This example demonstrates how to build an intelligent research assistant using AgentKit, arXiv dataset, and the DeepSeek R1 model. The assistant helps users find and analyze relevant academic papers based on their research queries.
Requirements To successfully follow this tutorial, you will need the following:
- Git installed
- Node 22+
pnpm- An Inngest account (for deployment)
- A Koyeb account (for the DeepSeek-R1 QWEN 32b model)
Use the link below to deploy the DeepSeek-R1 QWEN 32b model to Koyeb in one-click:
From there, you can choose to either run the application locally or deploy it to Koyeb.
Deploy the application to Koyeb using the following button:
Then, once your application is deployed, login into your Inngest dashboard and configure a new sync using the following url:
https://<YOUR_DOMAIN_PREFIX>.koyeb.app/api/inngest
Once the sync successful, navigate to the Functions tabs and invoke the arxiv-research-assistant function with this payload:
{
"data": {
"input": "What are the latest advancements in deep learning optimization techniques?"
}
}You should be redirected to the Runs view and see your workflow run.
-
Install dependencies:
pnpm install
-
Configure the deployed model URL in the
.envfile:
DEEPSEEK_API_KEY=test
DEEPSEEK_BASE_URL=https://<YOUR_DOMAIN_PREFIX>.koyeb.app/v1-
Start the development server:
pnpm dev
-
Start the Inngest Dev Server:
npx inngest-cli@latest devOpen the browser and navigate to http://127.0.0.1:8288/ to access the Inngest Dev Server.
Navigate to the Functions tabs and invoke the arxiv-research-assistant function with this payload:
{
"data": {
"input": "What are the latest advancements in deep learning optimization techniques?"
}
}You should be redirected to the Runs view and see your workflow run.