@andares/pi-toolkits 0.11.1 → 0.11.3

package/README.md

@@ -15,7 +15,7 @@
 
 In a nutshell: seven ready-to-use capabilities — **read-only consult mode** `ask`, **session-scoped boundary permission** `ask-perm`, **model favorites** `favorites`, **prompt stash** `stash`, **third-party code review** `third-review`, **dedicated compaction model** `compact-model` and **session startup automation** `autostart` — with more to come.
 
-**Requirements**: pi `>=0.99.0`. Every published pi release from 0.99.0 through 1.4.0 has been verified with the full gate (install → `tsc --noEmit` → full test suite → keybinding audit → hash compare of the host files this package patches), each in a sandbox with the whole `@earendil-works/*` family pinned to the version under test. Newer releases are verified as they ship and the floor moves only when the measurement says so; `pnpm check:compat` reproduces the result. Note: pi ≥ 0.99 hosts require host-provided packages to be declared in `peerDependencies` with a `"*"` range — `@earendil-works/pi-tui` and `typebox` (the `scratchpad` tool builds a TypeBox schema) are declared that way; an installed copy in `dependencies` would make pi warn.
+**Requirements**: pi `>=0.99.0`. Every published pi release from 0.99.0 through 1.5.0 has been verified with the full gate (install → `tsc --noEmit` → full test suite → keybinding audit → hash compare of the host files this package patches), each in a sandbox with the whole `@earendil-works/*` family pinned to the version under test. Newer releases are verified as they ship and the floor moves only when the measurement says so; `pnpm check:compat` reproduces the result. Note: pi ≥ 0.99 hosts require host-provided packages to be declared in `peerDependencies` with a `"*"` range — `@earendil-works/pi-tui` and `typebox` (the `scratchpad` tool builds a TypeBox schema) are declared that way; an installed copy in `dependencies` would make pi warn.
 
 ---
 
@@ -287,7 +287,7 @@ pi's `/compact` and auto-compact summarize the context **with the current sessio
 | auto-compact threshold reached | summarized by the compaction model |
 | context overflow recovery | summarized by the compaction model, retry behavior unchanged |
 
-**Parts that are identical to pi's built-in behavior after takeover** (the built-in implementation is mirrored verbatim — copied from 0.87.1, byte-identical across 1.1.0 → 1.4.0 apart from the two 1.4.0 fixes adopted here: summaries request the model's output limit instead of a fixed fraction of `reserveTokens`, and `/compact <instructions>` also reaches the split-turn prefix — and only the executing model changes; the conversation serializer itself is imported from the host, so it follows whatever the running host does, e.g. 1.2.0 stops feeding assistant thinking back into the summarization request):
+**Parts that are identical to pi's built-in behavior after takeover** (the built-in implementation is mirrored verbatim — copied from 0.87.1, byte-identical across 1.2.0 → 1.5.0 apart from the two 1.4.0 fixes adopted here: summaries request the model's output limit instead of a fixed fraction of `reserveTokens`, and `/compact <instructions>` also reaches the split-turn prefix, plus the 1.5.0 `cost.totalFromProvider` merge rule in the usage accounting — and only the executing model changes; the conversation serializer itself is imported from the host, so it follows whatever the running host does, e.g. 1.2.0 stops feeding assistant thinking back into the summarization request):
 
 - **Summary format** — the structured Goal / Progress / Key Decisions checkpoint; when a summary already exists, the iterative-update prompt is used so information carries over instead of being lost
 - **Split behavior** — the `keepRecentTokens` retention policy, and `firstKeptEntryId` / `tokensBefore` passed back unchanged
@@ -368,7 +368,7 @@ An **identical command line** declared by both sources runs once: a duplicate is
 
 **Key constraint: extension commands only**. Commands registered via `registerCommand` such as `/add-dir` (pi-add-dir) and `/cd` (pi-cd) work; built-in commands (`/compact`, `/reload`, `/resume`, …) are dispatched in the TUI layer and cannot be triggered from the extension layer; skill commands (`/skill:name`) are an expansion mechanism rather than an execution mechanism. The target extension you depend on (e.g. pi-add-dir) must be installed first.
 
-> **Implementation notes**: execution uses `sendUserMessage(cmd, { expandPromptTemplates: true })` — that option routes `/cmd args` into command dispatch instead of the model prompt path (the wrapper defaults to `false`; `sendUserMessage` and `_tryExecuteExtensionCommand` are byte-identical across 1.1.0–1.4.0, see the host-compatibility discipline in AGENTS.md). Pre-validation uses `pi.getCommands()` (accepting only `source: "extension"`). Depends on `smol-toml` (zero-dependency, ~10KB, TOML 1.0).
+> **Implementation notes**: execution uses `sendUserMessage(cmd, { expandPromptTemplates: true })` — that option routes `/cmd args` into command dispatch instead of the model prompt path (the wrapper defaults to `false`; `sendUserMessage` and `_tryExecuteExtensionCommand` are byte-identical across 1.2.0–1.5.0, see the host-compatibility discipline in AGENTS.md). Pre-validation uses `pi.getCommands()` (accepting only `source: "extension"`). Depends on `smol-toml` (zero-dependency, ~10KB, TOML 1.0).
 
 ---
 

package/README.zh-CN.md

@@ -15,7 +15,7 @@
 
 一句话介绍:目前包含七个开箱即用的功能——**只读咨询模式** `ask`、**会话级边界权限** `ask-perm`、**模型收藏** `favorites`、**提示词暂存** `stash`、**第三方代码评审** `third-review`、**独立压缩模型** `compact-model`、**会话启动自动化** `autostart`,后续持续追加。
 
-**版本要求**:pi `>=0.99.0`。0.99.0 到 1.4.0 的每一个已发布版本都用完整门禁实测过(install → `tsc --noEmit` → 全量测试 → 键位审计 → 对 patch 目标宿主文件做哈希比对),每个版本都在沙箱中把整个 `@earendil-works/*` 家族钉在同一版本。更新的版本随发布逐个验证,下限只在实测结果允许时移动;`pnpm check:compat` 可复现该结论。注:pi ≥ 0.99 的宿主要求「宿主提供的包」一律声明为 peerDependencies `"*"` —— `@earendil-works/pi-tui` 与 `typebox`(ask 的 `scratchpad` 用它构造参数 schema)都按此声明;若放进 `dependencies`,pi 会告警。
+**版本要求**:pi `>=0.99.0`。0.99.0 到 1.5.0 的每一个已发布版本都用完整门禁实测过(install → `tsc --noEmit` → 全量测试 → 键位审计 → 对 patch 目标宿主文件做哈希比对),每个版本都在沙箱中把整个 `@earendil-works/*` 家族钉在同一版本。更新的版本随发布逐个验证,下限只在实测结果允许时移动;`pnpm check:compat` 可复现该结论。注:pi ≥ 0.99 的宿主要求「宿主提供的包」一律声明为 peerDependencies `"*"` —— `@earendil-works/pi-tui` 与 `typebox`(ask 的 `scratchpad` 用它构造参数 schema)都按此声明;若放进 `dependencies`,pi 会告警。
 
 ---
 
@@ -287,7 +287,7 @@ pi 的 `/compact` 与 auto-compact 默认**用当前会话模型**做上下文
 | auto-compact 阈值触发 | 压缩模型总结 |
 | context overflow 恢复 | 压缩模型总结,重试行为不变 |
 
-**接管后与 pi 内建行为完全一致的部分**(逐字同步 pi 内建实现,复制自 0.87.1、在 1.1.0 → 1.4.0 内字节级不变 —— 除了此处采纳的两项 1.4.0 修正:摘要请求使用模型自身的输出上限而非 `reserveTokens` 的固定比例,以及 `/compact <instructions>` 也会进入 split-turn 前缀;只换执行模型;对话序列化函数(`serializeConversation`)是直接从宿主导入的,因此跟随宿主行为 —— 例如 1.2.0 起不再把 assistant thinking 回填进摘要请求):
+**接管后与 pi 内建行为完全一致的部分**(逐字同步 pi 内建实现,复制自 0.87.1、在 1.2.0 → 1.5.0 内字节级不变 —— 除了此处采纳的两项 1.4.0 修正:摘要请求使用模型自身的输出上限而非 `reserveTokens` 的固定比例,以及 `/compact <instructions>` 也会进入 split-turn 前缀,外加 1.5.0 的 `cost.totalFromProvider` 合并规则(用量统计);只换执行模型;对话序列化函数(`serializeConversation`)是直接从宿主导入的,因此跟随宿主行为 —— 例如 1.2.0 起不再把 assistant thinking 回填进摘要请求):
 
 - **摘要格式** —— Goal / Progress / Key Decisions 结构化 checkpoint;已有摘要时用迭代更新版 prompt,信息延续不丢
 - **切分行为** —— 保留最近 `keepRecentTokens` 的策略、`firstKeptEntryId` / `tokensBefore` 原样回传,不动
@@ -368,7 +368,7 @@ run = [
 
 **关键约束:只支持扩展命令**。`/add-dir`(pi-add-dir)、`/cd`(pi-cd)等 `registerCommand` 注册的命令可以;内置命令(`/compact`、`/reload`、`/resume`…)在 TUI 层分发,扩展层触发不了;skill 命令(`/skill:name`)是展开机制而非执行机制。依赖的目标扩展(如 pi-add-dir)需要先安装。
 
-> **实现说明**:执行用 `sendUserMessage(cmd, { expandPromptTemplates: true })` —— 该选项把 `/cmd args` 路由到命令分发而不是模型 prompt 路径(包装层默认 `false`,`sendUserMessage` 与 `_tryExecuteExtensionCommand` 在 1.1.0–1.4.0 全区间字节级一致,见 AGENTS.md 宿主兼容纪律)。预校验用 `pi.getCommands()`(只认 `source: "extension"`)。依赖 `smol-toml`(零依赖、~10KB、TOML 1.0)。
+> **实现说明**:执行用 `sendUserMessage(cmd, { expandPromptTemplates: true })` —— 该选项把 `/cmd args` 路由到命令分发而不是模型 prompt 路径(包装层默认 `false`,`sendUserMessage` 与 `_tryExecuteExtensionCommand` 在 1.2.0–1.5.0 全区间字节级一致,见 AGENTS.md 宿主兼容纪律)。预校验用 `pi.getCommands()`(只认 `source: "extension"`)。依赖 `smol-toml`(零依赖、~10KB、TOML 1.0)。
 
 ---
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@andares/pi-toolkits",
-  "version": "0.11.1",
+  "version": "0.11.3",
   "description": "Integrated toolkit extension for pi — ask mode (read-only Q&A), session-scoped permission checks, model favorites, prompt stash, third-party code review, session autostart, a dedicated compaction model, and an agent memory framework (/ask-init)",
   "type": "module",
   "license": "MIT",
@@ -47,9 +47,9 @@
     "smol-toml": "^1.9.0"
   },
   "devDependencies": {
-    "@earendil-works/pi-ai": "^1.4.0",
-    "@earendil-works/pi-coding-agent": "^1.4.0",
-    "@earendil-works/pi-tui": "^1.4.0",
+    "@earendil-works/pi-ai": "^1.5.0",
+    "@earendil-works/pi-coding-agent": "^1.5.0",
+    "@earendil-works/pi-tui": "^1.5.0",
     "@types/node": "^26.6.2",
     "typebox": "^1.3.27",
     "typescript": "^5.6.0",

package/src/features/ask/reasons.ts

@@ -2,7 +2,7 @@
  * ask feature — actionable rejection messages.
  *
  * The `tool_call` hook's `reason` is delivered to the model as the tool result
- * (verified against pi 1.4.0: `extensions/runner.js` returns the `{block,
+ * (verified against pi 1.5.0: `extensions/runner.js` returns the `{block,
  * reason}` result unchanged and pi-agent-core's `agent-loop.js` turns it into
  * `createErrorToolResult(reason)`; both paths are byte-identical to 1.2.0), so
  * it is the best place to name the nearest allowed alternative and stop the model

package/src/features/autostart/autostart.ts

@@ -48,7 +48,7 @@ export const RUNNABLE_REASONS: ReadonlySet<string> = new Set([
  * `-r` (`--resume`), `--session <path>` and `--fork <path>` all open an
  * existing session *before* the first runtime exists, and that runtime is
  * created without an explicit session_start event, so extensions see pi's
- * default reason — `startup` — not `resume` (verified on pi 1.1.0–1.4.0:
+ * default reason — `startup` — not `resume` (verified on pi 1.2.0–1.5.0:
  * `main.js` builds the initial session manager from those flags and calls
  * `createAgentSessionRuntime` with no `sessionStartEvent`;
  * `agent-session.js` defaults it to `{ reason: "startup" }`). `resume` is only

package/src/features/autostart/index.ts

@@ -27,7 +27,7 @@
  * that flag is what routes "/cmd args" through _tryExecuteExtensionCommand
  * instead of the model prompt path. The wrapper defaults it to false in this
  * package (each byte-identical across every published version of the supported
- * window, 1.1.0 → 1.4.0; see AGENTS.md host-compat discipline for the version
+ * window, 1.2.0 → 1.5.0; see AGENTS.md host-compat discipline for the version
  * matrix).
  *
  * Only extension commands can run this way. Built-in commands (/compact,

package/src/features/compact/compact.test.ts

@@ -525,6 +525,61 @@ describe("session_before_compact handler", () => {
 		expect(compaction.usage?.input).toBe(107);
 	});
 
+	it("keeps the cost provider-reported only when both split-turn parts report it", async () => {
+		// 1.5.0's combineUsage rule end-to-end: the merged compaction usage, not
+		// just the helper. A part that reports no provider cost is enough to make
+		// the sum "not provider-reported"; only both parts reporting it flips it
+		// to true (asserted below on a second run). The flag is cost provenance,
+		// not the token sum.
+		getCompactStore().setModel("google/gemini-2.5-flash");
+		const { handler } = registeredHandler();
+		const withFlag = (flag: boolean | undefined) => {
+			const usage = fakeUsage();
+			if (flag !== undefined) usage.cost.totalFromProvider = flag;
+			return usage;
+		};
+		const { ctx } = fakeCtx({
+			complete: vi.fn(async (_model, context) => {
+				const text = (context.messages[0]?.content as Array<{ text?: string }>)[0]
+					?.text;
+				return text?.includes("# Instructions")
+					? fakeResponse("TURN PREFIX", { usage: withFlag(true) })
+					: fakeResponse("HISTORY", { usage: withFlag(undefined) });
+			}),
+		});
+		const event = fakeEvent({
+			turnPrefixMessages: [userMessage("prefix msg")],
+			isSplitTurn: true,
+		});
+		const result = (await handler(
+			event as never,
+			ctx,
+		)) as SessionBeforeCompactResult | undefined;
+		// Read through the structural view, not the host's `CompleteUsage`:
+		// `cost.totalFromProvider` only exists on pi ≥ 1.5.0's types, so reading
+		// it directly from the host type breaks typecheck on the 1.2–1.4 window
+		// even though the runtime field we write is the same.
+		const mergedUsage = result!.compaction!.usage as unknown as UsageLike;
+		expect(mergedUsage.cost.totalFromProvider).toBe(false);
+
+		// Same split turn, but both summaries report the cost themselves → true.
+		const both = fakeCtx({
+			complete: vi.fn(async (_model, context) => {
+				const text = (context.messages[0]?.content as Array<{ text?: string }>)[0]
+					?.text;
+				return text?.includes("# Instructions")
+					? fakeResponse("TURN PREFIX", { usage: withFlag(true) })
+					: fakeResponse("HISTORY", { usage: withFlag(true) });
+			}),
+		});
+		const bothResult = (await handler(
+			event as never,
+			both.ctx,
+		)) as SessionBeforeCompactResult | undefined;
+		const bothUsage = bothResult!.compaction!.usage as unknown as UsageLike;
+		expect(bothUsage.cost.totalFromProvider).toBe(true);
+	});
+
 	it("keeps the previous summary as history text on split turns with no messages", async () => {
 		getCompactStore().setModel("google/gemini-2.5-flash");
 		const { handler } = registeredHandler();
@@ -704,4 +759,23 @@ describe("combineUsage", () => {
 		const none = combineUsage(fakeUsage(), fakeUsage());
 		expect(none.cacheWrite1h).toBeUndefined();
 	});
+
+	it("marks the cost provider-reported only when every part is (pi 1.5.0)", () => {
+		const withFlag = (flag?: boolean): UsageLike => {
+			const usage = fakeUsage();
+			if (flag !== undefined) usage.cost.totalFromProvider = flag;
+			return usage;
+		};
+		expect(
+			combineUsage(withFlag(true), withFlag(true)).cost.totalFromProvider,
+		).toBe(true);
+		expect(
+			combineUsage(withFlag(true), withFlag(false)).cost.totalFromProvider,
+		).toBe(false);
+		// Absent on either side is not "provider-reported"; the field is always
+		// written as an explicit boolean, exactly like the built-in.
+		expect(
+			combineUsage(withFlag(), withFlag(true)).cost.totalFromProvider,
+		).toBe(false);
+	});
 });

package/src/features/compact/index.ts

@@ -12,8 +12,10 @@
  *  - prompts, message assembly, split-turn handling, summary budgets, and
  *    the summary file-list tail are verbatim pi built-in behavior (copied
  *    from 0.87.1; `core/compaction/compaction.js` stayed byte-identical
- *    1.0.0 → 1.3.0, and 1.4.0 changed only the summary output budget and
- *    the split-turn focus handling mirrored below) — only the executing
+ *    1.0.0 → 1.3.0, 1.4.0 changed only the summary output budget and
+ *    the split-turn focus handling mirrored below, and 1.5.0 left it
+ *    untouched; the only mirrored helper that moved after that span is
+ *    `combineUsage`, see prompt.ts) — only the executing
  *    model differs. The one exception is `serializeConversation`, which is
  *    imported from the host instead of mirrored, so it tracks whatever the
  *    running host does (1.2.0 stopped feeding assistant thinking into the
@@ -133,7 +135,7 @@ async function completeSummary(
  * over).
  *
  * 1.4.0 changed two things this mirror covers, and we follow 1.4.0 for every
- * host in the supported window (1.1.0 → 1.4.0) because both are bug fixes
+ * host in the supported window (1.2.0 → 1.5.0) because both are bug fixes
  * rather than behavior we could sensibly keep per version:
  *  - summaries no longer cap maxTokens at 0.8/0.5 × reserveTokens; the
  *    request omits maxTokens so the provider uses the model's output limit

package/src/features/compact/prompt.ts

@@ -6,14 +6,16 @@
  * summaries produced by the configured compact model are formatted exactly
  * like pi's own compaction output. Only the executing model differs.
  * Verified byte-identical against every published version of the supported
- * window (1.1.0 → 1.4.0): core/usage-totals.js is unchanged across that
- * span, the only utils.js change (1.2.0 dropping assistant thinking from
- * serializeConversation) does not touch the strings mirrored here — that
- * serializer is imported from the host, not mirrored — and compaction.js
- * changed once, in 1.4.0, in the output budget and the split-turn focus
- * handling (both adopted here, see buildTurnPrefixPromptText). 0.87.0
- * shipped a different split-turn prefix prompt; the text mirrored here is
- * 0.87.1's fix, and the mirror is the only variant any supported host gets
+ * window (1.2.0 → 1.5.0): the only utils.js change (1.2.0 dropping assistant
+ * thinking from serializeConversation) does not touch the strings mirrored
+ * here — that serializer is imported from the host, not mirrored —
+ * compaction.js last changed in 1.4.0, in the output budget and the
+ * split-turn focus handling (both adopted here, see
+ * buildTurnPrefixPromptText), and core/usage-totals.js changed once past
+ * that point, in 1.5.0, which only added the `cost.totalFromProvider` merge
+ * rule mirrored below (the numeric fields are untouched). 0.87.0 shipped a
+ * different split-turn prefix prompt; the text mirrored here is 0.87.1's
+ * fix, and the mirror is the only variant any supported host gets
  * — intentional.
  *
  * Update these strings whenever pi changes its built-in prompts (diff
@@ -213,9 +215,12 @@ export function formatFileOperations(
 
 /**
  * Combine two provider usage records field-wise (pi 0.87.1 compaction.js
- * combineUsage, moved verbatim to core/usage-totals.js in 0.99.0 and
- * unchanged through 1.4.0): used when
- * a split turn produces two summarization calls.
+ * combineUsage, moved verbatim to core/usage-totals.js in 0.99.0; the
+ * numeric fields are unchanged through 1.5.0, which added the
+ * `cost.totalFromProvider` rule at the end of the cost block — a sum counts
+ * as provider-reported only when every part does, and the field is written
+ * as an explicit boolean either way): used when a split turn produces two
+ * summarization calls.
  */
 export function combineUsage<T extends UsageLike>(first: T, second: T): T {
 	return {
@@ -239,6 +244,10 @@ export function combineUsage<T extends UsageLike>(first: T, second: T): T {
 			cacheRead: first.cost.cacheRead + second.cost.cacheRead,
 			cacheWrite: first.cost.cacheWrite + second.cost.cacheWrite,
 			total: first.cost.total + second.cost.total,
+			// A sum is provider-reported only when every part is (pi 1.5.0).
+			totalFromProvider:
+				first.cost.totalFromProvider === true &&
+				second.cost.totalFromProvider === true,
 		},
 	} as T;
 }
@@ -258,5 +267,7 @@ export interface UsageLike {
 		cacheRead: number;
 		cacheWrite: number;
 		total: number;
+		/** pi 1.5.0: true only when the provider reported the cost itself. */
+		totalFromProvider?: boolean;
 	};
 }

package/src/features/favorites/constants.ts

@@ -9,9 +9,9 @@
  * keybinding conflicts arise.
  *
  * Verified against every published version of the supported range
- * (0.99.0 → 1.4.0) keybinding
- * tables + pi-tui input dispatch on WSL2 + Windows Terminal; re-run
- * `pnpm check:keybindings` after every host upgrade before trusting this.
+ * (0.99.0 → 1.5.0) keybinding tables + pi-tui input dispatch on WSL2 +
+ * Windows Terminal; re-run `pnpm check:keybindings` after every host upgrade
+ * before trusting this.
  */
 export const FAVORITES_STATUS_KEY = "favorites";
 
@@ -29,13 +29,19 @@ export const FAVORITES_FILE = "pi-toolkits-favorites.json";
  * (0.87–0.99 defaulted to `regular`, where the binding is inert). On
  * Windows/WSL `tui.altScreen.search` resolves to ctrl+f, and
  * `TuiAltScreen.handleViewportInput()` checks it before
- * `shouldDeferViewportInputToOverlay()`, so the keystroke opens transcript
- * search and the selector never sees it (verified in pi-tui 1.4.0
- * `dist/tui-alt-screen.js`: the search check is the first key branch and
- * returns `{ consume: true }` unconditionally; the 1.4.0 changes to that
- * file are exit/teardown only).
+ * `shouldDeferViewportInput()` (named `shouldDeferViewportInputToOverlay()`
+ * through 1.4.0), so the keystroke opens transcript
+ * search and the selector never sees it (verified in pi-tui 1.5.0
+ * `dist/tui-alt-screen.js`: the search check is still the first key branch
+ * and still returns `{ consume: true }` unconditionally; the 1.5.0 changes to
+ * that file are exit/teardown, word-selection joiners, and a rename of
+ * `shouldDeferViewportInputToOverlay()` to `shouldDeferViewportInput()` —
+ * which now also defers viewport keys to a focused component that declares
+ * `capturesViewportInput`, but that guard sits BELOW the search branch, so
+ * it does not rescue ctrl+f for the selector even though 1.5.0 gives the
+ * selector `capturesViewportInput = true`).
  *
- * Why ctrl+alt+f was discarded: clean at the pi layer across 1.1–1.4 (only
+ * Why ctrl+alt+f was discarded: clean at the pi layer across 1.1–1.5 (only
  * ctrl+alt+] is taken from that family) and byte-distinguishable without
  * kitty — but grabbed at the OS/app layer on the author's machine by a
  * Windows screenshot tool, so it never reached the terminal. Lesson: a key

package/src/features/favorites/favorites.test.ts

@@ -39,6 +39,7 @@ import {
 	setSharedFavoritesStore,
 } from "./store.js";
 import { applyCyclePatch } from "./patch-cycle.js";
+import { styleFavorite } from "./constants.js";
 import {
 	applyModelSelectorPatches,
 	setSelectorThemeProvider,
@@ -464,4 +465,81 @@ describe("selector patches", () => {
 			"claude-opus-4",
 		]);
 	});
+
+	it("mirrors pi 1.5.0's disambiguating model name after the id", () => {
+		// The name is shown only when it adds words the id does not convey
+		// (pi's getDistinctModelName rule, copied — it is not exported).
+		type LineItem = {
+			provider: string;
+			id: string;
+			model: { provider: string; id: string; name?: string };
+		};
+		const redundant: LineItem = {
+			provider: "anthropic",
+			id: "claude-sonnet-4",
+			model: { provider: "anthropic", id: "claude-sonnet-4", name: "Claude Sonnet 4" },
+		};
+		const renamed: LineItem = {
+			provider: "openrouter",
+			id: "minimax/minimax-m3",
+			model: {
+				provider: "openrouter",
+				id: "minimax/minimax-m3",
+				name: "MiniMax M3 [rasmus@openrouter, parasail]",
+			},
+		};
+		const lines: string[] = [];
+		const selector = fakeSelector({
+			activeModels: [],
+			filteredModels: [redundant, renamed] as unknown as FakeSelector["filteredModels"],
+			listContainer: {
+				clear() {},
+				addChild: (child: unknown) =>
+					lines.push((child as { text?: string }).text ?? ""),
+			},
+		});
+		selector.updateList();
+		const redundantLine = lines.find((l) => l.includes("claude-sonnet-4"));
+		const renamedLine = lines.find((l) => l.includes("minimax/minimax-m3"));
+		expect(redundantLine).toBeDefined();
+		expect(redundantLine).not.toContain("(Claude Sonnet 4)");
+		expect(renamedLine).toContain(
+			"(MiniMax M3 [rasmus@openrouter, parasail])",
+		);
+		// The full line pins the 1.5.0 column order (id → name → provider badge)
+		// and the favorite highlight not leaking ANSI codes into the name. The
+		// two spaces after → are the (empty) current-marker column; the fake
+		// selector has no currentModel, so no ✓ is drawn.
+		expect(redundantLine).toBe("→   claude-sonnet-4 [anthropic]");
+		expect(renamedLine).toBe(
+			"    minimax/minimax-m3 (MiniMax M3 [rasmus@openrouter, parasail]) [openrouter]",
+		);
+	});
+
+	it("keeps the favorite highlight while appending the disambiguating name", () => {
+		// Regression guard: the 1.5.0 name text must land AFTER the styled id
+		// (and outside its ANSI reset), or favorited ids would lose their color.
+		const favoriteModel = {
+			provider: "anthropic",
+			id: "claude-opus-4",
+			name: "Claude Opus 4 (max)",
+		};
+		getFavoritesStore().toggle(favoriteModel);
+		const lines: string[] = [];
+		const selector = fakeSelector({
+			activeModels: [],
+			filteredModels: [
+				{ provider: favoriteModel.provider, id: favoriteModel.id, model: favoriteModel },
+			] as unknown as FakeSelector["filteredModels"],
+			listContainer: {
+				clear() {},
+				addChild: (child: unknown) =>
+					lines.push((child as { text?: string }).text ?? ""),
+			},
+		});
+		selector.updateList();
+		expect(lines[0]).toBe(
+			`→   ${styleFavorite("claude-opus-4")} (Claude Opus 4 (max)) [anthropic]`,
+		);
+	});
 });

package/src/features/favorites/patch-cycle.ts

@@ -15,8 +15,9 @@
  *    model available" feedback); otherwise jump to it
  *  - the apply logic mirrors pi's _cycleScopedModel/_cycleAvailableModel
  *    (byte-identical across every published version in the supported window
- *    1.1.0 → 1.4.0 — `cycleModel` itself is untouched by 1.4.0's recovery and
- *    tool-restore rewrites:
+ *    1.2.0 → 1.5.0 — `cycleModel` itself is untouched by 1.4.0's recovery and
+ *    tool-restore rewrites and by 1.5.0's auth-routing/loadout-ordering
+ *    changes:
  *    state swap, session persistence, default-model persistence only when
  *    `options.persist` is set, thinking-level re-clamp, model_select emit
  *    with source "cycle")
@@ -32,9 +33,9 @@ import { getFavoritesStore, modelKey } from "./store.js";
 
 /**
  * pi's `Model<any>` — the exact type of `ModelCycleResult.model`
- * (pi-coding-agent 1.1–1.4; the field is unchanged across that span).
+ * (pi-coding-agent 1.2–1.5; the field is unchanged across that span).
  * Derived here instead of importing `Model` from @earendil-works/pi-ai,
- * whose index.d.ts (1.1 through 1.4.0 alike) re-exports it via an
+ * whose index.d.ts (1.2 through 1.5.0 alike) re-exports it via an
  * extensioned `"./types.ts"` specifier that not every TS resolver substitutes
  * to `types.d.ts` (tsc 5.9 bundler does; other tsserver builds don't).
  */
@@ -58,7 +59,7 @@ interface CycleSessionInternals {
 		setDefaultModelAndProvider(provider: string, id: string): void;
 	};
 	/**
-	 * pi 1.0–1.3: keeps a persisted default usable when cycling inside a
+	 * pi 1.0–1.5: keeps a persisted default usable when cycling inside a
 	 * non-empty --models scope.
 	 */
 	_addPersistedDefaultToNonEmptyScope(model: unknown): void;
@@ -141,7 +142,7 @@ export function applyCyclePatch(): boolean {
 			return original.call(this, direction, options);
 		}
 
-		// Same source and availability filtering as pi's stock cycling (1.0–1.3).
+		// Same source and availability filtering as pi's stock cycling (1.0–1.5).
 		const scoped = this._scopedModels.length > 0;
 		let candidates: CycleCandidate[];
 		if (scoped) {
@@ -186,7 +187,7 @@ export function applyCyclePatch(): boolean {
 				: this._getThinkingLevelForModelSwitch();
 		this.agent.state.model = next.model;
 		this.sessionManager.appendModelChange(next.model.provider, next.model.id);
-		// pi 1.0–1.3: only persist the default model when the caller asked for it.
+		// pi 1.0–1.5: only persist the default model when the caller asked for it.
 		if (options.persist) {
 			this.settingsManager.setDefaultModelAndProvider(
 				next.model.provider,

package/src/features/favorites/patch-selector.ts

@@ -21,8 +21,10 @@
  * upgraded internals), patches are skipped with a warning — the extension
  * keeps loading and ctrl+p falls back to stock behavior. The prototype
  * shape touched here (handleInput/updateList/filterModels and the instance
- * fields below) is byte-identical across the mandatory window
- * pi 1.1.0 → 1.4.0 (and exercised by the 0.99.0 → 1.0.4 backward probes).
+ * fields below) is unchanged across the mandatory window
+ * pi 1.2.0 → 1.5.0 (and exercised by the 0.99.0 → 1.1.0 backward probes).
+ * updateList gained one layout change in 1.5.0 — the disambiguating model
+ * name after the id — mirrored here; see distinctModelName below.
  */
 import {
 	ModelSelectorComponent,
@@ -61,7 +63,7 @@ interface SelectorPatchTarget {
 	handleInput(keyData: string): void;
 	updateList(): void;
 	filterModels(query: string): void;
-	/** pi 1.0–1.3: default-model badge source. */
+	/** pi 1.0–1.5: default-model badge source. */
 	isDefaultModel(model: ModelRef): boolean;
 	listContainer: ListContainerLike;
 	searchInput: SearchInputLike;
@@ -95,6 +97,37 @@ function sameModel(a: ModelRef, b: ModelRef): boolean {
 	return a.provider === b.provider && a.id === b.id;
 }
 
+/** Lowercase a label and split it into word tokens (pi's splitModelLabel). */
+function splitModelLabel(value: string): string[] {
+	return value
+		.toLowerCase()
+		.split(/[^\p{L}\p{N}]+/u)
+		.filter(Boolean);
+}
+
+/**
+ * Replica of pi 1.5.0's internal `getDistinctModelName`
+ * (`dist/modes/interactive/model-search.js`). It is not reachable from the
+ * package's `.` entry, so the rule is copied instead of imported: show the
+ * model name only when it has words the id does not already convey, which
+ * keeps renamed overrides visible without repeating `claude-sonnet-4` after
+ * `claude-sonnet-4-20250514`. Mirrored for every host in the window — the
+ * model `name` field exists across it, and one rendering rule is better than
+ * a version branch.
+ */
+function distinctModelName(id: string, name?: string): string | undefined {
+	const trimmed = name?.trim();
+	if (!trimmed) return undefined;
+	const idWords = splitModelLabel(id);
+	const idTokens = new Set(idWords);
+	const compactId = idWords.join("");
+	// Single-character words must match a whole id token: "4" is not covered
+	// by "mistral-small-2603".
+	const isCovered = (word: string): boolean =>
+		idTokens.has(word) || (word.length > 1 && compactId.includes(word));
+	return splitModelLabel(trimmed).every(isCovered) ? undefined : trimmed;
+}
+
 function buildHintText(state: SelectorPatchTarget): string {
 	const theme = themeProvider();
 	if (!theme) return "";
@@ -116,7 +149,7 @@ let patched = false;
 export function applyModelSelectorPatches(): boolean {
 	if (patched) return true;
 	// SAFETY: ModelSelectorComponent's members are TS-private (compile-time
-	// only); the structural view matches pi 1.0–1.3's prototype shape
+	// only); the structural view matches pi 1.0–1.5's prototype shape
 	// (byte-identical across that span) and every touched method is
 	// existence-checked below before any patching happens.
 	const proto =
@@ -155,10 +188,11 @@ export function applyModelSelectorPatches(): boolean {
 	};
 
 	// ── 2. List rendering: hint row + bold bright yellow favorites ────────
-	// Replicates the built-in updateList layout (pi 1.0–1.3: fixed-width cursor
-	// and ✓ current-marker columns, provider badge, `· default` badge, scroll
-	// indicator, empty/error states, refresh status, Model Name footer) with
-	// one change: favorited model ids render in the favorite style.
+	// Replicates the built-in updateList layout (fixed-width cursor and ✓
+	// current-marker columns, the 1.5.0 disambiguating model name, provider
+	// badge, `· default` badge, scroll indicator, empty/error states, refresh
+	// status, Model Name footer) with one change: favorited model ids render
+	// in the favorite style.
 	proto.updateList = function (this: SelectorPatchTarget): void {
 		const theme = themeProvider();
 		if (!theme) {
@@ -197,8 +231,10 @@ export function applyModelSelectorPatches(): boolean {
 			const currentMarker = isCurrent ? theme.fg("accent", "✓ ") : "  ";
 			const baseText = isSelected ? theme.fg("accent", item.id) : item.id;
 			const modelText = isFavorite ? styleFavorite(item.id) : baseText;
+			const name = distinctModelName(item.id, item.model.name);
+			const nameText = name ? theme.fg("muted", ` (${name})`) : "";
 			const providerBadge = theme.fg("muted", `[${item.provider}]`);
-			const line = `${cursor}${currentMarker}${modelText} ${providerBadge}${defaultBadge}`;
+			const line = `${cursor}${currentMarker}${modelText}${nameText} ${providerBadge}${defaultBadge}`;
 			list.addChild(new Text(line, 0, 0));
 		}
 		if (startIndex > 0 || endIndex < this.filteredModels.length) {

package/src/features/init/detect.test.ts

@@ -11,6 +11,10 @@ import {
 	scanRefs,
 } from "./detect.js";
 import { project, remove } from "./test-helpers.js";
+// Static import (not `await import` inside the test): the per-test timeout
+// must not have to cover transforming/loading the deterministic half of the
+// feature — that is what made this test flake under matrix load.
+import { stateToToml } from "./apply.js";
 
 describe("isBareRef", () => {
 	const text = "see memory/DESIGN.md, .agents/memory/X.md, my-memory/y, ~/.pi/agent/memory/z";
@@ -207,8 +211,7 @@ describe("review regressions", () => {
 		expect(countBareRefs(text, "implementations/")).toBe(1);
 	});
 
-	it("round-trips custom budgets through the state TOML keys", async () => {
-		const { stateToToml } = await import("./apply.js");
+	it("round-trips custom budgets through the state TOML keys", () => {
 		const parsed = parseState(
 			stateToToml({
 				version: 1,

package/src/features/stash/stash.ts

@@ -21,7 +21,7 @@ import type { KeyId } from "@earendil-works/pi-tui";
 /**
  * Stash hotkey.
  *
- * Compatibility rationale (verified against pi 0.99.0–1.4.0 keybindings, live
+ * Compatibility rationale (verified against pi 0.99.0–1.5.0 keybindings, live
  * matchesKey probes and terminal/OS behavior):
  *  - pi: no default binding for ctrl+alt+<letter> (only ctrl+alt+] is taken);
  *    extension shortcuts dispatch first while the editor is focused

package/src/lib/gates.ts

@@ -18,9 +18,9 @@
  *    promise. The queue serializes our own presentations; the host slot is
  *    still shared with other extensions (pi does not arbitrate globally).
  *    pi 1.4.0 stacks extension dialogs so a second one no longer strands the
- *    first (#11166), but the 1.1–1.3 hosts in the supported window still
+ *    first (#11166), but the 1.2–1.3 hosts in the supported window still
  *    replace it — the queue stays, and it also keeps one confirmation visible
- *    at a time instead of a pile of stacked dialogs on 1.4.0.
+ *    at a time instead of a pile of stacked dialogs on later hosts.
  *  - A gate that throws blocks the call (fail closed), matching pi's own
  *    `tool_call` handler contract.
  *

package/src/lib/model-config.ts

@@ -4,8 +4,8 @@
  * (review, compact-model — favorites keeps its own store-local key logic).
  *
  * Types are derived from pi-coding-agent's own exports (ModelRegistry)
- * rather than imported from @earendil-works/pi-ai: pi-ai's index.d.ts (1.1.0
- * through 1.4.0 alike) re-exports `Model`/`Api` via extensioned
+ * rather than imported from @earendil-works/pi-ai: pi-ai's index.d.ts (1.2.0
+ * through 1.5.0 alike) re-exports `Model`/`Api` via extensioned
  * `./types.ts` specifiers, which tsc 5.9 bundler resolution substitutes to
  * `.d.ts` but some tsserver builds do not. Deriving from ModelRegistry keeps
  * every resolver in agreement with the exact type

package/src/lib/model-picker.ts

@@ -9,11 +9,12 @@
  *
  * The component is driven through a small ModelRuntime facade backed by
  * ctx.modelRegistry (pi's extension-facing wrapper). SAFETY: verified
- * against pi 1.1.0–1.4.0 dist (modes/interactive/components/model-selector.js +
- * model-catalog-refresh.js — byte-identical across that span) — the
- * component only calls getAvailableSnapshot, getModel, getError and refresh
- * on modelRuntime, and ModelRegistry provides exactly those
- * (getAvailable/find/getError/refresh).
+ * against pi 1.2.0–1.5.0 dist (modes/interactive/components/model-selector.js +
+ * model-catalog-refresh.js) — the component only calls getAvailableSnapshot,
+ * getModel and getError on modelRuntime (plus refresh, which ModelRegistry
+ * provides) across that whole span. 1.5.0 changed the selector's rendering
+ * and key handling (page keys, disambiguating model names, viewport-input
+ * capture) but not the runtime surface this facade implements.
  *
  * `current` is passed as the component's currentModel: the built-in selector
  * preselects, ✓-marks and sorts-first its currentModel, so one Enter accepts
@@ -42,9 +43,10 @@ function modelRuntimeFacade(ctx: ExtensionContext): ModelRuntime {
 		refresh: (options?: { signal?: AbortSignal }) =>
 			ctx.modelRegistry.refresh(options),
 	};
-	// SAFETY: the selector only touches the four members above (pi 1.0–1.3
-	// model-selector.js); everything else on ModelRuntime is never reached
-	// from this component.
+	// SAFETY: the selector only touches the four members above (pi 1.0–1.5
+	// model-selector.js; the 1.5.0 rendering/page-key changes add no runtime
+	// call); everything else on ModelRuntime is never reached from this
+	// component.
 	return facade as unknown as ModelRuntime;
 }