Conversation
Adds a `none` field to `ReasoningDetails::Leveled` to indicate that a model supports completely disabling reasoning. This is now the first candidate returned by `lowest_effort()`, sitting below `xlow`. Several new OpenAI models are added: `gpt-5.4`, `gpt-5.4-pro`, `gpt-5.4-mini`, `gpt-5.4-nano`, `gpt-5.3-codex`, and `gpt-5-pro`. All of these carry the new `TEMP_REQUIRES_NO_REASONING` feature flag, which is also applied retroactively to the existing GPT-5 and GPT-5.1/ 5.2 families. When this flag is set and reasoning effort is anything other than `none`, `temperature` and `top_p` are stripped from the request and a warning is emitted — matching the constraint the OpenAI API enforces for these models. Two further improvements land for the OpenAI provider. Response verbosity (`low`/`medium`/`high`) can now be set via the catch-all `parameters.other.verbosity` config key, and is forwarded to the API through `TextConfig`. Message phase (`commentary`/`final_answer`) is now round-tripped; it is persisted to conversation metadata under the `openai_phase` key when a response is received, and restored when that message is later sent back as input context. The Google provider is updated to handle the new `none` field in pattern matches, and some model reasoning capability flags are corrected to better reflect reality. Signed-off-by: Jean Mertz <git@jeanmertz.com>
Signed-off-by: Jean Mertz <git@jeanmertz.com>
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds a
nonefield toReasoningDetails::Leveledto indicate that a model supports completely disabling reasoning. This is now the first candidate returned bylowest_effort(), sitting belowxlow.Several new OpenAI models are added:
gpt-5.4,gpt-5.4-pro,gpt-5.4-mini,gpt-5.4-nano,gpt-5.3-codex, andgpt-5-pro. All of these carry the newTEMP_REQUIRES_NO_REASONINGfeature flag, which is also applied retroactively to the existing GPT-5 and GPT-5.1/ 5.2 families. When this flag is set and reasoning effort is anything other thannone,temperatureandtop_pare stripped from the request and a warning is emitted — matching the constraint the OpenAI API enforces for these models.Two further improvements land for the OpenAI provider. Response verbosity (
low/medium/high) can now be set via the catch-allparameters.other.verbosityconfig key, and is forwarded to the API throughTextConfig. Message phase (commentary/final_answer) is now round-tripped; it is persisted to conversation metadata under theopenai_phasekey when a response is received, and restored when that message is later sent back as input context.The Google provider is updated to handle the new
nonefield in pattern matches, and some model reasoning capability flags are corrected to better reflect reality.