bug: Kimi Code OAuth shows incorrect Max Output (32K instead of 131K)

[myk1yt]

Bug Description

When selecting Settings > Providers > API Provider > Kimi Code > OAuth, the Max Output field shows 32,768 for all Kimi Code models. According to Kimi's official documentation, the correct values are:

• kimi-k3 (Kimi Code model ID: k3): max_completion_tokens defaults to 131,072, context window 1,048,576 (1M)
• kimi-k2.7-code (Kimi Code model ID: kimi-for-coding): max output = 256K - prompt_tokens, context window 262,144 (256K)

The Context Window displays correctly (values come from the server's /models API context_length field), but Max Output does not.

Root Cause

Three issues in the codebase:

1. Hardcoded default: kimiCodeDefaultModelInfo.maxTokens is set to 32_768 instead of 131_072
2. Missing schema field: The Zod schema in kimiCodeModelSchema doesn't include a max_tokens field, so even if the server returns it, it gets silently stripped
3. No override logic: mapKimiCodeModel() overrides contextWindow from the server response but does NOT override maxTokens

Impact

All Kimi Code models (kimi-for-coding, k3, k3-256k, kimi-for-coding-highspeed) display incorrect Max Output values. This affects the UI display and potentially the max_tokens parameter sent to the API.

Fix

A fix is ready in branch fix/kimi-code (commit 69f3ce3a8) with the following changes:

• Update kimiCodeDefaultModelInfo.maxTokens from 32_768 to 131_072
• Add kimiCodeModelDefaults for per-model fallback values
• Add max_tokens to kimiCodeModelSchema for server response parsing
• Override maxTokens in mapKimiCodeModel(): server value > model default > global default
• Add 4 new tests covering all fallback paths