- Clone the repository
- Populate the
.envfile as shown in.env.example - Run
npm install - Run
npm run build - Run
npm link - You're good to go! Use it in any Next.js repository.
- Run
npx dbagent init
→ Initializes the agent with a shallow context of your project - Run
npx dbagent "YOUR_QUERY_HERE"
→ Executes the query to create files, write, and update code accordingly
This agent is heavily prioritized on correctness and accuracy when creating database schemas and backend API routes.
The second priority is integrating correctly into the backend.
Since it performs a shallow scan to save resources, it currently overlooks CSS and focuses on the core backend structure of the project.
A deep scan feature will be introduced later to include frontend and build complete end-to-end functionality.
The core innovation(well atleast to me) here is the decision to store context locally in the user's project root, instead of relying on the model's memory — which can run out.
This allows the agent to re-use project understanding across multiple queries by sending context along each time.
-
gemini-2.5-flash
- Used for scanning and context gathering
- Thinking budget: 1024
- Chosen for speed
-
gemini-2.5-pro
- Used for code generation
- Thinking budget: 20,000
- Chosen for accuracy and correctness
Thank you for your attention to this matter!