From 5337f9243a1dbb33dc6fe8b5d5da808bd410af9c Mon Sep 17 00:00:00 2001
From: Rajeev91691 <rajeevnandan382@gmail.com>
Date: Mon, 20 Jul 2026 21:20:25 +0530
Subject: [PATCH] fix(ai-chat): respect max_tokens parameter and model limit in
 MoonshotProvider

Fixes #3408 by destructuring max_tokens from ICompleteArguments in MoonshotProvider.complete() and passing max_tokens || modelUsed.max_tokens || 1000 instead of hardcoding 1000. This ensures thinking models such as Kimi K3 have sufficient output token budget and do not hit early length truncation.
---
 .../moonshot/MoonshotProvider.test.ts         | 23 +++++++++++++++----
 .../providers/moonshot/MoonshotProvider.ts    |  3 ++-
 2 files changed, 21 insertions(+), 5 deletions(-)

diff --git a/src/backend/drivers/ai-chat/providers/moonshot/MoonshotProvider.test.ts b/src/backend/drivers/ai-chat/providers/moonshot/MoonshotProvider.test.ts
index 9ae61ace62..4f51b7744f 100644
--- a/src/backend/drivers/ai-chat/providers/moonshot/MoonshotProvider.test.ts
+++ b/src/backend/drivers/ai-chat/providers/moonshot/MoonshotProvider.test.ts
@@ -202,7 +202,7 @@ describe('MoonshotProvider.complete request shape', () => {
         usage: { prompt_tokens: 1, completion_tokens: 1 },
     };
 
-    it('forwards model + messages and locks max_tokens=1000', async () => {
+    it('forwards model + messages and passes max_tokens or falls back to model max_tokens', async () => {
         const { provider } = makeProvider();
         createMock.mockResolvedValueOnce(baseCompletion);
 
@@ -216,9 +216,24 @@ describe('MoonshotProvider.complete request shape', () => {
         const [args] = createMock.mock.calls[0]!;
         expect(args.model).toBe('kimi-k2.6');
         expect(args.messages).toEqual([{ role: 'user', content: 'hello' }]);
-        // max_tokens is hardcoded by the provider — the call to Moonshot
-        // should always cap at 1000 tokens of completion.
-        expect(args.max_tokens).toBe(1000);
+        expect(args.max_tokens).toBe(262_144);
+    });
+
+    it('passes custom max_tokens when provided in arguments', async () => {
+        const { provider } = makeProvider();
+        createMock.mockResolvedValueOnce(baseCompletion);
+
+        await withTestActor(() =>
+            provider.complete({
+                model: 'kimi-k3',
+                messages: [{ role: 'user', content: 'hello' }],
+                max_tokens: 4096,
+            }),
+        );
+
+        const [args] = createMock.mock.calls[0]!;
+        expect(args.model).toBe('kimi-k3');
+        expect(args.max_tokens).toBe(4096);
     });
 
     it('omits the `tools` key entirely when no tools are supplied', async () => {
diff --git a/src/backend/drivers/ai-chat/providers/moonshot/MoonshotProvider.ts b/src/backend/drivers/ai-chat/providers/moonshot/MoonshotProvider.ts
index 079a74601f..e6de47db8a 100644
--- a/src/backend/drivers/ai-chat/providers/moonshot/MoonshotProvider.ts
+++ b/src/backend/drivers/ai-chat/providers/moonshot/MoonshotProvider.ts
@@ -68,6 +68,7 @@ export class MoonshotProvider implements IChatProvider {
         stream,
         model,
         tools,
+        max_tokens,
     }: ICompleteArguments): Promise<IChatCompleteResult> {
         const actor = Context.get('actor');
         const availableModels = this.models();
@@ -89,7 +90,7 @@ export class MoonshotProvider implements IChatProvider {
                 messages,
                 model: modelUsed.id,
                 ...(tools ? { tools } : {}),
-                max_tokens: 1000,
+                max_tokens: max_tokens || modelUsed.max_tokens || 1000,
                 stream,
                 ...(stream
                     ? {