From ff9fefb79beac75e9a257aa43fce39db36fa828e Mon Sep 17 00:00:00 2001
From: Peter Steinberger <steipete@gmail.com>
Date: Sun, 26 Apr 2026 23:16:00 +0100
Subject: [PATCH] fix(agents): validate thinking with model catalog

---
 src/agents/agent-command.ts     | 14 +++++++++++--
 src/auto-reply/thinking.test.ts | 35 +++++++++++++++++++++++++++++++++
 src/auto-reply/thinking.ts      | 19 ++++++++++++++----
 3 files changed, 62 insertions(+), 6 deletions(-)

diff --git a/src/agents/agent-command.ts b/src/agents/agent-command.ts
index d9e57648577..af28a3d17dd 100644
--- a/src/agents/agent-command.ts
+++ b/src/agents/agent-command.ts
@@ -813,17 +813,27 @@ async function agentCommandInternal(
         catalog: catalogForThinking.length > 0 ? catalogForThinking : undefined,
       });
     }
-    if (!isThinkingLevelSupported({ provider, model, level: resolvedThinkLevel })) {
+    const catalogForThinking = modelCatalog ?? allowedModelCatalog;
+    const thinkingCatalog = catalogForThinking.length > 0 ? catalogForThinking : undefined;
+    if (
+      !isThinkingLevelSupported({
+        provider,
+        model,
+        level: resolvedThinkLevel,
+        catalog: thinkingCatalog,
+      })
+    ) {
       const explicitThink = Boolean(thinkOnce || thinkOverride);
       if (explicitThink) {
         throw new Error(
-          `Thinking level "${resolvedThinkLevel}" is not supported for ${provider}/${model}. Use one of: ${formatThinkingLevels(provider, model)}.`,
+          `Thinking level "${resolvedThinkLevel}" is not supported for ${provider}/${model}. Use one of: ${formatThinkingLevels(provider, model, ", ", thinkingCatalog)}.`,
         );
       }
       const fallbackThinkLevel = resolveSupportedThinkingLevel({
         provider,
         model,
         level: resolvedThinkLevel,
+        catalog: thinkingCatalog,
       });
       if (fallbackThinkLevel !== resolvedThinkLevel) {
         const previousThinkLevel = resolvedThinkLevel;
diff --git a/src/auto-reply/thinking.test.ts b/src/auto-reply/thinking.test.ts
index 7ab99adf885..ed32e6de8e8 100644
--- a/src/auto-reply/thinking.test.ts
+++ b/src/auto-reply/thinking.test.ts
@@ -12,6 +12,8 @@ let listThinkingLevelOptions: typeof import("./thinking.js").listThinkingLevelOp
 let listThinkingLevels: typeof import("./thinking.js").listThinkingLevels;
 let normalizeReasoningLevel: typeof import("./thinking.js").normalizeReasoningLevel;
 let normalizeThinkLevel: typeof import("./thinking.js").normalizeThinkLevel;
+let isThinkingLevelSupported: typeof import("./thinking.js").isThinkingLevelSupported;
+let formatThinkingLevels: typeof import("./thinking.js").formatThinkingLevels;
 let resolveSupportedThinkingLevel: typeof import("./thinking.js").resolveSupportedThinkingLevel;
 let resolveThinkingDefaultForModel: typeof import("./thinking.js").resolveThinkingDefaultForModel;
 
@@ -42,6 +44,8 @@ beforeEach(async () => {
     listThinkingLevels,
     normalizeReasoningLevel,
     normalizeThinkLevel,
+    isThinkingLevelSupported,
+    formatThinkingLevels,
     resolveSupportedThinkingLevel,
     resolveThinkingDefaultForModel,
   } = await loadFreshThinkingModuleForTest());
@@ -170,6 +174,37 @@ describe("listThinkingLevels", () => {
     expect(listThinkingLevelLabels("demo", "demo-model")).toEqual(["off", "on"]);
   });
 
+  it("passes catalog reasoning into provider thinking profiles for support checks", () => {
+    providerRuntimeMocks.resolveProviderThinkingProfile.mockImplementation(({ context }) => ({
+      levels:
+        context.reasoning === true
+          ? [{ id: "off" }, { id: "low" }, { id: "medium" }, { id: "high" }, { id: "max" }]
+          : [{ id: "off" }],
+      defaultLevel: "off",
+    }));
+    const catalog = [{ provider: "ollama", id: "gpt-oss:20b", name: "gpt-oss", reasoning: true }];
+
+    expect(
+      isThinkingLevelSupported({
+        provider: "ollama",
+        model: "gpt-oss:20b",
+        level: "max",
+        catalog,
+      }),
+    ).toBe(true);
+    expect(formatThinkingLevels("ollama", "gpt-oss:20b", ", ", catalog)).toBe(
+      "off, low, medium, high, max",
+    );
+    expect(
+      resolveSupportedThinkingLevel({
+        provider: "ollama",
+        model: "gpt-oss:20b",
+        level: "max",
+        catalog,
+      }),
+    ).toBe("max");
+  });
+
   it("maps stale unsupported levels to the largest profile level", () => {
     providerRuntimeMocks.resolveProviderThinkingProfile.mockReturnValue({
       levels: [{ id: "off" }, { id: "high" }],
diff --git a/src/auto-reply/thinking.ts b/src/auto-reply/thinking.ts
index e0abe1bac52..7d1aec14951 100644
--- a/src/auto-reply/thinking.ts
+++ b/src/auto-reply/thinking.ts
@@ -194,8 +194,11 @@ function supportsThinkingLevel(
   provider: string | null | undefined,
   model: string | null | undefined,
   level: ThinkLevel,
+  catalog?: ThinkingCatalogEntry[],
 ): boolean {
-  return resolveThinkingProfile({ provider, model }).levels.some((entry) => entry.id === level);
+  return resolveThinkingProfile({ provider, model, catalog }).levels.some(
+    (entry) => entry.id === level,
+  );
 }
 
 export function supportsXHighThinking(provider?: string | null, model?: string | null): boolean {
@@ -223,8 +226,10 @@ export function formatThinkingLevels(
   provider?: string | null,
   model?: string | null,
   separator = ", ",
+  catalog?: ThinkingCatalogEntry[],
 ): string {
-  return listThinkingLevelLabels(provider, model).join(separator);
+  const profile = resolveThinkingProfile({ provider, model, catalog });
+  return profile.levels.map(({ label }) => label).join(separator);
 }
 
 export function resolveThinkingDefaultForModel(params: {
@@ -262,8 +267,9 @@ export function isThinkingLevelSupported(params: {
   provider?: string | null;
   model?: string | null;
   level: ThinkLevel;
+  catalog?: ThinkingCatalogEntry[];
 }): boolean {
-  return supportsThinkingLevel(params.provider, params.model, params.level);
+  return supportsThinkingLevel(params.provider, params.model, params.level, params.catalog);
 }
 
 function resolveSupportedThinkingLevelFromProfile(
@@ -286,7 +292,12 @@ export function resolveSupportedThinkingLevel(params: {
   provider?: string | null;
   model?: string | null;
   level: ThinkLevel;
+  catalog?: ThinkingCatalogEntry[];
 }): ThinkLevel {
-  const profile = resolveThinkingProfile({ provider: params.provider, model: params.model });
+  const profile = resolveThinkingProfile({
+    provider: params.provider,
+    model: params.model,
+    catalog: params.catalog,
+  });
   return resolveSupportedThinkingLevelFromProfile(profile, params.level);
 }