diff --git a/Administration/config-yaml.md b/Administration/config-yaml.md index 981dd0c96..8dd45eabf 100644 --- a/Administration/config-yaml.md +++ b/Administration/config-yaml.md @@ -126,7 +126,7 @@ protocol: | Setting | Description | Default | Permitted Values | |---------|-------------|---------|-----------------| -| `mistral.enablePrefix` | Enable reply prefilling | `false` | `true`, `false` | +| `mistral.enablePrefix` | Enable reply prefilling. **The prefix will be echoed in the response** | `false` | `true`, `false` | ### Ollama Configuration @@ -136,6 +136,14 @@ protocol: ### Claude Configuration +!!! warning +**IMPORTANT!** + +Use with caution and only when the prompt prefix is static and doesn't change between requests. \{\{random\}\} macro, lorebooks, vectors, summaries, etc. will likely invalidate the cache and you'll just waste money on cache misses. Behavior may be unpredictable and no guarantees can or will be made. + +See: [Prompt Caching](https://docs.anthropic.com/en/docs/build-with-claude/prompt-caching) +!!! + | Setting | Description | Default | Permitted Values | |---------|-------------|---------|-----------------| | `claude.enableSystemPromptCache` | Enable system prompt caching | `false` | `true`, `false` |