Voxta Diagnostics
Every LLM call Voxta made — the full prompt, the raw result, tools and timings — plus average timings per step.
Diagnostics shows what Voxta actually sent and got back. Open it from Advanced → Diagnostics; from inside a chat it opens over the chat instead of leaving it.
Refresh, Clear and Download Report (Markdown; the selected call only, when one is selected) sit at the top.
Inferences
Every LLM call recorded, newest first. Filter by kind (TextGen, action inference, summarization…) or by text in the prompt. Open one to see:
- Prompt — the messages as sent, with templates resolved, contexts injected and history included.
- Thinking and Raw Result — what the model returned.
- Flags, Tools Offered, Tool Calls and Functions — what the model could do and what it did.
- Information — success, chat, session and message ids, the action, the service, start time and duration.
- Parameters and Response Fields — the generation settings sent, and what the service reported back.
Use it when a character acts oddly, a context doesn't seem to apply, or you want to see how a template edit renders.
Performance
Average time, counter and count for each measured step. Use it to find which service slows a chat down.