I propose that OpenAI enables developers to attach PDF files (or similar documents) to a model’s knowledge base once at setup. This would allow the model to refer to those documents continuously without requiring developers to resend portions of the documents in each API call. OpenAI could charge a base fee based on the size and number of attached PDFs, benefiting both developers and OpenAI.
Key Features:
- Persistent Knowledge Base: Allow models to store and reference PDFs or documents attached during setup. The model should use these documents for all future queries.
- Base Fee Model: Charge developers a base fee depending on the number and size of the PDFs attached. This creates a scalable revenue stream for OpenAI while offering developers greater efficiency.
- Increased Token Usage: By allowing this feature, developers can build applications more efficiently, which would lead to increased usage of tokens for actual queries, benefiting OpenAI.
Benefits:
- Developers can build applications faster and more efficiently without dealing with token limitations for large documents.
- OpenAI benefits from a new base fee revenue stream and increased API usage.
- Users will experience improved performance and reliability when working with large knowledge bases.
Conclusion:
This approach is a win-win for both OpenAI and developers, as it speeds up application development while driving more usage of OpenAI’s API services.