Merge pull request #459 from opticon454/feature/run-menu-picker-currently-loaded-model

feat(custom-model): promote the currently-loaded/last-used model in the Run-menu picker

docs/custom-model-endpoints.md

@@ -183,6 +183,28 @@ launch, with the endpoint's `defaultModelId` marked but not auto-chosen —
 the point of asking is letting one launch deliberately differ from the
 saved default, not just confirming it.
 
+The modal promotes exactly one row to the top of the list rather than
+always showing raw discovery order, so the zero-wait choice is the one
+under your thumb:
+
+- **"Currently loaded"** — a model from this host's own list that
+  llama-swap reports `ready` right now, queried via
+  `GET /api/model-endpoints/:id/running-status`. Bounded client-side to
+  ~800ms (`Promise.race`), on top of the route's own 5s server-side
+  timeout, so an endpoint that is asleep or firewalled cannot leave the
+  modal invisible for the full 5s after the Run menu has already closed.
+- **"Last used"** — shown only when nothing is currently loaded: the model
+  actually launched last for this exact (harness, endpoint) pair, read
+  from the per-device `codeman:customModelLastUsed:<mode>:<endpointId>`
+  localStorage key. Written by `runCustomModelEntry` /
+  `_quickStartWithCustomModelConfirm` only once the model is actually
+  applied, never on the mere click — declining the context-window warning
+  means this exact model cannot work with this CLI at all, so promoting it
+  next time would be actively wrong, not just premature.
+
+Neither tag reorders anything past that one promoted row, and both defer
+to "Default" when neither applies.
+
 **How the launch itself applies the endpoint depends on the harness.** For
 opencode, Codex, Gemini, Pi, Grok, DeepSeek and OMP (`runCustomModelEntry` →
 `_runCustomModelEntryOneShot`), the endpoint/model is folded into the SAME

src/web/public/i18n.js

@@ -312,6 +312,8 @@
       '运行菜单选择器会为此端点应用该模型。请先发现可用模型。',
     'Custom Endpoints': '自定义端点',
     'Choose a model': '选择模型',
+    'Currently loaded': '当前已加载',
+    'Last used': '上次使用',
     'That endpoint no longer exists': '该端点已不存在',
     'No models discovered for this endpoint yet': '此端点尚未发现任何模型',
     'Subagent Options': '子智能体选项',

src/web/public/session-ui.js

@@ -720,14 +720,101 @@ Object.assign(CodemanApp.prototype, {
     if (models.length === 1) {
       return this.runCustomModelEntry(mode, endpointId, models[0]);
     }
-    this._openCustomModelPickModal(mode, host);
+    await this._openCustomModelPickModal(mode, host);
   },
 
-  /** Renders the "which model" picker for a (harness, endpoint) pair with more than one discovered model. */
+  /** localStorage key for the last model launched on a given (harness, endpoint) pair — per-device by design, like every other `codeman:*` UI preference, never synced. */
-  _openCustomModelPickModal(mode, host) {
+  _customModelLastUsedKey(mode, endpointId) {
+    return `codeman:customModelLastUsed:${mode}:${endpointId}`;
+  },
+
+  /** Reads the last model chosen for this (harness, endpoint) pair, or null. Never throws — a blocked/full localStorage just means no promotion, not a broken picker. */
+  _getCustomModelLastUsed(mode, endpointId) {
+    try {
+      return localStorage.getItem(this._customModelLastUsedKey(mode, endpointId));
+    } catch {
+      return null;
+    }
+  },
+
+  /** Remembers `modelId` as the last one launched for this (harness, endpoint) pair. */
+  _setCustomModelLastUsed(mode, endpointId, modelId) {
+    try {
+      localStorage.setItem(this._customModelLastUsedKey(mode, endpointId), modelId);
+    } catch {
+      // best-effort — losing the "last used" hint is cosmetic, never worth surfacing
+    }
+  },
+
+  /**
+   * Best-effort lookup of the model llama-swap currently has loaded and ready on this
+   * endpoint, so the picker can offer it first instead of making the user remember what
+   * they picked last time it mattered. Mirrors `_watchLlamaSwapLoading`'s own
+   * `state === 'ready'` check. Returns null for a plain (non-llama-swap) server, an
+   * unreachable endpoint, or a loaded model this host no longer lists as discovered —
+   * never throws, since a failed probe should just skip promotion, not break the picker.
+   *
+   * ⚠️ Client-side bounded to ~800ms via Promise.race, on top of (never instead of) the
+   * route's own 5s server-side timeout (`RUNNING_TIMEOUT_MS`, custom-model-routes.ts) —
+   * a saved endpoint keeps its discovered models cached, so "the box behind this endpoint
+   * is asleep or firewalled" is a normal way to reach this path, not an exotic one, and
+   * the modal must not sit invisible (Run menu already closed, nothing else on screen)
+   * for the full 5s a slow/dead endpoint can take. The losing side of the race is left to
+   * resolve on its own — `.catch(() => null)` only stops an unhandled-rejection warning
+   * when it eventually fails, it never cancels the in-flight fetch.
+   *
+   * `timeoutMs` exists to let a test drive this in milliseconds instead of the real
+   * 800 — same reasoning as `_watchLlamaSwapLoading`'s own `pollIntervalMs`: this code
+   * runs inside a JSDOM window's own realm, whose `setTimeout` is not the one
+   * `vi.useFakeTimers()` patches, so a param is the only way to test the timeout without
+   * actually waiting on it. Real callers never pass it.
+   */
+  async _getCustomModelCurrentlyLoaded(host, timeoutMs = 800) {
+    const probe = this._apiJson(`/api/model-endpoints/${encodeURIComponent(host.id)}/running-status`).catch(
+      () => null
+    );
+    const timeout = new Promise((resolve) => setTimeout(() => resolve(null), timeoutMs));
+    const status = await Promise.race([probe, timeout]);
+    if (!status?.isLlamaSwap) return null;
+    const ready = (status.running || []).find((r) => r.state === 'ready' && (host.models || []).includes(r.model));
+    return ready?.model || null;
+  },
+
+  /**
+   * Renders the "which model" picker for a (harness, endpoint) pair with more than one
+   * discovered model. Async since it now awaits the currently-loaded-model probe below,
+   * so a SECOND call (a different custom-model entry clicked while the first one's probe
+   * is still in flight — the probe has its own 5s timeout) must not let the first call's
+   * later-arriving response clobber the second's already-rendered, already-correct modal.
+   * `_customModelPickGeneration` is the same guard-a-mutable-counter pattern
+   * `_watchLlamaSwapLoading` uses for the same reason: every DOM write below, including
+   * `_pendingCustomModelPick` itself, stays deferred until after the await, and a call
+   * that finds a newer generation already claimed bails out untouched rather than only
+   * skipping the model-list write and leaving title/hint/`_pendingCustomModelPick`
+   * inconsistent with what's on screen.
+   */
+  async _openCustomModelPickModal(mode, host) {
     const modal = document.getElementById('customModelPickModal');
     const list = document.getElementById('customModelPickList');
     if (!modal || !list) return;
+    const generation = (this._customModelPickGeneration = (this._customModelPickGeneration || 0) + 1);
+    const isCurrent = () => this._customModelPickGeneration === generation;
+
+    // Whichever model llama-swap actually has loaded right now beats a merely
+    // remembered choice — it's what a launch would attach to with zero wait, while
+    // "last used" might have been swapped out by another session since. Neither
+    // reorders past the top: exactly one model is promoted, everything else keeps
+    // its discovery order.
+    const currentlyLoaded = await this._getCustomModelCurrentlyLoaded(host);
+    if (!isCurrent()) return; // a newer pick opened (and possibly already rendered) while this probe was in flight
+    const lastUsed = currentlyLoaded ? null : this._getCustomModelLastUsed(mode, host.id);
+    const promoted = currentlyLoaded || lastUsed;
+    const models = [...(host.models || [])];
+    if (promoted && models.includes(promoted)) {
+      models.splice(models.indexOf(promoted), 1);
+      models.unshift(promoted);
+    }
+
     this._pendingCustomModelPick = { mode, endpointId: host.id };
     const cliLabel = (window.__codemanCustomModelClis || []).find((c) => c.id === mode)?.label || mode;
     // A static title (translatable by i18n.js's exact-string walker) plus a
@@ -736,13 +823,14 @@ Object.assign(CodemanApp.prototype, {
     document.getElementById('customModelPickTitle').textContent = 'Choose a model';
     document.getElementById('customModelPickHint').textContent =
       `${cliLabel} → ${host.label} — ${(host.models || []).length} models discovered.`;
-    list.innerHTML = (host.models || [])
+    list.innerHTML = models
       .map((m) => {
         const isDefault = m === host.defaultModelId;
+        const tag = m === currentlyLoaded ? 'Currently loaded' : m === lastUsed ? 'Last used' : isDefault ? 'Default' : null;
         const arg = escapeHtml(JSON.stringify(m));
         return `
           <button class="run-mode-option" onclick="app.chooseCustomModelAndRun(${arg})">
-            <span class="run-mode-dot ${escapeHtml(mode)}"></span>${escapeHtml(m)}${isDefault ? ' <span class="set-scope">Default</span>' : ''}
+            <span class="run-mode-dot ${escapeHtml(mode)}"></span>${escapeHtml(m)}${tag ? ` <span class="set-scope">${escapeHtml(tag)}</span>` : ''}
           </button>`;
       })
       .join('');
@@ -858,6 +946,12 @@ Object.assign(CodemanApp.prototype, {
    * (Codex, confirmed live) than on claude's own `--resume`-based restart.
    */
   async runCustomModelEntry(mode, endpointId, modelId) {
+    // "Last used" is recorded by each path itself, ONLY once the model is actually
+    // applied — never here, unconditionally, on the mere attempt. A context-window
+    // warning or a swap-conflict question can still say no after this call, and the
+    // context-warning case is the one that bites: declining it means this exact
+    // model cannot work with this CLI at all, so promoting it as "Last used" next
+    // time the picker opens would be actively wrong, not just premature.
     if (mode === 'claude') {
       return this._runCustomModelEntryViaRestart(mode, endpointId, modelId);
     }
@@ -950,7 +1044,15 @@ Object.assign(CodemanApp.prototype, {
       answered = { ...answered, confirmedSwap: true };
       data = await post({ ...bodyObj, customModel: { ...bodyObj.customModel, ...answered } });
     }
-    this._lastCustomModelLaunchResult = data?.success !== false ? data?.data : undefined;
+    const launched = data?.success !== false;
+    this._lastCustomModelLaunchResult = launched ? data?.data : undefined;
+    // Only once actually launched, and only for a call that carried a custom-model pick
+    // at all — `_launchQuickStartInstances` runs every quick-start body (custom-model or
+    // not) through this same function, so a plain launch must not fall through here with
+    // an undefined endpointId/modelId that quietly no-ops the (mode, endpointId) key.
+    if (launched && bodyObj.customModel) {
+      this._setCustomModelLastUsed(bodyObj.mode, bodyObj.customModel.endpointId, bodyObj.customModel.modelId);
+    }
     return data;
   },
 
@@ -1078,6 +1180,11 @@ Object.assign(CodemanApp.prototype, {
       return;
     }
 
+    // The apply has actually succeeded and both questions (if asked) are answered
+    // yes — only now is this a real "last used" for the picker's next open, not
+    // before either confirmation had a chance to decline it.
+    this._setCustomModelLastUsed(mode, endpointId, modelId);
+
     // The apply above already succeeded — the session IS pointed at the endpoint — but
     // llama-swap itself may still be unloading the old model and loading this one, which
     // can take well over a minute. Without this, a prompt sent during that window either

test/custom-model-one-shot-launch.test.ts

@@ -142,7 +142,7 @@ describe('_quickStartWithCustomModelConfirm', () => {
     })) as unknown as typeof fetch;
   }
 
-  it('returns the response directly when no confirmation is needed', async () => {
+  it('returns the response directly when no confirmation is needed, and records "last used"', async () => {
     const { win, app } = bootApp();
     withFetch(win, (body) => ({ success: true, data: { sessionId: 's1', modelSwapInProgress: false, body } }));
     const data = await app._quickStartWithCustomModelConfirm({
@@ -152,9 +152,17 @@ describe('_quickStartWithCustomModelConfirm', () => {
     expect(data.success).toBe(true);
     expect(data.data.sessionId).toBe('s1');
     expect(app._lastCustomModelLaunchResult).toEqual(data.data);
+    expect(win.localStorage.getItem('codeman:customModelLastUsed:codex:e')).toBe('m');
   });
 
-  it('confirming re-sends with confirmedSwap and returns the second response', async () => {
+  it('a plain launch with no customModel at all never touches the "last used" key (undefined endpointId/modelId would otherwise silently no-op it)', async () => {
+    const { win, app } = bootApp();
+    withFetch(win, () => ({ success: true, data: { sessionId: 's1' } }));
+    await app._quickStartWithCustomModelConfirm({ mode: 'codex' });
+    expect(win.localStorage.getItem('codeman:customModelLastUsed:codex:undefined')).toBeNull();
+  });
+
+  it('confirming re-sends with confirmedSwap, returns the second response, and only THEN records "last used"', async () => {
     const { win, app } = bootApp();
     app._confirmModelSwap = async () => true;
     let calls = 0;
@@ -183,9 +191,10 @@ describe('_quickStartWithCustomModelConfirm', () => {
     expect(calls).toBe(2);
     expect(data.data.sessionId).toBe('s1');
     expect(app._lastCustomModelLaunchResult.modelSwapInProgress).toBe(true);
+    expect(win.localStorage.getItem('codeman:customModelLastUsed:codex:e')).toBe('m');
   });
 
-  it('cancelling never re-sends, and reports a cancellation error', async () => {
+  it('cancelling never re-sends, reports a cancellation error, and must NEVER record "last used" for a launch that never happened', async () => {
     const { win, app } = bootApp();
     app._confirmModelSwap = async () => false;
     let calls = 0;
@@ -208,6 +217,7 @@ describe('_quickStartWithCustomModelConfirm', () => {
     expect(data.success).toBe(false);
     expect(data.error).toMatch(/cancelled/i);
     expect(app._lastCustomModelLaunchResult).toBeUndefined();
+    expect(win.localStorage.getItem('codeman:customModelLastUsed:codex:e')).toBeNull();
   });
 });
 

test/custom-model-run-menu-ui.test.ts

@@ -272,6 +272,146 @@ describe('Custom Model Endpoint Profiles: the "which model" picker', () => {
     expect(win.document.getElementById('customModelPickList')!.textContent).toContain('Default');
   });
 
+  it('promotes the model llama-swap currently has loaded and ready to the top of the list, tagged', async () => {
+    const { win, app } = bootApp({
+      hosts: [{ id: 'llama-box', label: 'llama.cpp', baseUrl: 'http://x', models: ['qwen3', 'llama3', 'phi4'] }],
+    });
+    const origApiJson = app._apiJson;
+    app._apiJson = async (path: string) => {
+      if (path === '/api/model-endpoints/llama-box/running-status') {
+        return { isLlamaSwap: true, running: [{ model: 'phi4', state: 'ready' }] };
+      }
+      return origApiJson(path);
+    };
+
+    await app.selectCustomModelEntry('claude', 'llama-box');
+
+    const buttons = [...win.document.getElementById('customModelPickList')!.querySelectorAll('button')];
+    expect(buttons.map((b) => b.textContent)).toHaveLength(3);
+    expect(buttons[0].textContent).toContain('phi4');
+    expect(buttons[0].textContent).toContain('Currently loaded');
+    // Nothing else got relabelled or reordered past the promoted row.
+    expect(buttons[1].textContent).toContain('qwen3');
+    expect(buttons[2].textContent).toContain('llama3');
+  });
+
+  it('falls back to the last model launched on this (harness, endpoint) pair when nothing is currently loaded', async () => {
+    const { win, app } = bootApp({
+      hosts: [{ id: 'llama-box', label: 'llama.cpp', baseUrl: 'http://x', models: ['qwen3', 'llama3', 'phi4'] }],
+    });
+    const origApiJson = app._apiJson;
+    app._apiJson = async (path: string) => {
+      if (path === '/api/model-endpoints/llama-box/running-status') return { isLlamaSwap: false, running: [] };
+      return origApiJson(path);
+    };
+    // Simulate a prior launch on this exact (harness, endpoint) pair having picked llama3.
+    win.localStorage.setItem('codeman:customModelLastUsed:claude:llama-box', 'llama3');
+
+    await app.selectCustomModelEntry('claude', 'llama-box');
+
+    const buttons = [...win.document.getElementById('customModelPickList')!.querySelectorAll('button')];
+    expect(buttons[0].textContent).toContain('llama3');
+    expect(buttons[0].textContent).toContain('Last used');
+    expect(buttons[0].textContent).not.toContain('Currently loaded');
+  });
+
+  it('prefers the currently-loaded model over a stale "last used" entry when both are present', async () => {
+    const { win, app } = bootApp({
+      hosts: [{ id: 'llama-box', label: 'llama.cpp', baseUrl: 'http://x', models: ['qwen3', 'llama3', 'phi4'] }],
+    });
+    const origApiJson = app._apiJson;
+    app._apiJson = async (path: string) => {
+      if (path === '/api/model-endpoints/llama-box/running-status') {
+        return { isLlamaSwap: true, running: [{ model: 'phi4', state: 'ready' }] };
+      }
+      return origApiJson(path);
+    };
+    win.localStorage.setItem('codeman:customModelLastUsed:claude:llama-box', 'llama3');
+
+    await app.selectCustomModelEntry('claude', 'llama-box');
+
+    const buttons = [...win.document.getElementById('customModelPickList')!.querySelectorAll('button')];
+    expect(buttons[0].textContent).toContain('phi4');
+    expect(buttons[0].textContent).toContain('Currently loaded');
+  });
+
+  it('is not fooled by a model llama-swap reports loaded but not yet ready, or one this host no longer lists', async () => {
+    const { win, app } = bootApp({
+      hosts: [{ id: 'llama-box', label: 'llama.cpp', baseUrl: 'http://x', models: ['qwen3', 'llama3'] }],
+    });
+    const origApiJson = app._apiJson;
+    app._apiJson = async (path: string) => {
+      if (path === '/api/model-endpoints/llama-box/running-status') {
+        // "loading", not "ready" — and a model id this host's own /v1/models no longer serves.
+        return { isLlamaSwap: true, running: [{ model: 'ghost-model', state: 'loading' }] };
+      }
+      return origApiJson(path);
+    };
+
+    await app.selectCustomModelEntry('claude', 'llama-box');
+
+    const list = win.document.getElementById('customModelPickList')!;
+    expect(list.textContent).not.toContain('Currently loaded');
+    expect(list.textContent).not.toContain('ghost-model');
+    const buttons = [...list.querySelectorAll('button')];
+    expect(buttons[0].textContent).toContain('qwen3');
+  });
+
+  it('remembers the launched model as "last used" only once the apply actually succeeds, not on the mere attempt', async () => {
+    const { win, app } = bootApp({});
+    app.activeSessionId = 'old-session';
+    // A real new session, and a real successful apply with no questions asked — the
+    // restart path only reaches its _setCustomModelLastUsed call past both.
+    app.run = async () => {
+      app.activeSessionId = 'new-session';
+    };
+    app._api = async () => ({
+      ok: true,
+      status: 200,
+      json: async () => ({ success: true, data: { customModel: { endpointId: 'llama-box' }, restarted: true } }),
+    });
+
+    await app.runCustomModelEntry('claude', 'llama-box', 'qwen3');
+
+    expect(win.localStorage.getItem('codeman:customModelLastUsed:claude:llama-box')).toBe('qwen3');
+  });
+
+  it('a slower currently-loaded probe for an earlier pick must never clobber a faster, later pick for a different endpoint', async () => {
+    const { win, app } = bootApp({});
+    const hostA = { id: 'host-a', label: 'Host A', baseUrl: 'http://a', models: ['a1', 'a2'] };
+    const hostB = { id: 'host-b', label: 'Host B', baseUrl: 'http://b', models: ['b1', 'b2'] };
+    let resolveA!: (v: unknown) => void;
+    const pendingA = new Promise((resolve) => {
+      resolveA = resolve;
+    });
+    app._apiJson = async (path: string) => {
+      if (path === '/api/model-endpoints/host-a/running-status') return pendingA;
+      if (path === '/api/model-endpoints/host-b/running-status') return { isLlamaSwap: false, running: [] };
+      return null;
+    };
+
+    // Host A's picker opens first but its probe never resolves until we say so below —
+    // Host B's opens second and resolves immediately, so it renders first.
+    const openA = app._openCustomModelPickModal('claude', hostA);
+    await app._openCustomModelPickModal('claude', hostB);
+
+    expect(win.document.getElementById('customModelPickHint')!.textContent).toContain('Host B');
+    expect(app._pendingCustomModelPick).toEqual({ mode: 'claude', endpointId: 'host-b' });
+
+    // Host A's probe finally answers, after Host B has already rendered.
+    resolveA({ isLlamaSwap: false, running: [] });
+    await openA;
+
+    // The late-arriving Host A response must be a no-op: still Host B on screen.
+    expect(win.document.getElementById('customModelPickHint')!.textContent).toContain('Host B');
+    expect(app._pendingCustomModelPick).toEqual({ mode: 'claude', endpointId: 'host-b' });
+    const listText = win.document.getElementById('customModelPickList')!.textContent;
+    expect(listText).toContain('b1');
+    expect(listText).toContain('b2');
+    expect(listText).not.toContain('a1');
+    expect(listText).not.toContain('a2');
+  });
+
   it('picking a row in the modal closes it and launches with that exact model', async () => {
     const { win, app } = bootApp({
       hosts: [{ id: 'llama-box', label: 'llama.cpp', baseUrl: 'http://localhost:8080', models: ['qwen3', 'llama3'] }],
@@ -333,6 +473,48 @@ describe('Custom Model Endpoint Profiles: the "which model" picker', () => {
   });
 });
 
+describe('Custom Model Endpoint Profiles: _getCustomModelCurrentlyLoaded is client-side bounded', () => {
+  // `timeoutMs` driven in milliseconds rather than the real 800 — same reasoning as
+  // `_watchLlamaSwapLoading`'s own `pollIntervalMs` a few describe blocks down: this
+  // code runs inside the JSDOM window's own realm, whose setTimeout vi.useFakeTimers()
+  // does not patch, so this is the only way to test the bound without actually waiting
+  // on it (or, worse, hanging on a promise that deliberately never resolves).
+
+  it('never lets an endpoint that never answers keep the picker waiting past the client-side bound', async () => {
+    const { app } = bootApp({
+      hosts: [{ id: 'llama-box', label: 'llama.cpp', baseUrl: 'http://x', models: ['qwen3'] }],
+    });
+    // A `running-status` probe that simply never resolves — the exact shape of an
+    // endpoint that is asleep or firewalled, distinct from one that answers an error.
+    app._apiJson = () => new Promise(() => {});
+
+    const result = await app._getCustomModelCurrentlyLoaded({ id: 'llama-box', models: ['qwen3'] }, 5);
+
+    expect(result).toBeNull();
+  });
+
+  it('an endpoint that answers well within the bound is unaffected by it', async () => {
+    const { app } = bootApp({});
+    app._apiJson = async () => ({ isLlamaSwap: true, running: [{ model: 'qwen3', state: 'ready' }] });
+
+    const result = await app._getCustomModelCurrentlyLoaded({ id: 'llama-box', models: ['qwen3'] }, 5);
+
+    expect(result).toBe('qwen3');
+  });
+
+  it('a rejected probe settles quietly to null rather than leaving an unhandled rejection once the timeout has already won the race', async () => {
+    const { app } = bootApp({});
+    app._apiJson = () => new Promise((_resolve, reject) => setTimeout(() => reject(new Error('boom')), 10));
+
+    const result = await app._getCustomModelCurrentlyLoaded({ id: 'llama-box', models: ['qwen3'] }, 2);
+    expect(result).toBeNull();
+    // Give the loser of the race a turn to actually reject and hit its own .catch —
+    // an unswallowed rejection here would surface as an "Unhandled Errors" failure
+    // for the whole test file, not a failed assertion in this test.
+    await new Promise((resolve) => setTimeout(resolve, 20));
+  });
+});
+
 describe('Custom Model Endpoint Profiles: applying a picked entry', () => {
   it('does not apply the endpoint to a session that was already open when the launch fails', async () => {
     const { app } = bootApp({});
@@ -1025,8 +1207,8 @@ describe("Custom Model Endpoint Profiles: requiresContextWarning (this CLI's own
     return { win, app, applyBodies };
   }
 
-  it('confirming the in-app context-warning modal re-sends the apply with confirmed:true', async () => {
+  it('confirming the in-app context-warning modal re-sends the apply with confirmed:true, and only THEN records "last used"', async () => {
-    const { app, applyBodies } = launchHarness([
+    const { win, app, applyBodies } = launchHarness([
       { requiresContextWarning: true, modelId: 'qwen3', contextLength: 16384, minSafeContextTokens: 40000 },
       { customModel: { endpointId: 'llama-box' }, restarted: true, modelSwapInProgress: false },
     ]);
@@ -1043,10 +1225,11 @@ describe("Custom Model Endpoint Profiles: requiresContextWarning (this CLI's own
       { endpointId: 'llama-box', modelId: 'qwen3' },
       { endpointId: 'llama-box', modelId: 'qwen3', confirmedContext: true },
     ]);
+    expect(win.localStorage.getItem('codeman:customModelLastUsed:claude:llama-box')).toBe('qwen3');
   });
 
-  it('declining the in-app context-warning modal keeps the native backend and never re-sends the apply', async () => {
+  it('declining the in-app context-warning modal keeps the native backend, never re-sends the apply, and must NEVER record this model as "last used" — it cannot work with this CLI at all', async () => {
-    const { app, applyBodies } = launchHarness([
+    const { win, app, applyBodies } = launchHarness([
       { requiresContextWarning: true, modelId: 'qwen3', contextLength: 16384, minSafeContextTokens: 40000 },
     ]);
     app._confirmContextWarning = async () => false;
@@ -1059,6 +1242,7 @@ describe("Custom Model Endpoint Profiles: requiresContextWarning (this CLI's own
 
     expect(applyBodies).toHaveLength(1); // no second (confirmed) call
     expect(toastMessage).toMatch(/context window too small/i);
+    expect(win.localStorage.getItem('codeman:customModelLastUsed:claude:llama-box')).toBeNull();
   });
 });