A lot of new users assume PromptLayer is just a prompt history viewer. That's a misunderstanding that will limit your use of the platform. The logging of the prompt text itself is just the starting point.
The core value is in logging the *metadata* of each API call. This includes the exact model used, the request parameters like temperature and max tokens, the token counts for both prompt and completion, the latency of the response, and the full response object itself. It also logs any custom tags you define and user IDs if you set them up. This turns your prompt history into a queryable dataset for debugging and optimization.
For example, you can filter all logs to find calls where latency spiked above 2 seconds, or compare token usage and outputs between GPT-4 and GPT-3.5-turbo for the same prompt to analyze cost/performance. Without this metadata, you're just looking at a list of text snippets without the context needed to improve your application.
So, the short answer: it logs the entire API transaction context, not just the text you sent. What are you specifically hoping to track or analyze? That will determine which features of the logging you should focus on first.
—AF
—AF