction.
   • Click "Create new Research" to start the research workflow.

3. Monitor Research Status:

   • You can track the status of your researches on the dashboard.
   • Researches in progress will be marked as "Running".
   • Completed researches will be marked as "Complete".

4. Read Research Reports:

   • Once a research is complete, a "Read" button will appear next to it.
   • Click "Read" to view the generated research report in a nicely formatted HTML page.

5. Ask Follow-up Questions:

   • On the research details page of a completed report, scroll to the "Ask a follow-up question" panel.
   • Type a question about the report (e.g. "What are the main risks identified?" or "Which companies were mentioned most?").
   • Answers are grounded in the report content and saved for future reference.

6. Re-run Researches:

   • If you want to re-run a previous research, click the "🔄" button next to the research entry. This will create a new research based on the same query and parameters.

📊 Performance & Limitations

• Research Time: Most researches complete within 5-10 minutes, depending on complexity.
• Research Depth: Configure depth from 1-5, with higher values producing more comprehensive reports.
• API Limits: Be aware of Cloudflare and Google Gemini API limits when running multiple researches. Browser Rendering on the Workers Free plan is limited to 10 minutes of usage per day, sufficient for most research tasks.

💬 Community & Support

• GitHub Issues: Report bugs or suggest features
• Discussions: Ask questions and share ideas

🖼️ Screenshots

Home page

New Research

Follow up questions

Reading a report

📜 License

This project is licensed under the MIT License - see the LICENSE file for details.