feat(agent): adaptive thinking for Kimi-family Anthropic endpoints

[webtecnica]

Kimi's Anthropic-compatible endpoints (api.moonshot.cn/anthropic, api.kimi.com/coding) implement the adaptive thinking contract — they accept thinking.type=adaptive + output_config.effort (all of low, medium, high, xhigh, max verified live) and return thinking blocks, and the replay-validation 400s that originally motivated dropping the parameter (#13848) no longer occur.

_supports_adaptive_thinking() now returns True for Kimi-family models, so they get thinking={type: adaptive, display: summarized} + output_config.effort via ADAPTIVE_EFFORT_MAP instead of nothing, and the blanket drop of the thinking parameter for Kimi-family endpoints is removed. MiniMax and other non-adaptive third parties keep the manual budget_tokens path; Claude behavior is unchanged.

What does this PR do?

Kimi's Anthropic-compatible endpoints (api.moonshot.cn/anthropic, api.kimi.com/coding) implement the adaptive thinking contract. Stop dropping the thinking parameter for the Kimi family and send it in adaptive form.

Related Issue

Fixes #67228

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)

Changes Made

• _supports_adaptive_thinking() returns True for Kimi-family models → thinking={type: adaptive, display: summarized} + output_config.effort via ADAPTIVE_EFFORT_MAP (never manual budget_tokens)
• the blanket drop of the thinking parameter for Kimi-family endpoints ([Bug]: 400 error for kimi-for-coding #13848-era mitigation) is removed
• MiniMax and other non-adaptive third parties keep the manual budget_tokens path; Claude behavior unchanged

How to Test

Test it against the kimi for coding endpoint:

• thinking.type="adaptive" + output_config.effort in {low, medium, high, xhigh, max} → all HTTP 200 with thinking blocks; display="summarized" accepted
• multi-turn tool-call replays in every history shape — signed thinking blocks verbatim, unsigned placeholder blocks, and no thinking blocks at all (the exact [Bug]: 400 error for kimi-for-coding #13848 trigger) — all return HTTP 200; the replay-validation 400s that motivated the original drop no longer occur

Checklist

• Commit messages follow Conventional Commits
• All tests pass
• Tests added for the changes
• No unrelated commits

[alt-glitch]

This was generated by AI during triage.

Duplicate of #67228: the production adaptive-thinking change is the same; only test-comment whitespace differs. The endpoint-contract decision noted on #67228 still applies.

[teknium1]

Thanks for isolating the Kimi change and retaining the existing replay coverage. Current main still suppresses Anthropic thinking for Kimi-family endpoints at agent/anthropic_adapter.py:2647-2649; this PR changes that exact path by classifying Kimi models as adaptive at agent/anthropic_adapter.py:255-256 and allowing the normal adaptive mapping at agent/anthropic_adapter.py:2648-2664.

The PR tests cover official and proxied Kimi-family endpoint shapes, effort mapping, disabled thinking, and retain the unsigned replay regression at tests/agent/test_kimi_coding_anthropic_thinking.py:172-209. No code-level blocking issue was found in static review.

Automated hermes-sweeper review.

[teknium1]

Closing as an exact duplicate of #67228 — the two diffs are byte-identical (same adapter change, same tests) and #67228 was submitted first (01:03 vs 01:11 UTC). Consolidating review onto #67228; if it lands, both approaches are covered since they're the same change. Thanks for the contribution — please coordinate on #67228 going forward.