tweakcc-fixed 2.8.27 → 2.8.29

package/README.md

@@ -66,9 +66,9 @@ That reach shows up in a few mechanisms the base doesn't have. The `<system-remi
 
 The extra patches cluster around a few themes: **memory** (a dream-mode consolidation pass, leaner memory types), **reasoning** (Opus defaulting to max effort, plus the experimental complexity router), **search** (the experimental fff backend), and a run of **correctness fixes** — an honest rewind-summary header, a "summarize from here" that actually starts at the rewind point, quieter empty system-reminders, and more.
 
-Two of those are worth calling out, and both ship off by default. **fff-first Bash search** routes Claude's grep, find, and rg through [fff](https://github.com/dmtrKovalenko/fff) and a warm-index daemon, so results come back ranked; it serves a query only when the result is provably identical to the real tool and falls back to the embedded ripgrep/ugrep on anything it can't match exactly, so correctness never rides on it. **The complexity router** reads how hard each task is and routes reasoning effort to match — routine work runs low, the hardest runs max — without switching models or churning the prompt cache, and an explicit `/effort` or `CLAUDE_CODE_EFFORT_LEVEL` always wins.
+Two of those are worth calling out, and both ship off by default. **fff-first Bash search** routes Claude's grep, find, and rg through [fff](https://github.com/dmtrKovalenko/fff) and a warm-index daemon, so results come back ranked; it serves a query only when the result is provably identical to the real tool and falls back to the embedded ripgrep/ugrep on anything it can't match exactly, so correctness never rides on it. **The complexity router** chooses reasoning effort for each user message without switching the main model. It supports Haiku routing or opt-in Jev decisions with asynchronous Haiku conversation summaries. See the [effort router setup and behavior](docs/effort-router.md), including credential storage, context sharing, and fallback behavior.
 
-Newest is **Better Claude in Chrome** — a menu item, not a patch, that installs [claude-browser-bridge](https://github.com/skrabe/claude-browser-bridge) and points Claude Code at your real, logged-in browser over the Chrome DevTools Protocol. It sees and _claims_ your existing signed-in tabs — a fuller, self-hosted take on Claude in Chrome, driven through a `/browser` skill and a programmable `run` primitive (script a whole flow — locate, fill, click, wait, read — in one call). Pick it and it fetches the bridge, wires up the MCP server and the skill, walks you through loading the extension once, and disables the built-in Claude in Chrome so you have one browser surface. Pure config plus a fetched repo, no `cli.js` patch; Reinstall and Uninstall live in the same menu.
+Newest is **Better Claude in Chrome** — a menu item, not a patch, that installs [claude-browser-bridge](https://github.com/skrabe/claude-browser-bridge) and points Claude Code at your real, logged-in browser over the Chrome DevTools Protocol. It sees and _claims_ your existing signed-in tabs — a fuller, self-hosted take on Claude in Chrome, driven through a `/browser` skill and a programmable `run` primitive (script a whole flow — locate, fill, click, wait, read — in one call). Pick it and it fetches the bridge, wires up the MCP server and the skill, walks you through loading the extension once, and disables the built-in Claude in Chrome — in the CLI and the Claude desktop app — so you have one browser surface. Pure config plus a fetched repo, no `cli.js` patch; Reinstall and Uninstall live in the same menu.
 
 <details>
 <summary>Every patch the fork adds</summary>
@@ -86,7 +86,7 @@ Each patch is tagged with how it behaves on `--apply`: **`[default on]`** applie
 **Reasoning**
 
 - `max-effort-default` **`[opt-in]`** — Opus defaults to max reasoning effort
-- `complexity-router` **`[opt-in]`** — route reasoning effort by task difficulty _(experimental)_
+- `complexity-router` **`[opt-in]`** — opt-in Opus 5.5 / Fable 5.1 model entries that route reasoning effort by task difficulty _(experimental)_
 
 **Search**