mirror of
https://github.com/ollama/ollama.git
synced 2026-10-21 02:51:35 +00:00
6e65d95ef5
Added cached prompt token counts to Ollama responses and compatibility usage fields. This carries local `llama-server` `cache_n` and MLX cache hits through `/api/generate`, `/api/chat`, OpenAI-compatible endpoints, and Anthropic-compatible `/v1/messages`. Cloud responses are passed through as-is, so cache counts will show up there once Cloud starts returning them. |
||
|---|---|---|
| .. | ||
| agent | ||
| cmd | ||
| create | ||
| imagegen | ||
| internal/mlxthread | ||
| mlxrunner | ||
| models | ||
| safetensors | ||
| server | ||
| tokenizer | ||
| tools | ||
| transfer | ||