Title: API Bug: invalid temperature: only 0.6 is allowed for this model returned for ALL temperature values (including 0.6)

What version of Kimi Code CLI is running?

hermes agent Not using Kimi CLI directly — the bug is at the API level. Tested with: - OpenAI Python SDK 2.32.0 - raw curl/urllib - Hermes Agent (latest)

Which open platform/subscription were you using?

Kimi Coding Plan (api.kimi.com/coding/v1)

Which model were you using?

kimi-for-coding Also tested: - kimi-k2.5 - kimi-k2-thinking - kimi-k2-turbo-preview - kimi-latest - kimi-k2

What platform is your computer?

Linux 6.8.0-107-generic x86_64 Bug is platform-independent — reproduced from: - Debian 13 container - Host Ubuntu VPS - Fresh container with only Python + openai SDK

What issue are you seeing?

Bug Description

The kimi-for-coding model on the Coding API endpoint returns 400 invalid temperature: only 0.6 is allowed for this model regardless of the temperature value sent, including the exact value 0.6 mentioned in the error message.
This makes the model completely unusable.

Environment

• Endpoint: https://api.kimi.com/coding/v1/chat/completions
• Model: kimi-for-coding
• Client: OpenAI Python SDK 2.32.0 / raw HTTP
• Header: User-Agent: KimiCLI/1.30.0
• API Key type: Kimi Coding Plan
• First observed: ~2026-04-17 16:49 UTC
• Status before: Working correctly earlier the same day

Reproduction

Test 1: temperature = 0.6 (exact value from error message)

curl -X POST https://api.kimi.com/coding/v1/chat/completions \
  -H "Authorization: Bearer $KIMI_API_KEY" \
  -H "Content-Type: application/json" \
  -H "User-Agent: KimiCLI/1.30.0" \
  -d '{"model":"kimi-for-coding","messages":[{"role":"user","content":"hi"}],"temperature":0.6}'
Result: 400 {"error":{"message":"invalid temperature: only 0.6 is allowed for this
model","type":"invalid_request_error"}}

Test 2: No temperature at all

curl -X POST https://api.kimi.com/coding/v1/chat/completions \
  -H "Authorization: Bearer $KIMI_API_KEY" \
  -H "Content-Type: application/json" \
  -H "User-Agent: KimiCLI/1.30.0" \
  -d '{"model":"kimi-for-coding","messages":[{"role":"user","content":"hi"}]}'
Result: Same 400 error

Test 3: temperature = 1.0

Result: Same 400 error

Test 4: temperature = 0.0

Result: Same 400 error

Test 5: Other model names (kimi-k2.5, kimi-k2-thinking, kimi-latest)

Result: Same 400 error on /coding/v1 endpoint

Expected Behavior

- temperature: 0.6 should be accepted (as the error message explicitly states)
- OR omitting temperature should use a safe default

Actual Behavior

Every single request returns HTTP 400 with the same temperature error. The model is completely unreachable.

Additional Context

- Tested with multiple API keys (same account and different accounts)
- Tested from fresh container with only openai SDK installed — same result
- The same code worked earlier today before ~16:49 UTC with no client-side changes
- Using the standard https://api.kimi.com/v1 endpoint (without /coding) returns 404


### What steps can reproduce the bug?

The API endpoint https://api.kimi.com/coding/v1/chat/completions returns HTTP 400 "invalid temperature: only 0.6 is
allowed for this model" for ALL temperature values, including the exact value 0.6 that the error message claims is
allowed.

Minimal reproduction:

curl -X POST https://api.kimi.com/coding/v1/chat/completions \
  -H "Authorization: Bearer $KIMI_API_KEY" \
  -H "Content-Type: application/json" \
  -H "User-Agent: KimiCLI/1.30.0" \
  -d '{"model":"kimi-for-coding","messages":[{"role":"user","content":"hi"}],"temperature":0.6}'

Expected: 200 OK with completion response
Actual: 400 Bad Request {"error":{"message":"invalid temperature: only 0.6 is allowed for this
model","type":"invalid_request_error"}}

The same error occurs when:
- temperature is omitted entirely
- temperature is set to any other value (0, 0.5, 1.0)
- using different model names on the same endpoint (kimi-k2.5, kimi-k2-thinking, kimi-latest)
- using a different API key

The same code worked correctly earlier today (~before 16:49 UTC on 2026-04-17) with no client-side changes.

### What is the expected behavior?

_No response_

### Additional information

_No response_

Start by running the supplied curl reproduction against https://api.kimi.com/coding/v1/chat/completions with temperature set to 0.6, omitted, and other values. Check the CLI's request path and parameter handling around the Kimi Coding API, then verify that valid requests receive a completion response instead of the reported 400 error.