Based on Portkey's multimodal support, smart education solutions can be built in four steps:
- Model Configuration: Add models that support vision (e.g. GPT-4V, LLaVA) to Gateway and upload API keys for each model.
- mix-and-match calling: use Python SDK to pass both text and image parameters (e.g., photo of math problem + text prompt "step-by-step solution to this problem")
- Optimization of results: Designing subject-specific prompt templates through the Prompts module ("You are a math teacher, explain in a way that middle schoolers can understand...")
- Deployment delivery: Deploy on-campus servers using the open-source version or achieve geographic coverage through the enterprise version of the cloud service
Realization effects: An online education platform with the help of this program:
- Image question recognition accuracy of 92%
- Increase peak hour response speeds by up to 3x with load balancing
- Utilizing smart caching to make answers to the same topics less costly 65%
The architecture is particularly suitable for homework correction, lab report analysis and other scenarios that require a combination of graphical understanding.
This answer comes from the articlePortkey: a development tool for connecting multiple AI models and managing applicationsThe
May not be reproduced without permission:AI productivity tools " How to build multimodal educational applications with Portkey?
