接入火山引擎的 KIMI K2.5 报错

[Greenplumwine]

What version of Kimi Code CLI is running?

kimi, version 1.12.0

Which open platform/subscription were you using?

/setup

Which model were you using?

火山引擎的 kimi k2.5

What platform is your computer?

Darwin 25.2.0 arm64 arm

What issue are you seeing?

使用 接入火山引擎的 kimi-k2.5 一直抛出
The parameter max_tokensspecified in the request is not valid: integer above maximum value, expected a value <= 32768, but got 50000 instead.
修改 max_context_size 和 reserved_context_size 都不行

What steps can reproduce the bug?

[models."volcengine-codeing-plan/ark-code-latest"]
provider = "volcengine-codeing-plan"
model = "ark-code-latest"
max_context_size = 65536
capabilities = ["video_in", "image_in", "thinking"]

[providers.volcengine-codeing-plan]
type = "anthropic"
base_url = "https://ark.cn-beijing.volces.com/api/coding"
api_key = "xxxx"

What is the expected behavior?

No response

Additional information

No response

[xiaoju111a]

try this
[models."volcengine-codeing-plan/ark-code-latest"]
provider = "volcengine-codeing-plan"
model = "ark-code-latest"
max_context_size = 65536
max_output_tokens = 32768 # 添加这一行
capabilities = ["video_in", "image_in", "thinking"]

[Greenplumwine]

max_output_tokens = 32768

LLM provider error: Error code: 400 - {'error': {'code': 'InvalidParameter', 'message': 'The parameter max_to kens specified in the request is not valid: integer above maximum value, expected a value <= 32768, but got 5
0000 instead. Request id: 0217710335433517d66892265258d8a5511543a67713e2f915196', 'param': 'max_tokens', 'type
': 'BadRequest'}}

加上后还是有这个问题

[wdvxdr1123]

接入火山使用openai_legacy会有thinking被丢失的问题， kimi官方api和火山都是类openai， 对于 thinking 部分的处理是一样的， 可以使用这样配置， 实测可以开启thinking

[providers.volcengine]
type = "kimi"
base_url = "https://ark.cn-beijing.volces.com/api/coding/v3"
api_key = "YOUR_TOKEN"