fix(anthropic): default new Claude models to the modern thinking contract - #42991
Merged
Conversation
…ract
New Anthropic models without a recognized version substring (claude-fable-5
and future named/numbered releases) were classified as legacy and routed down
the manual-thinking path, which made OpenRouter emit thinking.type.disabled —
a form reasoning-mandatory Claude models reject with a non-retryable HTTP 400.
Invert the brittle version-substring allowlists to default-to-modern (mirroring
_get_anthropic_max_output): unknown Claude models get the adaptive/xhigh/
no-sampling contract, with an explicit legacy list for older families. Non-Claude
Anthropic-Messages models (minimax, qwen3, …) keep the manual path.
- anthropic_adapter: _supports_adaptive_thinking / _supports_xhigh_effort /
_forbids_sampling_params now default unknown Claude models to modern; legacy
families enumerated in _LEGACY_MANUAL_THINKING_CLAUDE_SUBSTRINGS.
- openrouter profile: omit reasoning entirely (→ adaptive default) instead of
forwarding {enabled:false} for reasoning-mandatory Anthropic models; legacy
Anthropic + all non-Anthropic models still pass the disable form through.
- model_metadata + output-limit table: register claude-fable-5 (1M ctx, 128K out).
Tests assert the invariant ("unknown Claude model -> modern contract; legacy
stays manual; non-Claude unaffected"), not specific model names.
Contributor
🔎 Lint report:
|
alt-glitch
added
type/bug
Contributor
|
Verification comment from automated code review Reviewed this PR as part of a scheduled review pass. The change is clean and well-structured. Criteria matched: security-critical (model contract defaults), non-trivial logic (legacy→modern inversion), zero existing reviews, good test coverage (8 test functions across 3 files) Observations: The default-to-modern architecture is sound — inverting the allowlist to a legacy-blocklist means future Claude releases (fable, mythos, numbered 5.x) work without code changes. The _ANTHROPIC_REASONING_OPTIONAL_SUBSTRINGS in the OpenRouter plugin mirrors the same legacy list, ensuring consistency across the Anthropic adapter and OpenRouter profile. Test coverage is thorough: fable-class models route as adaptive, legacy stays manual, non-Claude Anthropic models use the manual path, 4.6 is adaptive-but-not-xhigh/no-sampling, and the OpenRouter reasoning-disable guard is tested for mandatory, legacy, non-Anthropic, and enabled-on-mandatory paths. No issues found. LGTM. 👍 |
alt-glitch
added a commit
that referenced
this pull request
Jun 9, 2026
…dels (#43012) The previous fix (#42991) only omitted reasoning when it was being disabled. But reasoning-mandatory Anthropic models (Claude 4.6+, fable) 400 with thinking.type.disabled on EVERY tool-continuation turn even when reasoning is enabled: chat_completions never replays signed thinking blocks, so the prior assistant tool_call has no thinking, and OpenRouter resolves "reasoning requested but history has none" by emitting thinking.type.disabled — which these models reject. Result: first turn works, every turn after the first tool call dies (HTTP 400, non-retryable). OpenRouter ignores reasoning.effort for adaptive Anthropic models anyway (the model self-decides), so the reasoning field is pointless for them on every turn and harmful on tool-replay turns. Omit it entirely → adaptive default. - openrouter profile: drop the reasoning field for reasoning-mandatory Anthropic models regardless of enabled/disabled; legacy Anthropic + non-Anthropic models unchanged. - tests: assert omission across enabled/disabled/effort variants; parity tests switched to a non-Anthropic reasoning model (deepseek) since Anthropic 4.6+ no longer carries a reasoning field. Verified live end-to-end: a tool-replay turn on anthropic/claude-fable-5 with reasoning enabled now builds extra_body=None and returns HTTP 200 (was 400).
ziliangpeng
pushed a commit
to ziliangpeng/hermes-agent
that referenced
this pull request
Jun 10, 2026
…ract (NousResearch#42991) New Anthropic models without a recognized version substring (claude-fable-5 and future named/numbered releases) were classified as legacy and routed down the manual-thinking path, which made OpenRouter emit thinking.type.disabled — a form reasoning-mandatory Claude models reject with a non-retryable HTTP 400. Invert the brittle version-substring allowlists to default-to-modern (mirroring _get_anthropic_max_output): unknown Claude models get the adaptive/xhigh/ no-sampling contract, with an explicit legacy list for older families. Non-Claude Anthropic-Messages models (minimax, qwen3, …) keep the manual path. - anthropic_adapter: _supports_adaptive_thinking / _supports_xhigh_effort / _forbids_sampling_params now default unknown Claude models to modern; legacy families enumerated in _LEGACY_MANUAL_THINKING_CLAUDE_SUBSTRINGS. - openrouter profile: omit reasoning entirely (→ adaptive default) instead of forwarding {enabled:false} for reasoning-mandatory Anthropic models; legacy Anthropic + all non-Anthropic models still pass the disable form through. - model_metadata + output-limit table: register claude-fable-5 (1M ctx, 128K out). Tests assert the invariant ("unknown Claude model -> modern contract; legacy stays manual; non-Claude unaffected"), not specific model names.
ziliangpeng
pushed a commit
to ziliangpeng/hermes-agent
that referenced
this pull request
Jun 10, 2026
…dels (NousResearch#43012) The previous fix (NousResearch#42991) only omitted reasoning when it was being disabled. But reasoning-mandatory Anthropic models (Claude 4.6+, fable) 400 with thinking.type.disabled on EVERY tool-continuation turn even when reasoning is enabled: chat_completions never replays signed thinking blocks, so the prior assistant tool_call has no thinking, and OpenRouter resolves "reasoning requested but history has none" by emitting thinking.type.disabled — which these models reject. Result: first turn works, every turn after the first tool call dies (HTTP 400, non-retryable). OpenRouter ignores reasoning.effort for adaptive Anthropic models anyway (the model self-decides), so the reasoning field is pointless for them on every turn and harmful on tool-replay turns. Omit it entirely → adaptive default. - openrouter profile: drop the reasoning field for reasoning-mandatory Anthropic models regardless of enabled/disabled; legacy Anthropic + non-Anthropic models unchanged. - tests: assert omission across enabled/disabled/effort variants; parity tests switched to a non-Anthropic reasoning model (deepseek) since Anthropic 4.6+ no longer carries a reasoning field. Verified live end-to-end: a tool-replay turn on anthropic/claude-fable-5 with reasoning enabled now builds extra_body=None and returns HTTP 200 (was 400).
alt-glitch
added a commit
that referenced
this pull request
Jun 10, 2026
…hropic models (#43436) * fix(openrouter): route reasoning_effort to verbosity for adaptive Anthropic models Reasoning-mandatory Anthropic models (Claude 4.6+/fable/mythos-class) over OpenRouter ignore reasoning.effort and use adaptive thinking. #42991 correctly stopped Hermes from sending a reasoning field to them (it 400s), but put nothing in its place — leaving agent.reasoning_effort a silent no-op on the OpenRouter path: the model always ran at its adaptive default (high) regardless of config. OpenRouter honors the requested effort on the top-level verbosity field instead (maps to Anthropic output_config.effort). Route the existing reasoning_config[effort] there for these models while still never emitting a reasoning field, preserving the #42991 fix. No new config arg — the value the user already sets via agent.reasoning_effort now flows to verbosity. - low/medium/high/xhigh/max pass through verbatim (OpenRouter accepts the extended scale for Claude; verified live HTTP 200 + monotonic token spend). - effort unset/none/disabled omits verbosity so the model keeps its default. - native Anthropic transport already correct; unchanged. Fixes #43432 * test(openrouter): cover real effort range (add minimal, frame max as passthrough) Adversarial review noted the verbosity tests looped over 'max' — a value parse_reasoning_effort can never produce — while omitting 'minimal', which it can. Align the routing test with the real config range (VALID_REASONING_EFFORTS = minimal/low/medium/high/xhigh) and keep a separate value-agnostic passthrough test that documents why xhigh/max must survive verbatim (TypedDict, no runtime literal validation; OpenRouter accepts the extended scale for Claude). * docs: explain reasoning_effort -> verbosity routing for adaptive Anthropic models Document that reasoning_effort transparently maps to OpenRouter's verbosity field for adaptive-thinking Anthropic models (Claude 4.6+/Fable/Mythos), where reasoning.effort is ignored. Note xhigh is the configurable ceiling (max is wire- only). Add verbosity as a top-level-kwarg example in the provider-plugin guide.
changman
pushed a commit
to changman/hermes-agent
that referenced
this pull request
Jun 10, 2026
…ract (NousResearch#42991) New Anthropic models without a recognized version substring (claude-fable-5 and future named/numbered releases) were classified as legacy and routed down the manual-thinking path, which made OpenRouter emit thinking.type.disabled — a form reasoning-mandatory Claude models reject with a non-retryable HTTP 400. Invert the brittle version-substring allowlists to default-to-modern (mirroring _get_anthropic_max_output): unknown Claude models get the adaptive/xhigh/ no-sampling contract, with an explicit legacy list for older families. Non-Claude Anthropic-Messages models (minimax, qwen3, …) keep the manual path. - anthropic_adapter: _supports_adaptive_thinking / _supports_xhigh_effort / _forbids_sampling_params now default unknown Claude models to modern; legacy families enumerated in _LEGACY_MANUAL_THINKING_CLAUDE_SUBSTRINGS. - openrouter profile: omit reasoning entirely (→ adaptive default) instead of forwarding {enabled:false} for reasoning-mandatory Anthropic models; legacy Anthropic + all non-Anthropic models still pass the disable form through. - model_metadata + output-limit table: register claude-fable-5 (1M ctx, 128K out). Tests assert the invariant ("unknown Claude model -> modern contract; legacy stays manual; non-Claude unaffected"), not specific model names.
changman
pushed a commit
to changman/hermes-agent
that referenced
this pull request
Jun 10, 2026
…dels (NousResearch#43012) The previous fix (NousResearch#42991) only omitted reasoning when it was being disabled. But reasoning-mandatory Anthropic models (Claude 4.6+, fable) 400 with thinking.type.disabled on EVERY tool-continuation turn even when reasoning is enabled: chat_completions never replays signed thinking blocks, so the prior assistant tool_call has no thinking, and OpenRouter resolves "reasoning requested but history has none" by emitting thinking.type.disabled — which these models reject. Result: first turn works, every turn after the first tool call dies (HTTP 400, non-retryable). OpenRouter ignores reasoning.effort for adaptive Anthropic models anyway (the model self-decides), so the reasoning field is pointless for them on every turn and harmful on tool-replay turns. Omit it entirely → adaptive default. - openrouter profile: drop the reasoning field for reasoning-mandatory Anthropic models regardless of enabled/disabled; legacy Anthropic + non-Anthropic models unchanged. - tests: assert omission across enabled/disabled/effort variants; parity tests switched to a non-Anthropic reasoning model (deepseek) since Anthropic 4.6+ no longer carries a reasoning field. Verified live end-to-end: a tool-replay turn on anthropic/claude-fable-5 with reasoning enabled now builds extra_body=None and returns HTTP 200 (was 400).
changman
pushed a commit
to changman/hermes-agent
that referenced
this pull request
Jun 10, 2026
…hropic models (NousResearch#43436) * fix(openrouter): route reasoning_effort to verbosity for adaptive Anthropic models Reasoning-mandatory Anthropic models (Claude 4.6+/fable/mythos-class) over OpenRouter ignore reasoning.effort and use adaptive thinking. NousResearch#42991 correctly stopped Hermes from sending a reasoning field to them (it 400s), but put nothing in its place — leaving agent.reasoning_effort a silent no-op on the OpenRouter path: the model always ran at its adaptive default (high) regardless of config. OpenRouter honors the requested effort on the top-level verbosity field instead (maps to Anthropic output_config.effort). Route the existing reasoning_config[effort] there for these models while still never emitting a reasoning field, preserving the NousResearch#42991 fix. No new config arg — the value the user already sets via agent.reasoning_effort now flows to verbosity. - low/medium/high/xhigh/max pass through verbatim (OpenRouter accepts the extended scale for Claude; verified live HTTP 200 + monotonic token spend). - effort unset/none/disabled omits verbosity so the model keeps its default. - native Anthropic transport already correct; unchanged. Fixes NousResearch#43432 * test(openrouter): cover real effort range (add minimal, frame max as passthrough) Adversarial review noted the verbosity tests looped over 'max' — a value parse_reasoning_effort can never produce — while omitting 'minimal', which it can. Align the routing test with the real config range (VALID_REASONING_EFFORTS = minimal/low/medium/high/xhigh) and keep a separate value-agnostic passthrough test that documents why xhigh/max must survive verbatim (TypedDict, no runtime literal validation; OpenRouter accepts the extended scale for Claude). * docs: explain reasoning_effort -> verbosity routing for adaptive Anthropic models Document that reasoning_effort transparently maps to OpenRouter's verbosity field for adaptive-thinking Anthropic models (Claude 4.6+/Fable/Mythos), where reasoning.effort is ignored. Note xhigh is the configurable ceiling (max is wire- only). Add verbosity as a top-level-kwarg example in the provider-plugin guide.
alt-glitch
added a commit
that referenced
this pull request
Jun 14, 2026
…ract (#42991) New Anthropic models without a recognized version substring (claude-fable-5 and future named/numbered releases) were classified as legacy and routed down the manual-thinking path, which made OpenRouter emit thinking.type.disabled — a form reasoning-mandatory Claude models reject with a non-retryable HTTP 400. Invert the brittle version-substring allowlists to default-to-modern (mirroring _get_anthropic_max_output): unknown Claude models get the adaptive/xhigh/ no-sampling contract, with an explicit legacy list for older families. Non-Claude Anthropic-Messages models (minimax, qwen3, …) keep the manual path. - anthropic_adapter: _supports_adaptive_thinking / _supports_xhigh_effort / _forbids_sampling_params now default unknown Claude models to modern; legacy families enumerated in _LEGACY_MANUAL_THINKING_CLAUDE_SUBSTRINGS. - openrouter profile: omit reasoning entirely (→ adaptive default) instead of forwarding {enabled:false} for reasoning-mandatory Anthropic models; legacy Anthropic + all non-Anthropic models still pass the disable form through. - model_metadata + output-limit table: register claude-fable-5 (1M ctx, 128K out). Tests assert the invariant ("unknown Claude model -> modern contract; legacy stays manual; non-Claude unaffected"), not specific model names.
alt-glitch
added a commit
that referenced
this pull request
Jun 14, 2026
…dels (#43012) The previous fix (#42991) only omitted reasoning when it was being disabled. But reasoning-mandatory Anthropic models (Claude 4.6+, fable) 400 with thinking.type.disabled on EVERY tool-continuation turn even when reasoning is enabled: chat_completions never replays signed thinking blocks, so the prior assistant tool_call has no thinking, and OpenRouter resolves "reasoning requested but history has none" by emitting thinking.type.disabled — which these models reject. Result: first turn works, every turn after the first tool call dies (HTTP 400, non-retryable). OpenRouter ignores reasoning.effort for adaptive Anthropic models anyway (the model self-decides), so the reasoning field is pointless for them on every turn and harmful on tool-replay turns. Omit it entirely → adaptive default. - openrouter profile: drop the reasoning field for reasoning-mandatory Anthropic models regardless of enabled/disabled; legacy Anthropic + non-Anthropic models unchanged. - tests: assert omission across enabled/disabled/effort variants; parity tests switched to a non-Anthropic reasoning model (deepseek) since Anthropic 4.6+ no longer carries a reasoning field. Verified live end-to-end: a tool-replay turn on anthropic/claude-fable-5 with reasoning enabled now builds extra_body=None and returns HTTP 200 (was 400).
alt-glitch
added a commit
that referenced
this pull request
Jun 14, 2026
…hropic models (#43436) * fix(openrouter): route reasoning_effort to verbosity for adaptive Anthropic models Reasoning-mandatory Anthropic models (Claude 4.6+/fable/mythos-class) over OpenRouter ignore reasoning.effort and use adaptive thinking. #42991 correctly stopped Hermes from sending a reasoning field to them (it 400s), but put nothing in its place — leaving agent.reasoning_effort a silent no-op on the OpenRouter path: the model always ran at its adaptive default (high) regardless of config. OpenRouter honors the requested effort on the top-level verbosity field instead (maps to Anthropic output_config.effort). Route the existing reasoning_config[effort] there for these models while still never emitting a reasoning field, preserving the #42991 fix. No new config arg — the value the user already sets via agent.reasoning_effort now flows to verbosity. - low/medium/high/xhigh/max pass through verbatim (OpenRouter accepts the extended scale for Claude; verified live HTTP 200 + monotonic token spend). - effort unset/none/disabled omits verbosity so the model keeps its default. - native Anthropic transport already correct; unchanged. Fixes #43432 * test(openrouter): cover real effort range (add minimal, frame max as passthrough) Adversarial review noted the verbosity tests looped over 'max' — a value parse_reasoning_effort can never produce — while omitting 'minimal', which it can. Align the routing test with the real config range (VALID_REASONING_EFFORTS = minimal/low/medium/high/xhigh) and keep a separate value-agnostic passthrough test that documents why xhigh/max must survive verbatim (TypedDict, no runtime literal validation; OpenRouter accepts the extended scale for Claude). * docs: explain reasoning_effort -> verbosity routing for adaptive Anthropic models Document that reasoning_effort transparently maps to OpenRouter's verbosity field for adaptive-thinking Anthropic models (Claude 4.6+/Fable/Mythos), where reasoning.effort is ignored. Note xhigh is the configurable ceiling (max is wire- only). Add verbosity as a top-level-kwarg example in the provider-plugin guide.
ether-btc
pushed a commit
to ether-btc/hermes-agent
that referenced
this pull request
Jun 15, 2026
…ract (NousResearch#42991) New Anthropic models without a recognized version substring (claude-fable-5 and future named/numbered releases) were classified as legacy and routed down the manual-thinking path, which made OpenRouter emit thinking.type.disabled — a form reasoning-mandatory Claude models reject with a non-retryable HTTP 400. Invert the brittle version-substring allowlists to default-to-modern (mirroring _get_anthropic_max_output): unknown Claude models get the adaptive/xhigh/ no-sampling contract, with an explicit legacy list for older families. Non-Claude Anthropic-Messages models (minimax, qwen3, …) keep the manual path. - anthropic_adapter: _supports_adaptive_thinking / _supports_xhigh_effort / _forbids_sampling_params now default unknown Claude models to modern; legacy families enumerated in _LEGACY_MANUAL_THINKING_CLAUDE_SUBSTRINGS. - openrouter profile: omit reasoning entirely (→ adaptive default) instead of forwarding {enabled:false} for reasoning-mandatory Anthropic models; legacy Anthropic + all non-Anthropic models still pass the disable form through. - model_metadata + output-limit table: register claude-fable-5 (1M ctx, 128K out). Tests assert the invariant ("unknown Claude model -> modern contract; legacy stays manual; non-Claude unaffected"), not specific model names.
ether-btc
pushed a commit
to ether-btc/hermes-agent
that referenced
this pull request
Jun 15, 2026
…dels (NousResearch#43012) The previous fix (NousResearch#42991) only omitted reasoning when it was being disabled. But reasoning-mandatory Anthropic models (Claude 4.6+, fable) 400 with thinking.type.disabled on EVERY tool-continuation turn even when reasoning is enabled: chat_completions never replays signed thinking blocks, so the prior assistant tool_call has no thinking, and OpenRouter resolves "reasoning requested but history has none" by emitting thinking.type.disabled — which these models reject. Result: first turn works, every turn after the first tool call dies (HTTP 400, non-retryable). OpenRouter ignores reasoning.effort for adaptive Anthropic models anyway (the model self-decides), so the reasoning field is pointless for them on every turn and harmful on tool-replay turns. Omit it entirely → adaptive default. - openrouter profile: drop the reasoning field for reasoning-mandatory Anthropic models regardless of enabled/disabled; legacy Anthropic + non-Anthropic models unchanged. - tests: assert omission across enabled/disabled/effort variants; parity tests switched to a non-Anthropic reasoning model (deepseek) since Anthropic 4.6+ no longer carries a reasoning field. Verified live end-to-end: a tool-replay turn on anthropic/claude-fable-5 with reasoning enabled now builds extra_body=None and returns HTTP 200 (was 400).
davidgut1982
pushed a commit
to davidgut1982/hermes-agent
that referenced
this pull request
Jun 17, 2026
…ract (NousResearch#42991) New Anthropic models without a recognized version substring (claude-fable-5 and future named/numbered releases) were classified as legacy and routed down the manual-thinking path, which made OpenRouter emit thinking.type.disabled — a form reasoning-mandatory Claude models reject with a non-retryable HTTP 400. Invert the brittle version-substring allowlists to default-to-modern (mirroring _get_anthropic_max_output): unknown Claude models get the adaptive/xhigh/ no-sampling contract, with an explicit legacy list for older families. Non-Claude Anthropic-Messages models (minimax, qwen3, …) keep the manual path. - anthropic_adapter: _supports_adaptive_thinking / _supports_xhigh_effort / _forbids_sampling_params now default unknown Claude models to modern; legacy families enumerated in _LEGACY_MANUAL_THINKING_CLAUDE_SUBSTRINGS. - openrouter profile: omit reasoning entirely (→ adaptive default) instead of forwarding {enabled:false} for reasoning-mandatory Anthropic models; legacy Anthropic + all non-Anthropic models still pass the disable form through. - model_metadata + output-limit table: register claude-fable-5 (1M ctx, 128K out). Tests assert the invariant ("unknown Claude model -> modern contract; legacy stays manual; non-Claude unaffected"), not specific model names.
davidgut1982
pushed a commit
to davidgut1982/hermes-agent
that referenced
this pull request
Jun 17, 2026
…dels (NousResearch#43012) The previous fix (NousResearch#42991) only omitted reasoning when it was being disabled. But reasoning-mandatory Anthropic models (Claude 4.6+, fable) 400 with thinking.type.disabled on EVERY tool-continuation turn even when reasoning is enabled: chat_completions never replays signed thinking blocks, so the prior assistant tool_call has no thinking, and OpenRouter resolves "reasoning requested but history has none" by emitting thinking.type.disabled — which these models reject. Result: first turn works, every turn after the first tool call dies (HTTP 400, non-retryable). OpenRouter ignores reasoning.effort for adaptive Anthropic models anyway (the model self-decides), so the reasoning field is pointless for them on every turn and harmful on tool-replay turns. Omit it entirely → adaptive default. - openrouter profile: drop the reasoning field for reasoning-mandatory Anthropic models regardless of enabled/disabled; legacy Anthropic + non-Anthropic models unchanged. - tests: assert omission across enabled/disabled/effort variants; parity tests switched to a non-Anthropic reasoning model (deepseek) since Anthropic 4.6+ no longer carries a reasoning field. Verified live end-to-end: a tool-replay turn on anthropic/claude-fable-5 with reasoning enabled now builds extra_body=None and returns HTTP 200 (was 400).
santhreal
pushed a commit
to santhreal/hermes-agent
that referenced
this pull request
Jul 13, 2026
…ract (NousResearch#42991) New Anthropic models without a recognized version substring (claude-fable-5 and future named/numbered releases) were classified as legacy and routed down the manual-thinking path, which made OpenRouter emit thinking.type.disabled — a form reasoning-mandatory Claude models reject with a non-retryable HTTP 400. Invert the brittle version-substring allowlists to default-to-modern (mirroring _get_anthropic_max_output): unknown Claude models get the adaptive/xhigh/ no-sampling contract, with an explicit legacy list for older families. Non-Claude Anthropic-Messages models (minimax, qwen3, …) keep the manual path. - anthropic_adapter: _supports_adaptive_thinking / _supports_xhigh_effort / _forbids_sampling_params now default unknown Claude models to modern; legacy families enumerated in _LEGACY_MANUAL_THINKING_CLAUDE_SUBSTRINGS. - openrouter profile: omit reasoning entirely (→ adaptive default) instead of forwarding {enabled:false} for reasoning-mandatory Anthropic models; legacy Anthropic + all non-Anthropic models still pass the disable form through. - model_metadata + output-limit table: register claude-fable-5 (1M ctx, 128K out). Tests assert the invariant ("unknown Claude model -> modern contract; legacy stays manual; non-Claude unaffected"), not specific model names.
santhreal
pushed a commit
to santhreal/hermes-agent
that referenced
this pull request
Jul 13, 2026
…dels (NousResearch#43012) The previous fix (NousResearch#42991) only omitted reasoning when it was being disabled. But reasoning-mandatory Anthropic models (Claude 4.6+, fable) 400 with thinking.type.disabled on EVERY tool-continuation turn even when reasoning is enabled: chat_completions never replays signed thinking blocks, so the prior assistant tool_call has no thinking, and OpenRouter resolves "reasoning requested but history has none" by emitting thinking.type.disabled — which these models reject. Result: first turn works, every turn after the first tool call dies (HTTP 400, non-retryable). OpenRouter ignores reasoning.effort for adaptive Anthropic models anyway (the model self-decides), so the reasoning field is pointless for them on every turn and harmful on tool-replay turns. Omit it entirely → adaptive default. - openrouter profile: drop the reasoning field for reasoning-mandatory Anthropic models regardless of enabled/disabled; legacy Anthropic + non-Anthropic models unchanged. - tests: assert omission across enabled/disabled/effort variants; parity tests switched to a non-Anthropic reasoning model (deepseek) since Anthropic 4.6+ no longer carries a reasoning field. Verified live end-to-end: a tool-replay turn on anthropic/claude-fable-5 with reasoning enabled now builds extra_body=None and returns HTTP 200 (was 400).
santhreal
pushed a commit
to santhreal/hermes-agent
that referenced
this pull request
Jul 13, 2026
…hropic models (NousResearch#43436) * fix(openrouter): route reasoning_effort to verbosity for adaptive Anthropic models Reasoning-mandatory Anthropic models (Claude 4.6+/fable/mythos-class) over OpenRouter ignore reasoning.effort and use adaptive thinking. NousResearch#42991 correctly stopped Hermes from sending a reasoning field to them (it 400s), but put nothing in its place — leaving agent.reasoning_effort a silent no-op on the OpenRouter path: the model always ran at its adaptive default (high) regardless of config. OpenRouter honors the requested effort on the top-level verbosity field instead (maps to Anthropic output_config.effort). Route the existing reasoning_config[effort] there for these models while still never emitting a reasoning field, preserving the NousResearch#42991 fix. No new config arg — the value the user already sets via agent.reasoning_effort now flows to verbosity. - low/medium/high/xhigh/max pass through verbatim (OpenRouter accepts the extended scale for Claude; verified live HTTP 200 + monotonic token spend). - effort unset/none/disabled omits verbosity so the model keeps its default. - native Anthropic transport already correct; unchanged. Fixes NousResearch#43432 * test(openrouter): cover real effort range (add minimal, frame max as passthrough) Adversarial review noted the verbosity tests looped over 'max' — a value parse_reasoning_effort can never produce — while omitting 'minimal', which it can. Align the routing test with the real config range (VALID_REASONING_EFFORTS = minimal/low/medium/high/xhigh) and keep a separate value-agnostic passthrough test that documents why xhigh/max must survive verbatim (TypedDict, no runtime literal validation; OpenRouter accepts the extended scale for Claude). * docs: explain reasoning_effort -> verbosity routing for adaptive Anthropic models Document that reasoning_effort transparently maps to OpenRouter's verbosity field for adaptive-thinking Anthropic models (Claude 4.6+/Fable/Mythos), where reasoning.effort is ignored. Note xhigh is the configurable ceiling (max is wire- only). Add verbosity as a top-level-kwarg example in the provider-plugin guide.
donbowman
pushed a commit
to donbowman/hermes-agent
that referenced
this pull request
Jul 13, 2026
…ract (NousResearch#42991) New Anthropic models without a recognized version substring (claude-fable-5 and future named/numbered releases) were classified as legacy and routed down the manual-thinking path, which made OpenRouter emit thinking.type.disabled — a form reasoning-mandatory Claude models reject with a non-retryable HTTP 400. Invert the brittle version-substring allowlists to default-to-modern (mirroring _get_anthropic_max_output): unknown Claude models get the adaptive/xhigh/ no-sampling contract, with an explicit legacy list for older families. Non-Claude Anthropic-Messages models (minimax, qwen3, …) keep the manual path. - anthropic_adapter: _supports_adaptive_thinking / _supports_xhigh_effort / _forbids_sampling_params now default unknown Claude models to modern; legacy families enumerated in _LEGACY_MANUAL_THINKING_CLAUDE_SUBSTRINGS. - openrouter profile: omit reasoning entirely (→ adaptive default) instead of forwarding {enabled:false} for reasoning-mandatory Anthropic models; legacy Anthropic + all non-Anthropic models still pass the disable form through. - model_metadata + output-limit table: register claude-fable-5 (1M ctx, 128K out). Tests assert the invariant ("unknown Claude model -> modern contract; legacy stays manual; non-Claude unaffected"), not specific model names.
donbowman
pushed a commit
to donbowman/hermes-agent
that referenced
this pull request
Jul 13, 2026
…dels (NousResearch#43012) The previous fix (NousResearch#42991) only omitted reasoning when it was being disabled. But reasoning-mandatory Anthropic models (Claude 4.6+, fable) 400 with thinking.type.disabled on EVERY tool-continuation turn even when reasoning is enabled: chat_completions never replays signed thinking blocks, so the prior assistant tool_call has no thinking, and OpenRouter resolves "reasoning requested but history has none" by emitting thinking.type.disabled — which these models reject. Result: first turn works, every turn after the first tool call dies (HTTP 400, non-retryable). OpenRouter ignores reasoning.effort for adaptive Anthropic models anyway (the model self-decides), so the reasoning field is pointless for them on every turn and harmful on tool-replay turns. Omit it entirely → adaptive default. - openrouter profile: drop the reasoning field for reasoning-mandatory Anthropic models regardless of enabled/disabled; legacy Anthropic + non-Anthropic models unchanged. - tests: assert omission across enabled/disabled/effort variants; parity tests switched to a non-Anthropic reasoning model (deepseek) since Anthropic 4.6+ no longer carries a reasoning field. Verified live end-to-end: a tool-replay turn on anthropic/claude-fable-5 with reasoning enabled now builds extra_body=None and returns HTTP 200 (was 400).
donbowman
pushed a commit
to donbowman/hermes-agent
that referenced
this pull request
Jul 13, 2026
…hropic models (NousResearch#43436) * fix(openrouter): route reasoning_effort to verbosity for adaptive Anthropic models Reasoning-mandatory Anthropic models (Claude 4.6+/fable/mythos-class) over OpenRouter ignore reasoning.effort and use adaptive thinking. NousResearch#42991 correctly stopped Hermes from sending a reasoning field to them (it 400s), but put nothing in its place — leaving agent.reasoning_effort a silent no-op on the OpenRouter path: the model always ran at its adaptive default (high) regardless of config. OpenRouter honors the requested effort on the top-level verbosity field instead (maps to Anthropic output_config.effort). Route the existing reasoning_config[effort] there for these models while still never emitting a reasoning field, preserving the NousResearch#42991 fix. No new config arg — the value the user already sets via agent.reasoning_effort now flows to verbosity. - low/medium/high/xhigh/max pass through verbatim (OpenRouter accepts the extended scale for Claude; verified live HTTP 200 + monotonic token spend). - effort unset/none/disabled omits verbosity so the model keeps its default. - native Anthropic transport already correct; unchanged. Fixes NousResearch#43432 * test(openrouter): cover real effort range (add minimal, frame max as passthrough) Adversarial review noted the verbosity tests looped over 'max' — a value parse_reasoning_effort can never produce — while omitting 'minimal', which it can. Align the routing test with the real config range (VALID_REASONING_EFFORTS = minimal/low/medium/high/xhigh) and keep a separate value-agnostic passthrough test that documents why xhigh/max must survive verbatim (TypedDict, no runtime literal validation; OpenRouter accepts the extended scale for Claude). * docs: explain reasoning_effort -> verbosity routing for adaptive Anthropic models Document that reasoning_effort transparently maps to OpenRouter's verbosity field for adaptive-thinking Anthropic models (Claude 4.6+/Fable/Mythos), where reasoning.effort is ignored. Note xhigh is the configurable ceiling (max is wire- only). Add verbosity as a top-level-kwarg example in the provider-plugin guide.
Gravezzz
pushed a commit
to Gravezzz/hermes-agent
that referenced
this pull request
Jul 21, 2026
…ract (NousResearch#42991) New Anthropic models without a recognized version substring (claude-fable-5 and future named/numbered releases) were classified as legacy and routed down the manual-thinking path, which made OpenRouter emit thinking.type.disabled — a form reasoning-mandatory Claude models reject with a non-retryable HTTP 400. Invert the brittle version-substring allowlists to default-to-modern (mirroring _get_anthropic_max_output): unknown Claude models get the adaptive/xhigh/ no-sampling contract, with an explicit legacy list for older families. Non-Claude Anthropic-Messages models (minimax, qwen3, …) keep the manual path. - anthropic_adapter: _supports_adaptive_thinking / _supports_xhigh_effort / _forbids_sampling_params now default unknown Claude models to modern; legacy families enumerated in _LEGACY_MANUAL_THINKING_CLAUDE_SUBSTRINGS. - openrouter profile: omit reasoning entirely (→ adaptive default) instead of forwarding {enabled:false} for reasoning-mandatory Anthropic models; legacy Anthropic + all non-Anthropic models still pass the disable form through. - model_metadata + output-limit table: register claude-fable-5 (1M ctx, 128K out). Tests assert the invariant ("unknown Claude model -> modern contract; legacy stays manual; non-Claude unaffected"), not specific model names.
Gravezzz
pushed a commit
to Gravezzz/hermes-agent
that referenced
this pull request
Jul 21, 2026
…dels (NousResearch#43012) The previous fix (NousResearch#42991) only omitted reasoning when it was being disabled. But reasoning-mandatory Anthropic models (Claude 4.6+, fable) 400 with thinking.type.disabled on EVERY tool-continuation turn even when reasoning is enabled: chat_completions never replays signed thinking blocks, so the prior assistant tool_call has no thinking, and OpenRouter resolves "reasoning requested but history has none" by emitting thinking.type.disabled — which these models reject. Result: first turn works, every turn after the first tool call dies (HTTP 400, non-retryable). OpenRouter ignores reasoning.effort for adaptive Anthropic models anyway (the model self-decides), so the reasoning field is pointless for them on every turn and harmful on tool-replay turns. Omit it entirely → adaptive default. - openrouter profile: drop the reasoning field for reasoning-mandatory Anthropic models regardless of enabled/disabled; legacy Anthropic + non-Anthropic models unchanged. - tests: assert omission across enabled/disabled/effort variants; parity tests switched to a non-Anthropic reasoning model (deepseek) since Anthropic 4.6+ no longer carries a reasoning field. Verified live end-to-end: a tool-replay turn on anthropic/claude-fable-5 with reasoning enabled now builds extra_body=None and returns HTTP 200 (was 400).
Gravezzz
pushed a commit
to Gravezzz/hermes-agent
that referenced
this pull request
Jul 21, 2026
…hropic models (NousResearch#43436) * fix(openrouter): route reasoning_effort to verbosity for adaptive Anthropic models Reasoning-mandatory Anthropic models (Claude 4.6+/fable/mythos-class) over OpenRouter ignore reasoning.effort and use adaptive thinking. NousResearch#42991 correctly stopped Hermes from sending a reasoning field to them (it 400s), but put nothing in its place — leaving agent.reasoning_effort a silent no-op on the OpenRouter path: the model always ran at its adaptive default (high) regardless of config. OpenRouter honors the requested effort on the top-level verbosity field instead (maps to Anthropic output_config.effort). Route the existing reasoning_config[effort] there for these models while still never emitting a reasoning field, preserving the NousResearch#42991 fix. No new config arg — the value the user already sets via agent.reasoning_effort now flows to verbosity. - low/medium/high/xhigh/max pass through verbatim (OpenRouter accepts the extended scale for Claude; verified live HTTP 200 + monotonic token spend). - effort unset/none/disabled omits verbosity so the model keeps its default. - native Anthropic transport already correct; unchanged. Fixes NousResearch#43432 * test(openrouter): cover real effort range (add minimal, frame max as passthrough) Adversarial review noted the verbosity tests looped over 'max' — a value parse_reasoning_effort can never produce — while omitting 'minimal', which it can. Align the routing test with the real config range (VALID_REASONING_EFFORTS = minimal/low/medium/high/xhigh) and keep a separate value-agnostic passthrough test that documents why xhigh/max must survive verbatim (TypedDict, no runtime literal validation; OpenRouter accepts the extended scale for Claude). * docs: explain reasoning_effort -> verbosity routing for adaptive Anthropic models Document that reasoning_effort transparently maps to OpenRouter's verbosity field for adaptive-thinking Anthropic models (Claude 4.6+/Fable/Mythos), where reasoning.effort is ignored. Note xhigh is the configurable ceiling (max is wire- only). Add verbosity as a top-level-kwarg example in the provider-plugin guide.
leewenjie
pushed a commit
to leewenjie/hermes-agent
that referenced
this pull request
Aug 7, 2026
…ract (NousResearch#42991) New Anthropic models without a recognized version substring (claude-fable-5 and future named/numbered releases) were classified as legacy and routed down the manual-thinking path, which made OpenRouter emit thinking.type.disabled — a form reasoning-mandatory Claude models reject with a non-retryable HTTP 400. Invert the brittle version-substring allowlists to default-to-modern (mirroring _get_anthropic_max_output): unknown Claude models get the adaptive/xhigh/ no-sampling contract, with an explicit legacy list for older families. Non-Claude Anthropic-Messages models (minimax, qwen3, …) keep the manual path. - anthropic_adapter: _supports_adaptive_thinking / _supports_xhigh_effort / _forbids_sampling_params now default unknown Claude models to modern; legacy families enumerated in _LEGACY_MANUAL_THINKING_CLAUDE_SUBSTRINGS. - openrouter profile: omit reasoning entirely (→ adaptive default) instead of forwarding {enabled:false} for reasoning-mandatory Anthropic models; legacy Anthropic + all non-Anthropic models still pass the disable form through. - model_metadata + output-limit table: register claude-fable-5 (1M ctx, 128K out). Tests assert the invariant ("unknown Claude model -> modern contract; legacy stays manual; non-Claude unaffected"), not specific model names.
leewenjie
pushed a commit
to leewenjie/hermes-agent
that referenced
this pull request
Aug 7, 2026
…dels (NousResearch#43012) The previous fix (NousResearch#42991) only omitted reasoning when it was being disabled. But reasoning-mandatory Anthropic models (Claude 4.6+, fable) 400 with thinking.type.disabled on EVERY tool-continuation turn even when reasoning is enabled: chat_completions never replays signed thinking blocks, so the prior assistant tool_call has no thinking, and OpenRouter resolves "reasoning requested but history has none" by emitting thinking.type.disabled — which these models reject. Result: first turn works, every turn after the first tool call dies (HTTP 400, non-retryable). OpenRouter ignores reasoning.effort for adaptive Anthropic models anyway (the model self-decides), so the reasoning field is pointless for them on every turn and harmful on tool-replay turns. Omit it entirely → adaptive default. - openrouter profile: drop the reasoning field for reasoning-mandatory Anthropic models regardless of enabled/disabled; legacy Anthropic + non-Anthropic models unchanged. - tests: assert omission across enabled/disabled/effort variants; parity tests switched to a non-Anthropic reasoning model (deepseek) since Anthropic 4.6+ no longer carries a reasoning field. Verified live end-to-end: a tool-replay turn on anthropic/claude-fable-5 with reasoning enabled now builds extra_body=None and returns HTTP 200 (was 400).
leewenjie
pushed a commit
to leewenjie/hermes-agent
that referenced
this pull request
Aug 7, 2026
…hropic models (NousResearch#43436) * fix(openrouter): route reasoning_effort to verbosity for adaptive Anthropic models Reasoning-mandatory Anthropic models (Claude 4.6+/fable/mythos-class) over OpenRouter ignore reasoning.effort and use adaptive thinking. NousResearch#42991 correctly stopped Hermes from sending a reasoning field to them (it 400s), but put nothing in its place — leaving agent.reasoning_effort a silent no-op on the OpenRouter path: the model always ran at its adaptive default (high) regardless of config. OpenRouter honors the requested effort on the top-level verbosity field instead (maps to Anthropic output_config.effort). Route the existing reasoning_config[effort] there for these models while still never emitting a reasoning field, preserving the NousResearch#42991 fix. No new config arg — the value the user already sets via agent.reasoning_effort now flows to verbosity. - low/medium/high/xhigh/max pass through verbatim (OpenRouter accepts the extended scale for Claude; verified live HTTP 200 + monotonic token spend). - effort unset/none/disabled omits verbosity so the model keeps its default. - native Anthropic transport already correct; unchanged. Fixes NousResearch#43432 * test(openrouter): cover real effort range (add minimal, frame max as passthrough) Adversarial review noted the verbosity tests looped over 'max' — a value parse_reasoning_effort can never produce — while omitting 'minimal', which it can. Align the routing test with the real config range (VALID_REASONING_EFFORTS = minimal/low/medium/high/xhigh) and keep a separate value-agnostic passthrough test that documents why xhigh/max must survive verbatim (TypedDict, no runtime literal validation; OpenRouter accepts the extended scale for Claude). * docs: explain reasoning_effort -> verbosity routing for adaptive Anthropic models Document that reasoning_effort transparently maps to OpenRouter's verbosity field for adaptive-thinking Anthropic models (Claude 4.6+/Fable/Mythos), where reasoning.effort is ignored. Note xhigh is the configurable ceiling (max is wire- only). Add verbosity as a top-level-kwarg example in the provider-plugin guide.
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.
Summary
New Anthropic models (claude-fable-5 and future named/numbered releases) now route correctly instead of dying with a non-retryable HTTP 400 on the first API call.
Root cause: Hermes decided Anthropic thinking behavior by substring-matching the model name against hardcoded version numbers (
"4.6","4.7","4.8").claude-fable-5contains none of them, so it fell through to the legacy manual-thinking path. Over OpenRouter that path forwardsreasoning: {enabled: false}when reasoning is off, which OpenRouter translates into Anthropic'sthinking: {type: "disabled"}— a form reasoning-mandatory Claude models (4.6+, fable-class) reject outright ("thinking.type.disabled" is not supported for this model).The defect is the asymmetry vs.
_get_anthropic_max_output, which already defaults unknown models to the newest behavior. The thinking-mode classifiers did the opposite — defaulting unknown models to legacy — so every new un-numbered model re-breaks.Changes
agent/anthropic_adapter.py: invert the version-substring allowlists._supports_adaptive_thinking/_supports_xhigh_effort/_forbids_sampling_paramsnow default unknown Claude models to the modern contract, with an explicit legacy list (_LEGACY_MANUAL_THINKING_CLAUDE_SUBSTRINGS) for older families. Non-Claude Anthropic-Messages models (minimax, qwen3, …) stay on the manual path.plugins/model-providers/openrouter/__init__.py: for reasoning-mandatory Anthropic models, omitreasoningentirely (→ adaptive default) instead of forwarding{enabled: false}. Legacy Anthropic + all non-Anthropic models still pass the disable form through unchanged.agent/model_metadata.py+_ANTHROPIC_OUTPUT_LIMITS: registerclaude-fable-5(1M context, 128K output).No
"fable"magic string in any behavior path — the durable fix is the default direction, so the next un-numbered release works with zero code changes.Validation
Live OpenRouter calls against
anthropic/claude-fable-5:{enabled: false})thinking.type.disabledeffort: none)Tests assert the invariant ("unknown Claude → modern; legacy stays manual; 4.6 adaptive-but-not-xhigh; non-Claude unaffected"), not specific model names.
tests/agent/test_anthropic_adapter.py+tests/providers/— 258 passed; full reasoning/thinking sweep across run_agent/model_metadata/fast_command — 301 passed.