I. Purpose
To support academic research at Academia Sinica, the Department of Information Technology Services provides a locally deployed Large Language Model (LLM) API service. All research data transmission, processing, and generated outputs are handled exclusively on internal servers within Academia Sinica, reducing the risk of data leakage and unauthorized disclosure of sensitive information.
II. Eligibility Requirements
Must be a researcher at Academia Sinica (Users who are at the level of Assistant Research Fellow or above, or Assistant Research Specialist/Engineer/Scientist or above)
III. Application and Modification
Users must first obtain an API Key for the AI API service before connecting to backend model services.
- API Key Application :
Apply online at the AI API application website
Alternatively, users may sign in to the Academia Sinica Portal and navigate to: All > Personal Services > AI Suite > AI API -
Multiple API Keys :
To facilitate separate management of different research projects or programs, users may apply for multiple API Keys as needed. Please keep your API Keys secure and do not share them with others.
-
Validity and Renewal :
Each API Key is valid for up to one year. Users must apply for renewal before the expiration date. API Keys that expire without renewal will become permanently invalid.
IV. Service Information and Specifications
- Service Endpoint and Technical Documentation
For the service endpoint (Base URL), the latest supported models, and code integration examples, please refer to the "Service Guide" available on the AI API application website.
- Model Updates
The Department of Information Technology Services reserves the right to modify or retire supported models based on technological developments or resource availability. Users may refer to the above webpage for the latest model information.
V. Usage Limits and Important Notes
To ensure fair access to the service for all researchers, the following usage limits and requirements apply.
- Usage Limits :
- Request Rate Limit: Each API Key may submit up to 60 API requests per minute (60 Requests Per Minute, RPM).
- Usage Quota: The daily usage limit is 2,000,000 Tokens.
- Concurrent Connections: Each API Key is limited to three concurrent connections.
- Important Notes
- Users must comply with the Academia Sinica Information Security Regulations.
- This service is intended solely for academic research conducted within Academia Sinica. API Keys must not be shared with others, connected to any external systems, published on public platforms, or used with unauthorized third-party applications.
VI. Technical Support and Troubleshooting
If you encounter any issues while using this service, please contact the IT Service Desk through the online support system or call +886-2-2789-8855 for assistance.
VII. Data Privacy and Security Commitment
This service is a fully self-hosted, on-premises system operated entirely within the Academia. All data strictly remains within the internal network with zero retention. System logs capture only essential operational information, and no submitted content is used for any model training or fine-tuning.
