fix(anthropic): broaden Kimi thinking-suppression to custom endpoints#17455
Merged
Conversation
The guard that drops Anthropic's `thinking` kwarg for Kimi endpoints was matched on `https://api.kimi.com/coding` only. Users configuring a custom Kimi-compatible gateway (or an official Moonshot host) with `api_mode: anthropic_messages` fall through to the generic third-party path, which strips thinking blocks AND still sends `thinking={enabled,...}` → upstream rejects with HTTP 400 "reasoning_content is missing in assistant tool call message at index N" on the next request after a tool call. Replace `_is_kimi_coding_endpoint` callers (history replay + thinking kwarg gate) with `_is_kimi_family_endpoint(base_url, model)` that also matches the `api.kimi.com` / `moonshot.ai` / `moonshot.cn` hosts and Kimi/Moonshot family model names (`kimi-`, `moonshot-`, `k1.`, `k2.`, …) for custom / proxied endpoints. Keeps the UA-header check in `build_anthropic_client` URL-only — the `claude-code/0.1.0` header is an official-Kimi contract. Plumbs optional `model` through `convert_messages_to_anthropic` so the unsigned reasoning_content→thinking block synthesised for Kimi's history validation survives the third-party signature-stripping pass on custom hosts too. Closes #17057.
6 tasks
donald131
pushed a commit
to donald131/hermes-agent
that referenced
this pull request
May 2, 2026
…NousResearch#17455) The guard that drops Anthropic's `thinking` kwarg for Kimi endpoints was matched on `https://api.kimi.com/coding` only. Users configuring a custom Kimi-compatible gateway (or an official Moonshot host) with `api_mode: anthropic_messages` fall through to the generic third-party path, which strips thinking blocks AND still sends `thinking={enabled,...}` → upstream rejects with HTTP 400 "reasoning_content is missing in assistant tool call message at index N" on the next request after a tool call. Replace `_is_kimi_coding_endpoint` callers (history replay + thinking kwarg gate) with `_is_kimi_family_endpoint(base_url, model)` that also matches the `api.kimi.com` / `moonshot.ai` / `moonshot.cn` hosts and Kimi/Moonshot family model names (`kimi-`, `moonshot-`, `k1.`, `k2.`, …) for custom / proxied endpoints. Keeps the UA-header check in `build_anthropic_client` URL-only — the `claude-code/0.1.0` header is an official-Kimi contract. Plumbs optional `model` through `convert_messages_to_anthropic` so the unsigned reasoning_content→thinking block synthesised for Kimi's history validation survives the third-party signature-stripping pass on custom hosts too. Closes NousResearch#17057.
02356abc
pushed a commit
to 02356abc/hermes-agent
that referenced
this pull request
May 14, 2026
…NousResearch#17455) The guard that drops Anthropic's `thinking` kwarg for Kimi endpoints was matched on `https://api.kimi.com/coding` only. Users configuring a custom Kimi-compatible gateway (or an official Moonshot host) with `api_mode: anthropic_messages` fall through to the generic third-party path, which strips thinking blocks AND still sends `thinking={enabled,...}` → upstream rejects with HTTP 400 "reasoning_content is missing in assistant tool call message at index N" on the next request after a tool call. Replace `_is_kimi_coding_endpoint` callers (history replay + thinking kwarg gate) with `_is_kimi_family_endpoint(base_url, model)` that also matches the `api.kimi.com` / `moonshot.ai` / `moonshot.cn` hosts and Kimi/Moonshot family model names (`kimi-`, `moonshot-`, `k1.`, `k2.`, …) for custom / proxied endpoints. Keeps the UA-header check in `build_anthropic_client` URL-only — the `claude-code/0.1.0` header is an official-Kimi contract. Plumbs optional `model` through `convert_messages_to_anthropic` so the unsigned reasoning_content→thinking block synthesised for Kimi's history validation survives the third-party signature-stripping pass on custom hosts too. Closes NousResearch#17057.
jsboige
pushed a commit
to jsboige/hermes-agent
that referenced
this pull request
May 14, 2026
…NousResearch#17455) The guard that drops Anthropic's `thinking` kwarg for Kimi endpoints was matched on `https://api.kimi.com/coding` only. Users configuring a custom Kimi-compatible gateway (or an official Moonshot host) with `api_mode: anthropic_messages` fall through to the generic third-party path, which strips thinking blocks AND still sends `thinking={enabled,...}` → upstream rejects with HTTP 400 "reasoning_content is missing in assistant tool call message at index N" on the next request after a tool call. Replace `_is_kimi_coding_endpoint` callers (history replay + thinking kwarg gate) with `_is_kimi_family_endpoint(base_url, model)` that also matches the `api.kimi.com` / `moonshot.ai` / `moonshot.cn` hosts and Kimi/Moonshot family model names (`kimi-`, `moonshot-`, `k1.`, `k2.`, …) for custom / proxied endpoints. Keeps the UA-header check in `build_anthropic_client` URL-only — the `claude-code/0.1.0` header is an official-Kimi contract. Plumbs optional `model` through `convert_messages_to_anthropic` so the unsigned reasoning_content→thinking block synthesised for Kimi's history validation survives the third-party signature-stripping pass on custom hosts too. Closes NousResearch#17057.
dannyJ848
pushed a commit
to dannyJ848/hermes-agent
that referenced
this pull request
May 17, 2026
…NousResearch#17455) The guard that drops Anthropic's `thinking` kwarg for Kimi endpoints was matched on `https://api.kimi.com/coding` only. Users configuring a custom Kimi-compatible gateway (or an official Moonshot host) with `api_mode: anthropic_messages` fall through to the generic third-party path, which strips thinking blocks AND still sends `thinking={enabled,...}` → upstream rejects with HTTP 400 "reasoning_content is missing in assistant tool call message at index N" on the next request after a tool call. Replace `_is_kimi_coding_endpoint` callers (history replay + thinking kwarg gate) with `_is_kimi_family_endpoint(base_url, model)` that also matches the `api.kimi.com` / `moonshot.ai` / `moonshot.cn` hosts and Kimi/Moonshot family model names (`kimi-`, `moonshot-`, `k1.`, `k2.`, …) for custom / proxied endpoints. Keeps the UA-header check in `build_anthropic_client` URL-only — the `claude-code/0.1.0` header is an official-Kimi contract. Plumbs optional `model` through `convert_messages_to_anthropic` so the unsigned reasoning_content→thinking block synthesised for Kimi's history validation survives the third-party signature-stripping pass on custom hosts too. Closes NousResearch#17057.
gweeteve
pushed a commit
to gweeteve/hermes-agent
that referenced
this pull request
Jun 2, 2026
…NousResearch#17455) The guard that drops Anthropic's `thinking` kwarg for Kimi endpoints was matched on `https://api.kimi.com/coding` only. Users configuring a custom Kimi-compatible gateway (or an official Moonshot host) with `api_mode: anthropic_messages` fall through to the generic third-party path, which strips thinking blocks AND still sends `thinking={enabled,...}` → upstream rejects with HTTP 400 "reasoning_content is missing in assistant tool call message at index N" on the next request after a tool call. Replace `_is_kimi_coding_endpoint` callers (history replay + thinking kwarg gate) with `_is_kimi_family_endpoint(base_url, model)` that also matches the `api.kimi.com` / `moonshot.ai` / `moonshot.cn` hosts and Kimi/Moonshot family model names (`kimi-`, `moonshot-`, `k1.`, `k2.`, …) for custom / proxied endpoints. Keeps the UA-header check in `build_anthropic_client` URL-only — the `claude-code/0.1.0` header is an official-Kimi contract. Plumbs optional `model` through `convert_messages_to_anthropic` so the unsigned reasoning_content→thinking block synthesised for Kimi's history validation survives the third-party signature-stripping pass on custom hosts too. Closes NousResearch#17057.
Egavasyug
pushed a commit
to Egavasyug/hermes-agent
that referenced
this pull request
Jun 10, 2026
…NousResearch#17455) The guard that drops Anthropic's `thinking` kwarg for Kimi endpoints was matched on `https://api.kimi.com/coding` only. Users configuring a custom Kimi-compatible gateway (or an official Moonshot host) with `api_mode: anthropic_messages` fall through to the generic third-party path, which strips thinking blocks AND still sends `thinking={enabled,...}` → upstream rejects with HTTP 400 "reasoning_content is missing in assistant tool call message at index N" on the next request after a tool call. Replace `_is_kimi_coding_endpoint` callers (history replay + thinking kwarg gate) with `_is_kimi_family_endpoint(base_url, model)` that also matches the `api.kimi.com` / `moonshot.ai` / `moonshot.cn` hosts and Kimi/Moonshot family model names (`kimi-`, `moonshot-`, `k1.`, `k2.`, …) for custom / proxied endpoints. Keeps the UA-header check in `build_anthropic_client` URL-only — the `claude-code/0.1.0` header is an official-Kimi contract. Plumbs optional `model` through `convert_messages_to_anthropic` so the unsigned reasoning_content→thinking block synthesised for Kimi's history validation survives the third-party signature-stripping pass on custom hosts too. Closes NousResearch#17057.
Seven74AI
pushed a commit
to Seven74AI/hermes-agent
that referenced
this pull request
Jun 13, 2026
…NousResearch#17455) The guard that drops Anthropic's `thinking` kwarg for Kimi endpoints was matched on `https://api.kimi.com/coding` only. Users configuring a custom Kimi-compatible gateway (or an official Moonshot host) with `api_mode: anthropic_messages` fall through to the generic third-party path, which strips thinking blocks AND still sends `thinking={enabled,...}` → upstream rejects with HTTP 400 "reasoning_content is missing in assistant tool call message at index N" on the next request after a tool call. Replace `_is_kimi_coding_endpoint` callers (history replay + thinking kwarg gate) with `_is_kimi_family_endpoint(base_url, model)` that also matches the `api.kimi.com` / `moonshot.ai` / `moonshot.cn` hosts and Kimi/Moonshot family model names (`kimi-`, `moonshot-`, `k1.`, `k2.`, …) for custom / proxied endpoints. Keeps the UA-header check in `build_anthropic_client` URL-only — the `claude-code/0.1.0` header is an official-Kimi contract. Plumbs optional `model` through `convert_messages_to_anthropic` so the unsigned reasoning_content→thinking block synthesised for Kimi's history validation survives the third-party signature-stripping pass on custom hosts too. Closes NousResearch#17057.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes #17057.
Summary
Kimi-compatible custom endpoints +
api_mode: anthropic_messages+ thinking no longer fail with HTTP 400 after a tool call. The gate that drops Anthropic'sthinkingkwarg for Kimi was matched onhttps://api.kimi.com/codingonly; users with a private gateway fronting Kimi (or an official Moonshot host) fell through to the generic third-party path, which strips thinking blocks AND still sendsthinking={enabled,...}. Upstream then rejects the replay withreasoning_content is missing in assistant tool call message at index N.Changes
agent/anthropic_adapter.py: new_is_kimi_family_endpoint(base_url, model)coveringapi.kimi.com/coding*URLs, host matches onapi.kimi.com/moonshot.ai/moonshot.cn, and Kimi/Moonshot family model names (kimi-,moonshot-,k1.,k2.,k25,k2.5). Strips vendor prefix somoonshotai/kimi-k2.5is recognised the same askimi-k2.5.build_anthropic_kwargs, and unsigned-thinking preservation inconvert_messages_to_anthropic.convert_messages_to_anthropicgrows an optionalmodelparam so the custom-endpoint Kimi branch gets the signal it needs.build_anthropic_clientUA-header check stays URL-only —claude-code/0.1.0is an official-Kimi-only contract.tests/agent/test_kimi_coding_anthropic_thinking.py: custom-endpoint × {kimi-2.6, kimi-k2.5, moonshot-v1-*, kimi_thinking, vendor-prefixed} matrix; negative MiniMax test; replay test confirming unsigned reasoning_content→thinking survives the third-party strip on custom Kimi hosts. Retargeted the staletest_kimi_root_endpoint_unaffectedwhose "we should never see it" premise didn't survive user-configurableapi_mode.Validation
tests/agent/test_kimi_coding_anthropic_thinking.py: 17/17 passing (7 new)tests/agent/ -k "anthropic or kimi or moonshot or thinking": 349/349 passingtests/run_agent/ -k "kimi or moonshot or reasoning or thinking": 144 passed, 1 skippedbase_url=http://custom-endpoint.example.com,model=kimi-2.6,reasoning_effort=medium) now produces outbound kwargs withthinkingabsent.Root cause (one line)
_is_kimi_coding_endpointwas hostname-only and hardcoded to Kimi's/codingURL; any other host speaking Anthropic Messages to a Kimi-family model hit the generic third-party path, which is incompatible with Kimi's reasoning_content semantics.