You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: README.md
+2-2Lines changed: 2 additions & 2 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -32,8 +32,8 @@ This repo is a Bun workspace monorepo with two user-facing integrations and one
32
32
-**Persistent Claude cache controls**: manage Anthropic 1-hour prompt caching from `/claude-cache` with explicit, automatic, or hybrid modes.
33
33
-**Cache keepalive**: use `/claude-cachekeep always` or `/claude-cachekeep HH-HH` to pre-warm hybrid cache anchors for active sessions before the 1-hour TTL expires.
34
34
-**Fast mode toggle**: use `/claude-fast on|off` to request Anthropic fast mode for supported Opus models.
35
-
-**Fable/Mythos reasoning visibility**: request summarized adaptive thinking for Claude Fable 5and Mythos 5 so agents can display reasoning summaries instead of blank signed-thinking blocks.
36
-
-**Fablecontent-filter recovery (OpenCode)**: when Fable ends a session response with Anthropic's `refusal` content-filter reason, transparently retry with Opus 4.8 for 10 successful model responses. After each Opus response, a zero-output Fable prewarm advances the same prompt cache using the OAuth account that served the filtered Fable request. The latest Opus cache boundary is retained so a later refusal can bridge back to Opus even after more than 20 Fable content blocks; the selected OpenCode model remains Fable. The TUI sidebar shows the per-session recovery countdown and return to Fable, while OpenCode Desktop receives ignored `promptAsync` notices for both transitions.
35
+
-**Adaptive reasoning visibility**: request summarized adaptive thinking for Claude Fable 5, Mythos 5, and Opus 5. OpenCode receives native `low`, `medium`, `high`, `xhigh`, and `max` Opus 5 effort variants rather than legacy manual-thinking budgets.
36
+
-**Fable/Opus 5 content-filter recovery (OpenCode)**: when Fable or Opus 5 ends a session response with Anthropic's `refusal` content-filter reason, transparently retry that source-model family with Opus 4.8 for 10 successful model responses. Recovery state and pending cache warms remain independent when a session switches between Fable and Opus 5. After each downgraded response, a zero-output prewarm advances the source model's prompt cache using the OAuth account that served the filtered request. The latest Opus cache boundary is retained so a later refusal can bridge back even after more than 20 content blocks. The TUI sidebar and OpenCode Desktop report both transitions while preserving the selected model.
37
37
-**Live quota visibility**: use `/claude-quota` to see main and fallback quota state, reset times, and refresh errors.
38
38
-**Quota sidebar widget**: register the OpenCode TUI plugin in `tui.json` to render a live sidebar with per-account quota, routing, cache, and health state.
39
39
-**Killswitch**: per-account hard-block thresholds that stop requests before hitting Anthropic's rate limits, with synthetic 429 retry-after when all accounts are exhausted.
Copy file name to clipboardExpand all lines: packages/opencode/README.md
+2-2Lines changed: 2 additions & 2 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -31,8 +31,8 @@ This repo is a Bun workspace monorepo with two user-facing integrations and one
31
31
-**Persistent Claude cache controls**: manage Anthropic 1-hour prompt caching from `/claude-cache` with explicit, automatic, or hybrid modes.
32
32
-**Cache keepalive**: use `/claude-cachekeep always` or `/claude-cachekeep HH-HH` to pre-warm hybrid cache anchors for active sessions before the 1-hour TTL expires.
33
33
-**Fast mode toggle**: use `/claude-fast on|off` to request Anthropic fast mode for supported Opus models.
34
-
-**Fable/Mythos reasoning visibility**: request summarized adaptive thinking for Claude Fable 5and Mythos 5 so agents can display reasoning summaries instead of blank signed-thinking blocks.
35
-
-**Fablecontent-filter recovery**: when Fable ends a session response with Anthropic's `refusal` content-filter reason, transparently retry with Opus 4.8 for 10 successful model responses. After each Opus response, a zero-output Fable prewarm advances the same prompt cache using the OAuth account that served the filtered Fable request. The latest Opus cache boundary is retained so a later refusal can bridge back to Opus even after more than 20 Fable content blocks; the selected OpenCode model remains Fable. The TUI sidebar shows the per-session recovery countdown and return to Fable, while OpenCode Desktop receives ignored `promptAsync` notices for both transitions.
34
+
-**Adaptive reasoning visibility**: request summarized adaptive thinking for Claude Fable 5, Mythos 5, and Opus 5. OpenCode receives native `low`, `medium`, `high`, `xhigh`, and `max` Opus 5 effort variants rather than legacy manual-thinking budgets.
35
+
-**Fable/Opus 5 content-filter recovery**: when Fable or Opus 5 ends a session response with Anthropic's `refusal` content-filter reason, transparently retry that source-model family with Opus 4.8 for 10 successful model responses. Recovery state and pending cache warms remain independent when a session switches between Fable and Opus 5. After each downgraded response, a zero-output prewarm advances the source model's prompt cache using the OAuth account that served the filtered request. The latest Opus cache boundary is retained so a later refusal can bridge back even after more than 20 content blocks. The TUI sidebar and OpenCode Desktop report both transitions while preserving the selected model.
36
36
-**Live quota visibility**: use `/claude-quota` to see main and fallback quota state, reset times, and refresh errors.
37
37
-**Killswitch**: per-account hard-block thresholds that stop requests before hitting Anthropic's rate limits, with synthetic 429 retry-after when all accounts are exhausted.
38
38
-**User-owned Cloudflare relay**: optionally provision your own Worker relay to reduce repeated client upload bytes for large OpenCode or Pi requests.
0 commit comments