tkmind/tkmind_go
1
0
Fork 0
Code Issues Pull Requests Actions 4 Packages Projects Releases Wiki Activity

feat: configurable Anthropic prompt-cache TTL (5m/1h) (#11576)

Browse Source 
Signed-off-by: Michael Neale <michael.neale@gmail.com>
Co-authored-by: Galadriel <galadriel@buzz.agent>
This commit is contained in:
Michael Neale
2026-08-31 00:43:58 +00:00
committed by GitHub
parent 815bd0b69a
commit fb15d4eade
9 changed files with 425 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files
documentation/docs/guides/config-files.md
+1
View File
@@ -48,6 +48,7 @@ The following settings can be configured at the root level of your config.yaml f
|---------|---------|---------|---------|-----------|
| `GOOSE_TEMPERATURE` | Model response randomness | Float between 0.0 and 1.0 | Model-specific | No |
| `GOOSE_MAX_TOKENS` | Maximum number of tokens for each model response (truncates longer responses) | Positive integer | Model-specific | No |
| `GOOSE_CACHE_TTL` | Anthropic prompt-cache TTL; `1h` keeps the cached prefix alive across idle gaps at a higher cache-write rate. Headless runs always use `5m` | "5m", "1h" | "5m" | No |
| `GOOSE_MODE` | [Tool execution behavior](/docs/guides/managing-tools/goose-permissions) | "auto", "approve", "chat", "smart_approve" | "auto" | No |
| `GOOSE_MAX_TURNS` | [Maximum number of turns](/docs/guides/sessions/smart-context-management#maximum-turns) allowed without user input | Integer (e.g., 10, 50, 100) | 1000 | No |
| `GOOSE_PLANNER_PROVIDER` | Provider for [planning mode](/docs/guides/context-engineering/creating-plans) | Same as `GOOSE_PROVIDER` options | Falls back to `GOOSE_PROVIDER` | No |
documentation/docs/guides/environment-variables.md
+1
View File
@@ -21,6 +21,7 @@ These are the minimum required variables to get started with goose.
| `GOOSE_FAST_MODEL` | Overrides the provider's default fast model used for auxiliary calls (tool-selection, classification, session titles) | Model name (e.g., "gpt-4o-mini", "google/gemini-2.5-flash") | Provider-specific default |
| `GOOSE_TEMPERATURE` | Sets the [temperature](https://medium.com/@kelseyywang/a-comprehensive-guide-to-llm-temperature-%EF%B8%8F-363a40bbc91f) for model responses | Float between 0.0 and 1.0 | Model-specific default |
| `GOOSE_MAX_TOKENS` | Sets the maximum number of tokens for each model response (truncates longer responses) | Positive integer (e.g., 4096, 8192) | Model-specific default |
| `GOOSE_CACHE_TTL` | Sets the Anthropic prompt-cache TTL. `1h` keeps the cached prefix alive across idle gaps (e.g. stepping away mid-session) but bills cache writes at 2x input instead of 1.25x, so it only pays off for sessions that actually idle. Headless runs (`goose run`, subagents, scheduled recipes) always use `5m` | `5m`, `1h` | `5m` |
**Examples**