[Kimi K2.6] reasoning_content is missing in assistant tool call messages

[beeckrumi0000]

Description

Summary

When using OpenCode with Kimi K2.6, tool calls fail with:

[Moonshot AI] thinking is enabled but reasoning_content is missing in assistant tool call message at index N

Environment

• OpenCode: latest
• Provider: Moonshot AI / Kimi K2.6
• Model: kimi-k2.6
• Thinking mode: enabled
• Tool calls: enabled

Steps to reproduce

1. Use Kimi K2.6 with thinking enabled.
2. Ask the agent to perform a task that requires tool calls.
3. After the first assistant/tool-call roundtrip, the next request fails.

Expected behavior

OpenCode should preserve and resend reasoning_content in assistant messages that contain tool calls.

Actual behavior

Moonshot rejects the request because reasoning_content is missing from the assistant tool call message.

Notes

This appears to happen specifically with Kimi K2.6 and tool calling.
The issue seems related to reasoning-content preservation across the tool-call roundtrip.

Plugins

Warp

OpenCode version

1.15.11

Steps to reproduce

No response

Screenshot and/or share link

Operating System

Ubuntu 22.04

Terminal

Warp

[github-actions]

This issue might be a duplicate of existing issues. Please check:

• thinking is enabled but reasoning_content is missing in assistant tool call message at index ... #25758: Same exact error message ("thinking is enabled but reasoning_content is missing in assistant tool call message at index N") affecting Kimi K2.6 and DeepSeek
• Bug: OpenCode Go bridge layer incompatible with kimi-k2.6 tool calls #26331: Directly covers the Kimi K2.6 bridge layer failing to preserve reasoning_content in tool call message history
• DeepSeek thinking mode: reasoning_content not passed back for tool call turns, causing 400 errors #24722: DeepSeek thinking mode — reasoning_content not passed back for tool call turns (same root cause pattern)

If your issue is specific to a different scenario or version, please add details clarifying how it differs from the above.

[rekram1-node]

can u send the session?

opencode export > session.json

do not run /export

[MrCh1ppy]

I’ve seen at least five issues regarding the same problem, and there are quite a few pull requests as well. It’s strange that these PRs still haven’t been reviewed.

[rekram1-node]

@MrCh1ppy thats because vast majority of PRs:

1. dont understand issue
2. are ai generated while not understanding issue (we like ai generated but if u dont understand it then i dont read it)
3. dont fix issue

[MrCh1ppy]

I understand that many PRs may not actually fix the issue.
But #25011 seems to target the Moonshot/Kimi message-format problem and was closed by automated cleanup without review.
So this still looks like a maintainer review/backlog problem, not only a PR quality problem.

[rekram1-node]

That was my own PR haha and it's not a fix for this issue it's a separate one and I havent flushed it out yet