@oh-my-pi/pi-coding-agent 15.8.3 → 15.9.1

package/src/prompts/system/title-system.md

@@ -1,2 +1,3 @@
-Generate a 3-6 word title for a coding session from the user's first message. Capture the main task or topic.
-Output ONLY the title. No quotes or trailing punctuation.
+Need generate 3-6 word title from first message; capture main task
+Output title only; no quotes no punctuation
+If message has no concrete task yet (greeting, small talk, vague), output exactly: none

package/src/prompts/system/workflow-notice.md

@@ -62,7 +62,7 @@ Scale to the ask: "find any bugs" → a few finders, single-vote verify. "thorou
 </patterns>
 
 <execution>
-- Decompose the surface first; capture it in `todo_write` when it spans phases.
+- Decompose the surface first; capture it in `todo` when it spans phases.
 - Prefer `schema=` for any agent whose output you branch on.
 - After a fan-out returns, YOU own correctness: read the artifacts, run the gate, verify before acting. Subagents do the legwork; they don't get the last word.
 - Keep going until the task is closed — a returned fan-out is a step, not a stopping point.

package/src/prompts/tools/browser.md

@@ -1,41 +1,41 @@
-Drives a real Chromium tab with full puppeteer access via JS execution.
+Drives real Chromium tab; full puppeteer access via JS execution.
 
 <instruction>
-- For static web content (articles, docs, issues/PRs, JSON, PDFs, feeds), prefer the `read` tool with a URL — reader-mode text without spinning up a browser. Use this tool when you need JS execution, authentication, or interactive actions.
+- For static web content (articles, docs, issues/PRs, JSON, PDFs, feeds), prefer `read` tool with URL — reader-mode text without spinning up browser. Use this tool when Need JS execution, authentication, or interactive actions.
 - Three actions only:
-  - `open` — acquire (or reuse) a named tab. `name` defaults to `"main"`. Optional `url` navigates after the tab is ready. Optional `viewport` sets dimensions. Optional `dialogs: "accept" | "dismiss"` auto-handles `alert`/`confirm`/`beforeunload` so navigation/clicks don't hang (default: leave dialogs unhandled — page hangs until caller wires `page.on('dialog', …)`).
-  - `close` — release a tab by `name`, or every tab with `all: true`. For spawned-app browsers, set `kill: true` to terminate the process tree (default leaves it running).
-  - `run` — execute JS against an existing tab. `code` is the body of an async function with `page`, `browser`, `tab`, `display`, `assert`, `wait` in scope. The function's return value is JSON-stringified into the tool result; multiple `display(value)` calls accumulate text/images.
+  - `open` — acquire or reuse named tab. `name` defaults `"main"`. Optional `url` navigates after tab ready. Optional `viewport` sets dimensions. Optional `dialogs: "accept" | "dismiss"` auto-handles `alert`/`confirm`/`beforeunload` so navigation/clicks don't hang (default: leave dialogs unhandled — page hangs until caller wires `page.on('dialog', …)`).
+  - `close` — release tab by `name`, or every tab with `all: true`. For spawned-app browsers, set `kill: true` to terminate process tree (default leaves running).
+  - `run` — execute JS against existing tab. `code` is body of async function with `page`, `browser`, `tab`, `display`, `assert`, `wait` in scope. Function's return value JSON-stringified into tool result; multiple `display(value)` calls accumulate text/images.
 - Tabs survive across `run` calls and across in-process subagents. Open once, reuse many times.
-- Browser kinds, selected by the `app` field on `open`:
+- Browser kinds, selected by `app` field on `open`:
   - default (no `app`) → headless Chromium with stealth patches.
-  - `app.path` → spawn an absolute binary (Electron/CDP). If a running instance already exposes a CDP port, it is reused; otherwise stale instances are killed and a fresh one is spawned. No stealth patches — never tamper with a real desktop app.
-  - `app.cdp_url` → connect to an existing CDP endpoint (e.g. `http://127.0.0.1:9222`).
-  - `app.target` (with `path`/`cdp_url`) — substring matched against url+title to pick a BrowserWindow when the app exposes several.
-- Inside `run`, `tab` exposes high-level helpers; reach for `page` (raw puppeteer Page) when you need anything they don't cover.
-  - `tab.goto(url, { waitUntil? })` — clears the element cache and navigates.
-  - `tab.observe({ includeAll?, viewportOnly? })` — accessibility snapshot. Returns `{ url, title, viewport, scroll, elements: [{ id, role, name, value, states, … }] }`. Element ids are stable until the next observe/goto.
-  - `tab.id(n)` — resolves an element id from the most recent observe to a real `ElementHandle` you can `.click()`, `.type()`, etc.
+  - `app.path` → spawn absolute binary (Electron/CDP). If running instance already exposes CDP port, reused; otherwise stale instances killed, fresh one spawned. No stealth patches — NEVER tamper with real desktop app.
+  - `app.cdp_url` → connect to existing CDP endpoint (e.g. `http://127.0.0.1:9222`).
+  - `app.target` (with `path`/`cdp_url`) — substring matched against url+title to pick BrowserWindow when app exposes several.
+- Inside `run`, `tab` exposes high-level helpers; reach for `page` (raw puppeteer Page) when Need anything they don't cover.
+  - `tab.goto(url, { waitUntil? })` — clears element cache and navigates.
+  - `tab.observe({ includeAll?, viewportOnly? })` — accessibility snapshot. Returns `{ url, title, viewport, scroll, elements: [{ id, role, name, value, states, … }] }`. Element ids stable until next observe/goto.
+  - `tab.id(n)` — resolves element id from most recent observe to real `ElementHandle` you can `.click()`, `.type()`, etc.
   - `tab.click(selector)` / `tab.type(selector, text)` / `tab.fill(selector, value)` / `tab.press(key, { selector? })` / `tab.scroll(dx, dy)` — selector-based actions.
-  - `tab.waitFor(selector)` — waits until the selector is attached, returns the resolved `ElementHandle` for chaining (e.g. `const btn = await tab.waitFor('text/Submit'); await btn.click();`).
-  - `tab.drag(from, to)` — drag from one point to another. Each endpoint is either a selector string (drag center-to-center) or a `{ x, y }` viewport-coordinate point (e.g. for canvases, sliders).
-  - `tab.scrollIntoView(selector)` — scroll the matching element to the center of the viewport (use before clicking off-screen elements).
-  - `tab.select(selector, …values)` — set the selected option(s) on a `<select>`. Returns the values that ended up selected. `tab.fill` NEVER works for selects.
-  - `tab.uploadFile(selector, …filePaths)` — attach files to an `<input type="file">`. Paths resolve relative to cwd.
-  - `tab.waitForUrl(pattern, { timeout? })` — pattern is a substring or `RegExp`. Polls `location.href` so it works for SPA pushState navigations, not just real navigations. Returns the matched URL.
-  - `tab.waitForResponse(pattern, { timeout? })` — pattern is a substring, `RegExp`, or `(response) => boolean`. Returns the raw puppeteer `HTTPResponse` (call `.text()` / `.json()` / `.status()` / `.headers()` on it).
-  - `tab.evaluate(fn, …args)` — sugar for `page.evaluate` with the abort signal already wired. Use this instead of dropping to `page.evaluate` for ad-hoc DOM reads.
-  - `tab.screenshot({ selector?, fullPage?, save?, silent? })` — auto-attaches the image to the tool output unless `silent: true`. Saves full-res to `save` (or `browser.screenshotDir` setting) and a downscaled copy to the model.
+  - `tab.waitFor(selector)` — waits until selector attached, returns resolved `ElementHandle` for chaining (e.g. `const btn = await tab.waitFor('text/Submit'); await btn.click();`).
+  - `tab.drag(from, to)` — drag from one point to another. Each endpoint either selector string (drag center-to-center) or `{ x, y }` viewport-coordinate point (for canvases, sliders).
+  - `tab.scrollIntoView(selector)` — scroll matching element to center of viewport (use before clicking off-screen elements).
+  - `tab.select(selector, …values)` — set selected option(s) on `<select>`. Returns values that ended up selected. `tab.fill` NEVER works for selects.
+  - `tab.uploadFile(selector, …filePaths)` — attach files to `<input type="file">`. Paths resolve relative to cwd.
+  - `tab.waitForUrl(pattern, { timeout? })` — pattern substring or `RegExp`. Polls `location.href` so works for SPA pushState navigations, not just real navigations. Returns matched URL.
+  - `tab.waitForResponse(pattern, { timeout? })` — pattern substring, `RegExp`, or `(response) => boolean`. Returns raw puppeteer `HTTPResponse` (call `.text()` / `.json()` / `.status()` / `.headers()` on it).
+  - `tab.evaluate(fn, …args)` — sugar for `page.evaluate` with abort signal already wired. Use this instead of dropping to `page.evaluate` for ad-hoc DOM reads.
+  - `tab.screenshot({ selector?, fullPage?, save?, silent? })` — captures screenshot and **auto-attaches to tool output for you to view** (unless `silent: true`). `save` is **strictly optional**: OMIT when you just want to look at page — downscaled image shown regardless, full-res capture written to temp file automatically. Pass `save` (a path) ONLY when deliberately need to keep full-res copy on disk for later use; `browser.screenshotDir` does same for every shot. NEVER invent `save` path for throwaway/temporal screenshot.
   - `tab.extract(format = "markdown")` — Readability-extracted page content.
-- Selectors accept CSS as well as puppeteer query handlers: `aria/Sign in`, `text/Continue`, `xpath/…`, `pierce/…`. Playwright-style `p-aria/[name="…"]`, `p-text/…`, etc. are normalized.
-- Default to `tab.observe()` over `tab.screenshot()` for understanding page state. Screenshot only when visual appearance matters.
+- Selectors accept CSS plus puppeteer query handlers: `aria/Sign in`, `text/Continue`, `xpath/…`, `pierce/…`. Playwright-style `p-aria/[name="…"]`, `p-text/…` normalized.
+- Default `tab.observe()` over `tab.screenshot()` for page state. Screenshot only when visual appearance matters.
 </instruction>
 
 <critical>
-- You MUST call `open` before `run`. `run` does not implicitly create a tab.
-- You NEVER screenshot just to "see what's on the page" — `tab.observe()` returns structured data with element ids you can act on immediately.
-- After a `tab.goto()` or any navigation, prior element ids from `tab.observe()` are invalidated. Re-observe before referencing them.
-- `code` runs with full Node access. Treat it as your code, not sandboxed code.
+- MUST call `open` before `run`. `run` does not implicitly create tab.
+- NEVER screenshot just to "see what's on page" — `tab.observe()` returns structured data with element ids you can act on immediately.
+- After `tab.goto()` or any navigation, prior element ids from `tab.observe()` invalidated. Re-observe before referencing them.
+- `code` runs with full Node access. Treat as your code, not sandboxed code.
 </critical>
 
 <examples>
@@ -46,7 +46,10 @@ Drives a real Chromium tab with full puppeteer access via JS execution.
 # Click an observed element by id
 `{"action":"run","name":"docs","code":"const obs = await tab.observe(); const link = obs.elements.find(e => e.role === 'link' && e.name === 'Sign in'); assert(link, 'Sign in link missing'); await (await tab.id(link.id)).click();"}`
 
-# Save a full-page screenshot to disk
+# Take a transient screenshot just to look at the page — NO save path needed; the image is shown to you
+`{"action":"run","name":"docs","code":"await tab.screenshot();"}`
+
+# Persist a full-page screenshot to disk (only when you deliberately need to keep the file)
 `{"action":"run","name":"docs","code":"await tab.screenshot({ fullPage: true, save: 'screenshot.png' });"}`
 
 # Fill and submit a form via selectors
@@ -66,5 +69,5 @@ Drives a real Chromium tab with full puppeteer access via JS execution.
 </examples>
 
 <output>
-- Per call: any `display(value)` outputs (text/images) followed by the JSON-stringified return value of the `code` function. `run` always produces at least a status line.
+- Per call: any `display(value)` outputs (text/images) followed by JSON-stringified return value of `code` function. `run` always produces at least status line.
 </output>

package/src/prompts/tools/render-mermaid.md

@@ -5,5 +5,5 @@ Parameters:
 - `config` (optional): JSON render configuration (spacing and layout options).
 Behavior:
 - Returns ASCII diagram text.
-- Saves full output to `artifact://<id>` when storage is available.
-- Returns error when Mermaid input is invalid or rendering fails.
+- Saves full output to `artifact://<id>` when storage available.
+- Returns error when Mermaid input invalid or rendering fails.

package/src/prompts/tools/search.md

@@ -2,7 +2,7 @@ Searches files using powerful regex matching.
 
 <instruction>
 - Supports Rust regex syntax (RE2-style — no lookaround or backreferences). Use line anchors or post-filters instead of (?!…)/(?<!…)
-- `paths` is required and accepts either one string or an array of files, directories, globs, or internal URLs
+- `paths` accepts either one string or an array of files, directories, globs, or internal URLs. Optional: when omitted or empty it searches the workspace root (`.`). Prefer scoping to specific paths when you know them.
 - For multiple targets, pass an array with one target per element: `["src", "tests"]`.
 - Cross-line patterns are detected from literal `\n` or escaped `\\n` in `pattern`
 </instruction>

package/src/prompts/tools/task.md

@@ -38,6 +38,8 @@ Subagents have no conversation history. Every fact, file path, and direction the
 - Pass large payloads via `local://<path>` URIs, not inline. {{#if contextEnabled}} (other than the context){{/if}}
 {{#if contextEnabled}}- Put shared constraints in `context` once; do not duplicate across assignments.{{/if}}
 - Prefer agents that investigate **and** edit in one pass; only spin a read-only discovery step when affected files are genuinely unknown.
+- **Read-only agents**: Agents tagged READ-ONLY (e.g. `explore`) have no edit/write/command tools. NEVER hand them an assignment that requires changing files or running commands — they cannot do it and the turn is wasted. Use them to investigate and report back; do the edits yourself or delegate to a writing agent (`task`, `oracle`, `designer`).
+- **No reasoning offload**: NEVER offload reasoning, analysis, design, or decision-making to `quick_task` or `explore` — they run minimal-effort / small models for mechanical lookups and data collection only. Keep judgment and synthesis in your own context; delegate hard thinking to `task`, `plan`, or `oracle`.
 </rules>
 
 <parallelization>
@@ -71,7 +73,7 @@ Parallel when tasks touch disjoint files or are independent refactors/tests.
 Agent spawning is disabled for this context.
 {{else}}
 {{#list agents join="\n"}}
-# {{name}}
+# {{name}}{{#if readOnly}} — READ-ONLY (no edit/write/exec tools){{/if}}
 {{description}}
 {{/list}}
 {{/if}}

package/src/prompts/tools/{todo-write.md → todo.md}

@@ -50,9 +50,9 @@ Allowed `op` values are only `init`, `start`, `done`, `drop`, `rm`, `append`, an
 </examples>
 
 <critical>
-When the user hands you a multi-step plan — a phased todo, a numbered or bulleted checklist, or "N bugs/items/tasks" to work through: 
-- You MUST `init` the list with EVERY item as its own task before doing the work. 
-- Enumerate all of them; 
-- NEVER summarize the plan into fewer tasks, sample "the important ones", drop items, or rely on memory to track the rest. 
+When the user hands you a multi-step plan — a phased todo, a numbered or bulleted checklist, or "N bugs/items/tasks" to work through:
+- You MUST `init` the list with EVERY item as its own task before doing the work.
+- Enumerate all of them;
+- NEVER summarize the plan into fewer tasks, sample "the important ones", drop items, or rely on memory to track the rest.
 The entire point is to remember every one.
-</critical>
+</critical>

package/src/sdk.ts

@@ -39,6 +39,7 @@ import { createAutoresearchExtension } from "./autoresearch";
 import { loadCapability } from "./capability";
 import { type Rule, ruleCapability, setActiveRules } from "./capability/rule";
 import { bucketRules } from "./capability/rule-buckets";
+import { shouldEnableAppendOnlyContext } from "./config/append-only-context-mode";
 import { ModelRegistry } from "./config/model-registry";
 import {
 	formatModelString,
@@ -101,7 +102,7 @@ import {
 import { AgentSession } from "./session/agent-session";
 import { resolveAuthBrokerConfig } from "./session/auth-broker-config";
 import { AuthBrokerClient, AuthStorage, RemoteAuthCredentialStore } from "./session/auth-storage";
-import { type CustomMessage, convertToLlm } from "./session/messages";
+import { type CustomMessage, convertToLlm, wrapSteeringForModel } from "./session/messages";
 import { getRestorableSessionModels, SessionManager } from "./session/session-manager";
 import { closeAllConnections } from "./ssh/connection-manager";
 import { unmountAll } from "./ssh/sshfs-mount";
@@ -646,24 +647,6 @@ function registerPythonCleanup(): void {
 	postmortem.register("python-cleanup", disposeAllKernelSessions);
 }
 
-/**
- * Resolve whether to enable append-only context mode based on the setting and provider.
- *
- * - `"on"` → always enable
- * - `"off"` → never enable
- * - `"auto"` → enable for DeepSeek (prefix-caching provider)
- */
-function resolveAppendOnlyMode(setting: "auto" | "on" | "off" | undefined, provider: string): boolean {
-	switch (setting ?? "auto") {
-		case "on":
-			return true;
-		case "off":
-			return false;
-		default:
-			return provider === "deepseek";
-	}
-}
-
 function customToolToDefinition(tool: CustomTool): ToolDefinition {
 	const definition: ToolDefinition & { [TOOL_DEFINITION_MARKER]: true } = {
 		name: tool.name,
@@ -1259,6 +1242,7 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 			getModelString: () => (hasExplicitModel && model ? formatModelString(model) : undefined),
 			getActiveModelString,
 			getPlanModeState: () => session?.getPlanModeState(),
+			getPlanReferencePath: () => session?.getPlanReferencePath() ?? "local://PLAN.md",
 			getGoalModeState: () => session?.getGoalModeState(),
 			getGoalRuntime: () => session?.goalRuntime,
 			getUsageStatistics: () => sessionManager.getUsageStatistics(),
@@ -1904,7 +1888,8 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 			return obfuscateMessages(obfuscator, converted);
 		};
 		const transformContext = async (messages: AgentMessage[], _signal?: AbortSignal) => {
-			return await extensionRunner.emitContext(messages);
+			const withContext = await extensionRunner.emitContext(messages);
+			return wrapSteeringForModel(withContext);
 		};
 		const onPayload = async (payload: unknown, _model?: Model) => {
 			return await extensionRunner.emitBeforeProviderRequest(payload);
@@ -2027,7 +2012,7 @@ export async function createAgentSession(options: CreateAgentSessionOptions = {}
 			getToolChoice: () => session?.nextToolChoice(),
 			telemetry: options.telemetry,
 			appendOnlyContext: model
-				? resolveAppendOnlyMode(settings.get("provider.appendOnlyContext"), model.provider)
+				? shouldEnableAppendOnlyContext(settings.get("provider.appendOnlyContext"), model)
 					? new AppendOnlyContextManager()
 					: undefined
 				: undefined,

package/src/session/agent-session.ts

@@ -100,6 +100,7 @@ import { type AsyncJob, type AsyncJobDeliveryState, AsyncJobManager } from "../a
 import { classifyDifficulty } from "../auto-thinking/classifier";
 import { reset as resetCapabilities } from "../capability";
 import type { Rule } from "../capability/rule";
+import { shouldEnableAppendOnlyContext } from "../config/append-only-context-mode";
 import { MODEL_ROLE_IDS, type ModelRegistry } from "../config/model-registry";
 import {
 	extractExplicitThinkingSelector,
@@ -203,7 +204,7 @@ import type { CheckpointState } from "../tools/checkpoint";
 import { outputMeta } from "../tools/output-meta";
 import { normalizeLocalScheme, resolveToCwd } from "../tools/path-utils";
 import { isAutoQaEnabled } from "../tools/report-tool-issue";
-import { getLatestTodoPhasesFromEntries, type TodoItem, type TodoPhase } from "../tools/todo-write";
+import { getLatestTodoPhasesFromEntries, type TodoItem, type TodoPhase } from "../tools/todo";
 import { ToolAbortError, ToolError } from "../tools/tool-errors";
 import { clampTimeout } from "../tools/tool-timeouts";
 import { parseCommandArgs } from "../utils/command-args";
@@ -232,7 +233,7 @@ import type {
 	SessionContext,
 	SessionManager,
 } from "./session-manager";
-import { getLatestCompactionEntry, getRestorableSessionModels } from "./session-manager";
+import { EPHEMERAL_MODEL_CHANGE_ROLE, getLatestCompactionEntry, getRestorableSessionModels } from "./session-manager";
 import type { ShakeMode, ShakeResult } from "./shake-types";
 import { ToolChoiceQueue } from "./tool-choice-queue";
 import { YieldQueue } from "./yield-queue";
@@ -1293,6 +1294,12 @@ export class AgentSession {
 		this.#standingResolveHandler = handler ?? undefined;
 	}
 
+	#sessionSwitchReconciler: (() => Promise<void>) | undefined;
+
+	setSessionSwitchReconciler(reconciler: (() => Promise<void>) | null): void {
+		this.#sessionSwitchReconciler = reconciler ?? undefined;
+	}
+
 	/** Provider-scoped mutable state store for transport/session caches. */
 	get providerSessionState(): Map<string, ProviderSessionState> {
 		return this.#providerSessionState;
@@ -1807,21 +1814,21 @@ export class AgentSession {
 				if (toolName === "edit" && details?.path) {
 					this.#invalidateFileCacheForPath(details.path);
 				}
-				if (toolName === "todo_write" && !isError && Array.isArray(details?.phases)) {
+				if (toolName === "todo" && !isError && Array.isArray(details?.phases)) {
 					this.setTodoPhases(details.phases);
 				}
-				if (toolName === "todo_write" && isError) {
+				if (toolName === "todo" && isError) {
 					const errorText = content?.find(part => part.type === "text")?.text;
 					const reminderText = [
 						"<system-reminder>",
-						"todo_write failed, so todo progress is not visible to the user.",
-						errorText ? `Failure: ${errorText}` : "Failure: todo_write returned an error.",
-						"Fix the todo payload and call todo_write again before continuing.",
+						"todo failed, so todo progress is not visible to the user.",
+						errorText ? `Failure: ${errorText}` : "Failure: todo returned an error.",
+						"Fix the todo payload and call todo again before continuing.",
 						"</system-reminder>",
 					].join("\n");
 					await this.sendCustomMessage(
 						{
-							customType: "todo-write-error-reminder",
+							customType: "todo-error-reminder",
 							content: reminderText,
 							display: false,
 							details: { toolName, errorText },
@@ -4639,6 +4646,7 @@ export class AgentSession {
 		this.agent.steer({
 			role: "user",
 			content,
+			steering: true,
 			attribution: "user",
 			timestamp: Date.now(),
 		});
@@ -4994,7 +5002,7 @@ export class AgentSession {
 	// splice mutated canonical `#todoPhases` between tool calls, so the model
 	// observed phase totals shrinking ("5 → 4") after marking tasks done. The
 	// `tasks.todoClearDelay` setting is now inert; completed tasks survive
-	// until the next explicit `todo_write` call removes them via `rm`/`drop`.
+	// until the next explicit `todo` call removes them via `rm`/`drop`.
 
 	/**
 	 * Abort current operation and wait for agent to become idle.
@@ -5225,7 +5233,11 @@ export class AgentSession {
 	 * Validates API key, saves to session log but NOT to settings.
 	 * @throws Error if no API key available for the model
 	 */
-	async setModelTemporary(model: Model, thinkingLevel?: ThinkingLevel): Promise<void> {
+	async setModelTemporary(
+		model: Model,
+		thinkingLevel?: ThinkingLevel,
+		options?: { ephemeral?: boolean },
+	): Promise<void> {
 		const previousEditMode = this.#resolveActiveEditMode();
 		const apiKey = await this.#modelRegistry.getApiKey(model, this.sessionId);
 		if (!apiKey) {
@@ -5234,7 +5246,10 @@ export class AgentSession {
 
 		this.#clearActiveRetryFallback();
 		this.#setModelWithProviderSessionReset(model);
-		this.sessionManager.appendModelChange(`${model.provider}/${model.id}`, "temporary");
+		this.sessionManager.appendModelChange(
+			`${model.provider}/${model.id}`,
+			options?.ephemeral ? EPHEMERAL_MODEL_CHANGE_ROLE : "temporary",
+		);
 		this.settings.getStorage()?.recordModelUsage(`${model.provider}/${model.id}`);
 
 		// Apply explicit thinking level if given; otherwise prefer the model's
@@ -6553,16 +6568,16 @@ export class AgentSession {
 			return undefined;
 		}
 
-		if (!this.#toolRegistry.has("todo_write")) {
-			logger.warn("Eager todo enforcement skipped because todo_write is unavailable", {
+		if (!this.#toolRegistry.has("todo")) {
+			logger.warn("Eager todo enforcement skipped because todo is unavailable", {
 				activeToolNames: this.agent.state.tools.map(tool => tool.name),
 			});
 			return undefined;
 		}
 
-		const todoWriteToolChoice = buildNamedToolChoice("todo_write", this.model);
-		if (!todoWriteToolChoice) {
-			logger.warn("Eager todo enforcement skipped because the current model does not support forcing todo_write", {
+		const todoToolChoice = buildNamedToolChoice("todo", this.model);
+		if (!todoToolChoice) {
+			logger.warn("Eager todo enforcement skipped because the current model does not support forcing todo", {
 				modelApi: this.model?.api,
 				modelId: this.model?.id,
 			});
@@ -6580,7 +6595,7 @@ export class AgentSession {
 				attribution: "agent",
 				timestamp: Date.now(),
 			},
-			toolChoice: todoWriteToolChoice,
+			toolChoice: todoToolChoice,
 		};
 	}
 	/**
@@ -6682,7 +6697,7 @@ export class AgentSession {
 		if (!targetModel) return false;
 
 		try {
-			await this.setModelTemporary(targetModel);
+			await this.setModelTemporary(targetModel, undefined, { ephemeral: true });
 			logger.debug("Context promotion switched model on overflow", {
 				from: `${currentModel.provider}/${currentModel.id}`,
 				to: `${targetModel.provider}/${targetModel.id}`,
@@ -6735,8 +6750,8 @@ export class AgentSession {
 	 */
 	#syncAppendOnlyContext(model: Model | null | undefined): void {
 		const setting = this.settings.get("provider.appendOnlyContext") ?? "auto";
+		const enable = shouldEnableAppendOnlyContext(setting, model);
 		const providerId = model?.provider;
-		const enable = setting === "on" || (setting === "auto" && providerId === "deepseek");
 		const prev = this.#lastAppendOnlyResolution;
 		if (prev && prev.enable === enable && prev.providerId === providerId) return;
 		this.#lastAppendOnlyResolution = { enable, providerId };
@@ -7848,7 +7863,7 @@ export class AgentSession {
 		const nextThinkingLevel = selector.thinkingLevel ?? currentThinkingLevel;
 
 		this.#setModelWithProviderSessionReset(candidate);
-		this.sessionManager.appendModelChange(`${candidate.provider}/${candidate.id}`, "temporary");
+		this.sessionManager.appendModelChange(`${candidate.provider}/${candidate.id}`, EPHEMERAL_MODEL_CHANGE_ROLE);
 		this.settings.getStorage()?.recordModelUsage(`${candidate.provider}/${candidate.id}`);
 		this.setThinkingLevel(nextThinkingLevel);
 		if (!this.#activeRetryFallback) {
@@ -7921,7 +7936,7 @@ export class AgentSession {
 		const thinkingToApply =
 			currentThinkingLevel === lastAppliedFallbackThinkingLevel ? originalThinkingLevel : currentThinkingLevel;
 		this.#setModelWithProviderSessionReset(primaryModel);
-		this.sessionManager.appendModelChange(`${primaryModel.provider}/${primaryModel.id}`, "temporary");
+		this.sessionManager.appendModelChange(`${primaryModel.provider}/${primaryModel.id}`, EPHEMERAL_MODEL_CHANGE_ROLE);
 		this.settings.getStorage()?.recordModelUsage(`${primaryModel.provider}/${primaryModel.id}`);
 		this.setThinkingLevel(thinkingToApply);
 		this.#clearActiveRetryFallback();
@@ -8948,6 +8963,14 @@ export class AgentSession {
 				this.#resetMnemopiConversationTrackingIfMnemopi();
 			}
 			this.#reconnectToAgent();
+			try {
+				await this.#sessionSwitchReconciler?.();
+			} catch (error) {
+				logger.warn("Failed to reconcile session mode after switch", {
+					targetSessionFile: sessionPath,
+					error: String(error),
+				});
+			}
 			return true;
 		} catch (error) {
 			this.sessionManager.restoreState(previousSessionState);

package/src/session/history-storage.ts

@@ -209,10 +209,6 @@ CREATE TRIGGER IF NOT EXISTS history_ai AFTER INSERT ON history BEGIN
 			logger.debug("HistoryStorage FTS query failed, using substring only", { error: String(error) });
 		}
 
-		if (ftsRows.length >= safeLimit) {
-			return ftsRows.map(row => this.#toEntry(row));
-		}
-
 		// 2. Substring fallback (token-AND LIKE). Catches infix matches FTS5's
 		//    prefix-only wildcard cannot reach (e.g. "mit" -> "commit"). Bounded
 		//    by safeLimit, ordered by recency - no full-table load into JS.
@@ -227,27 +223,24 @@ CREATE TRIGGER IF NOT EXISTS history_ai AFTER INSERT ON history BEGIN
 			return subRows.map(row => this.#toEntry(row));
 		}
 
-		const seen = new Set<number>();
-		const merged: HistoryEntry[] = [];
+		const rowsById = new Map<number, HistoryRow>();
 		for (const row of ftsRows) {
-			if (seen.has(row.id)) continue;
-			seen.add(row.id);
-			merged.push(this.#toEntry(row));
+			rowsById.set(row.id, row);
 		}
 		for (const row of subRows) {
-			if (merged.length >= safeLimit) break;
-			if (seen.has(row.id)) continue;
-			seen.add(row.id);
-			merged.push(this.#toEntry(row));
+			if (!rowsById.has(row.id)) rowsById.set(row.id, row);
 		}
-		return merged;
+
+		return [...rowsById.values()]
+			.sort((a, b) => b.created_at - a.created_at || b.id - a.id)
+			.slice(0, safeLimit)
+			.map(row => this.#toEntry(row));
 	}
 
 	/**
-	 * IDs of the sessions whose stored prompts match `query`, ordered by match
-	 * relevance (most relevant/recent first) and de-duplicated. Prompts with no
-	 * recorded session are skipped. Used to augment session ranking in the
-	 * resume picker with prompts that the 4KB session-list prefix never sees.
+	 * IDs of the sessions whose stored prompts match `query`, ordered by prompt
+	 * recency and de-duplicated. Used to augment session ranking in the resume
+	 * picker with prompts that the 4KB session-list prefix never sees.
 	 */
 	matchingSessionIds(query: string, limit = 500): string[] {
 		const seen = new Set<string>();