dot-inference 2.2.0__tar.gz → 3.0.0__tar.gz

{dot_inference-2.2.0 → dot_inference-3.0.0}/.pre-commit-config.yaml

@@ -1,6 +1,6 @@
-# Verbatim legal texts: the AGPL and the DCO both forbid modification,
-# so no hook may normalise their whitespace or line endings.
-exclude: '^(LICENSE\.md|DCO)$'
+# Verbatim legal text (MIT licence): no hook may normalise its whitespace
+# or line endings.
+exclude: '^LICENSE\.md$'
 
 repos:
   - repo: https://github.com/pre-commit/pre-commit-hooks

{dot_inference-2.2.0 → dot_inference-3.0.0}/AUTHORS.md

@@ -7,6 +7,7 @@ here does not confer any ownership of, or rights in, the project.
 ## Kannon For Deep Tech
 
 - Adrien Marjollet
+- Charles Capon
 - Clément Olivier
 - Jean Fechter
 - Louis Le Tarnec

{dot_inference-2.2.0 → dot_inference-3.0.0}/CHANGELOG.md

@@ -7,6 +7,29 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [3.0.0] - 2026-10-15
+
+### Added
+
+- Registry: `supports_temperature` (default `true`, can be overridden per
+  provider), independent of `supports_reasoning`. OpenAI's `gpt-5*` and
+  `gpt-6*` models are marked `false`: their API accepts only the default
+  temperature (1).
+
+### Changed
+
+- **Breaking** — temperature no longer depends on reasoning: a temperature is
+  sent whenever one is set, reasoning model or not (gpt-oss, Claude, Mistral,
+  Qwen… honour it), and dropped only for a model whose registry entry says
+  `supports_temperature: false`. A per-call `temperature=` is treated the
+  same way.
+- **Breaking** — `GenerationSettings.temperature` defaults to `None` instead of
+  `0`: unset, no temperature is sent and the provider applies its own default.
+  Set `DOTI_GENERATION__TEMPERATURE=0` to keep the previous behaviour for
+  non-reasoning models.
+- Licence: from `AGPL-3.0-or-later` to `MIT`. Releases already published
+  stay under the AGPL.
+
 ## [2.2.0] - 2026-10-08
 
 ### Added
@@ -107,7 +130,8 @@ this release.
 - `LLMCallable` and `EmbeddingCallable` return a `Coroutine` rather than the
   looser `Awaitable`, matching what `asyncio.run_coroutine_threadsafe` accepts.
 
-[Unreleased]: https://gitlab.com/deepika6190303/deepika-open-toolbox/dot-inference/-/compare/v2.2.0...main
+[Unreleased]: https://gitlab.com/deepika6190303/deepika-open-toolbox/dot-inference/-/compare/v3.0.0...main
+[3.0.0]: https://gitlab.com/deepika6190303/deepika-open-toolbox/dot-inference/-/tags/v3.0.0
 [2.2.0]: https://gitlab.com/deepika6190303/deepika-open-toolbox/dot-inference/-/tags/v2.2.0
 [2.1.0]: https://gitlab.com/deepika6190303/deepika-open-toolbox/dot-inference/-/tags/v2.1.0
 [2.0.3]: https://gitlab.com/deepika6190303/deepika-open-toolbox/dot-inference/-/tags/v2.0.3

dot_inference-3.0.0/CONTRIBUTING.md

@@ -0,0 +1,37 @@
+# Contributing
+
+Thanks for your interest in this project. Contributions are welcome.
+
+This project is maintained by **Kannon For Deep Tech** and released under the
+MIT License (`MIT`).
+
+## 1. Licensing of your contribution
+
+By contributing to this project, you agree that your contribution is licensed
+under the MIT License, the same terms as the rest of the project. You remain
+the author and copyright holder of your contribution.
+
+You confirm that you are entitled to do so: that your contribution is your
+original work or that you hold the necessary rights over it, and that it does
+not infringe any third-party rights. If you are contributing in the course of
+your employment or under a service agreement, you confirm that your employer or
+client has authorised you to contribute under these terms.
+
+## 2. Contractors and employees of Kannon For Deep Tech
+
+If you are working for Kannon For Deep Tech as an employee or under a service
+agreement, the assignment of rights set out in your contract governs, and takes
+precedence over this file.
+
+## 3. Submitting a change
+
+1. Open an issue first for anything substantial, so we can agree on the approach.
+2. Fork, branch, and keep commits focused.
+3. Add or update tests where relevant.
+4. If you would like to be credited individually, add yourself to
+   [AUTHORS.md](AUTHORS.md) in the same pull request.
+5. Open a pull request.
+
+## 4. Questions
+
+For anything about licensing, contact <louis.letarnec@deepika.ai>.

dot_inference-3.0.0/LICENSE.md

@@ -0,0 +1,21 @@
+# MIT License
+
+Copyright (c) 2026 Kannon For Deep Tech
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

{dot_inference-2.2.0 → dot_inference-3.0.0}/PKG-INFO

@@ -1,12 +1,12 @@
 Metadata-Version: 2.5
 Name: dot-inference
-Version: 2.2.0
+Version: 3.0.0
 Summary: Unified async Python library for LLM inference and embeddings across multiple providers
 Project-URL: Homepage, https://gitlab.com/deepika6190303/deepika-open-toolbox/dot-inference
 Project-URL: Repository, https://gitlab.com/deepika6190303/deepika-open-toolbox/dot-inference
 Project-URL: Issues, https://gitlab.com/deepika6190303/deepika-open-toolbox/dot-inference/-/issues
 Author-email: Kannon For Deep Tech <louis.letarnec@deepika.ai>
-License-Expression: AGPL-3.0-or-later
+License-Expression: MIT
 License-File: LICENSE.md
 Keywords: embeddings,inference,llm,providers
 Classifier: Development Status :: 3 - Alpha
@@ -29,7 +29,7 @@ Description-Content-Type: text/markdown
 
 [![PyPI](https://img.shields.io/pypi/v/dot-inference)](https://pypi.org/project/dot-inference/)
 ![Python Version](https://img.shields.io/badge/python-3.12%2B-blue)
-[![Licence: AGPL v3](https://img.shields.io/badge/licence-AGPL--3.0--or--later-blue)](LICENSE.md)
+[![Licence: MIT](https://img.shields.io/badge/licence-MIT-blue)](LICENSE.md)
 [![Pipeline](https://gitlab.com/deepika6190303/deepika-open-toolbox/dot-inference/badges/main/pipeline.svg)](https://gitlab.com/deepika6190303/deepika-open-toolbox/dot-inference/-/pipelines)
 
 **One async interface for LLM inference and embeddings, across every provider.**
@@ -313,7 +313,7 @@ package is covered, anything underscore-prefixed is internal and may change in
 any release. Public names are never removed without a deprecation period.
 
 ```toml
-dependencies = ["dot-inference>=2.0,<3"]
+dependencies = ["dot-inference>=3.0,<4"]
 ```
 
 See [docs/VERSIONING.md](docs/VERSIONING.md) for the full policy.
@@ -337,16 +337,12 @@ See [docs/VERSIONING.md](docs/VERSIONING.md) for the full policy.
 
 ## Contributing
 
-Contributions are welcome. See [CONTRIBUTING.md](CONTRIBUTING.md) for the DCO
-sign-off requirement, the licensing terms that apply to contributions, and how to
-submit a change.
+Contributions are welcome. See [CONTRIBUTING.md](CONTRIBUTING.md) for the
+licensing terms that apply to contributions and how to submit a change.
 
 ## Licence
 
 Copyright (C) 2026 Kannon For Deep Tech (deepika)
 
-This software is distributed under the GNU Affero General Public License,
-version 3 or later — see [LICENSE.md](LICENSE.md).
-
-A commercial licence is available for use in proprietary environments.
-Contact: louis.letarnec@deepika.ai
+This software is distributed under the MIT License — see
+[LICENSE.md](LICENSE.md).

{dot_inference-2.2.0 → dot_inference-3.0.0}/README.md

@@ -2,7 +2,7 @@
 
 [![PyPI](https://img.shields.io/pypi/v/dot-inference)](https://pypi.org/project/dot-inference/)
 ![Python Version](https://img.shields.io/badge/python-3.12%2B-blue)
-[![Licence: AGPL v3](https://img.shields.io/badge/licence-AGPL--3.0--or--later-blue)](LICENSE.md)
+[![Licence: MIT](https://img.shields.io/badge/licence-MIT-blue)](LICENSE.md)
 [![Pipeline](https://gitlab.com/deepika6190303/deepika-open-toolbox/dot-inference/badges/main/pipeline.svg)](https://gitlab.com/deepika6190303/deepika-open-toolbox/dot-inference/-/pipelines)
 
 **One async interface for LLM inference and embeddings, across every provider.**
@@ -286,7 +286,7 @@ package is covered, anything underscore-prefixed is internal and may change in
 any release. Public names are never removed without a deprecation period.
 
 ```toml
-dependencies = ["dot-inference>=2.0,<3"]
+dependencies = ["dot-inference>=3.0,<4"]
 ```
 
 See [docs/VERSIONING.md](docs/VERSIONING.md) for the full policy.
@@ -310,16 +310,12 @@ See [docs/VERSIONING.md](docs/VERSIONING.md) for the full policy.
 
 ## Contributing
 
-Contributions are welcome. See [CONTRIBUTING.md](CONTRIBUTING.md) for the DCO
-sign-off requirement, the licensing terms that apply to contributions, and how to
-submit a change.
+Contributions are welcome. See [CONTRIBUTING.md](CONTRIBUTING.md) for the
+licensing terms that apply to contributions and how to submit a change.
 
 ## Licence
 
 Copyright (C) 2026 Kannon For Deep Tech (deepika)
 
-This software is distributed under the GNU Affero General Public License,
-version 3 or later — see [LICENSE.md](LICENSE.md).
-
-A commercial licence is available for use in proprietary environments.
-Contact: louis.letarnec@deepika.ai
+This software is distributed under the MIT License — see
+[LICENSE.md](LICENSE.md).

{dot_inference-2.2.0 → dot_inference-3.0.0}/docs/ARCHITECTURE.md

@@ -148,7 +148,7 @@ sequenceDiagram
     App->>RateLimit: await client.call(messages)
     RateLimit->>RateLimit: Sliding-window check / wait
     RateLimit->>Compat: forward call
-    Compat->>Compat: Build request kwargs (temperature or reasoning_effort)
+    Compat->>Compat: Build request kwargs (reasoning_effort if the model reasons, temperature if set and the model accepts it)
 
     loop retry up to 4 attempts on transient errors
         Compat->>API: POST /chat/completions

{dot_inference-2.2.0 → dot_inference-3.0.0}/docs/DEVELOPMENT.md

@@ -1,8 +1,8 @@
 # Development Guide
 
 > Looking to contribute? Start with [CONTRIBUTING.md](../CONTRIBUTING.md) at the
-> repository root: it covers the DCO sign-off and the licensing terms your
-> contribution falls under. This guide covers the practical side.
+> repository root: it covers the licensing terms your contribution
+> falls under. This guide covers the practical side.
 
 ## Requirements
 
@@ -53,11 +53,6 @@ and re-run in CI, so most of the style is applied for you.
 - Use meaningful variable and function names
 - Add docstrings for public functions and classes
 - Keep functions focused and small
-- Add the SPDX headers to any new file:
-  ```python
-  # SPDX-FileCopyrightText: Kannon For Deep Tech and contributors
-  # SPDX-License-Identifier: AGPL-3.0-or-later
-  ```
 
 ## Testing
 

{dot_inference-2.2.0 → dot_inference-3.0.0}/docs/VERSIONING.md

@@ -1,7 +1,7 @@
 # Versioning and stability
 
 `dot-inference` follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
-This page says what that actually covers, so `>=2.0,<3` means something concrete.
+This page says what that actually covers, so `>=3.0,<4` means something concrete.
 
 ## What is public
 
@@ -94,11 +94,11 @@ manual. Use them to validate a release without committing to it.
 ### From PyPI — the supported way
 
 ```toml
-dependencies = ["dot-inference>=2.0,<3"]
+dependencies = ["dot-inference>=3.0,<4"]
 ```
 
 The upper bound is what makes the guarantees above useful: you get every fix and
-feature of the 2.x line, and never an unannounced break.
+feature of the 3.x line, and never an unannounced break.
 
 ### From git — development only
 

{dot_inference-2.2.0 → dot_inference-3.0.0}/examples/demo.py

@@ -1,6 +1,3 @@
-# SPDX-FileCopyrightText: Kannon For Deep Tech
-# SPDX-License-Identifier: AGPL-3.0-or-later
-
 """
 dot-inference demo script
 

{dot_inference-2.2.0 → dot_inference-3.0.0}/examples/demo_streaming.py

@@ -1,6 +1,3 @@
-# SPDX-FileCopyrightText: Kannon For Deep Tech
-# SPDX-License-Identifier: AGPL-3.0-or-later
-
 """
 dot-inference streaming demo
 

{dot_inference-2.2.0 → dot_inference-3.0.0}/pyproject.toml

@@ -1,10 +1,10 @@
 [project]
 name = "dot-inference"
-version = "2.2.0"
+version = "3.0.0"
 description = "Unified async Python library for LLM inference and embeddings across multiple providers"
 readme = "README.md"
 authors = [{ name = "Kannon For Deep Tech", email = "louis.letarnec@deepika.ai" }]
-license = "AGPL-3.0-or-later"
+license = "MIT"
 license-files = ["LICENSE.md"]
 keywords = ["llm", "inference", "embeddings", "providers"]
 

{dot_inference-2.2.0 → dot_inference-3.0.0}/scripts/test_all_models.py

@@ -1,6 +1,3 @@
-# SPDX-FileCopyrightText: Kannon For Deep Tech
-# SPDX-License-Identifier: AGPL-3.0-or-later
-
 """Ad-hoc live smoke test for every model × provider route in the registry.
 
 Makes a minimal real call per (model, provider) where a key is available,

{dot_inference-2.2.0 → dot_inference-3.0.0}/src/dot_inference/__init__.py

@@ -1,6 +1,3 @@
-# SPDX-FileCopyrightText: Kannon For Deep Tech
-# SPDX-License-Identifier: AGPL-3.0-or-later
-
 from dot_inference._client import (
     EmbeddingClient,
     LLMClient,

{dot_inference-2.2.0 → dot_inference-3.0.0}/src/dot_inference/_client.py

@@ -1,6 +1,3 @@
-# SPDX-FileCopyrightText: Kannon For Deep Tech
-# SPDX-License-Identifier: AGPL-3.0-or-later
-
 """Class-based LLM client interface.
 
 Provides ``LLMClient`` and ``EmbeddingClient`` — abstract base classes with
@@ -371,10 +368,10 @@ def _warn_compatibility(
     # separately in LLMClient._resolve_reasoning_effort.
     _validate_reasoning_effort_for_model(model, reasoning_effort, config.supports_reasoning)
 
-    if config.supports_reasoning and temperature:
+    if not config.supports_temperature and temperature is not None:
         logger.warning(
-            "Model '%s' is a reasoning model — the temperature setting (%.2f) "
-            "will be ignored in favor of reasoning_effort.",
+            "Model '%s' does not accept a temperature — the temperature setting "
+            "(%.2f) will not be sent.",
             model,
             temperature,
         )

{dot_inference-2.2.0 → dot_inference-3.0.0}/src/dot_inference/_cost_tracker.py

@@ -1,6 +1,3 @@
-# SPDX-FileCopyrightText: Kannon For Deep Tech
-# SPDX-License-Identifier: AGPL-3.0-or-later
-
 """Cost and token usage tracker for aggregating across multiple LLM calls."""
 
 from __future__ import annotations

{dot_inference-2.2.0 → dot_inference-3.0.0}/src/dot_inference/_llm_registry.py

@@ -1,6 +1,3 @@
-# SPDX-FileCopyrightText: Kannon For Deep Tech
-# SPDX-License-Identifier: AGPL-3.0-or-later
-
 """Model registry — loads model configurations from models.yaml.
 
 The built-in ``models.yaml`` ships with the package.  Users can extend or
@@ -31,6 +28,8 @@ class ProviderModelConfig:
     # Useful when reasoning works through one provider (e.g. OpenRouter) but
     # not another (e.g. native SDK doesn't expose the parameter yet).
     supports_reasoning: bool | None = None
+    # Per-provider override for temperature support (None = inherit from model).
+    supports_temperature: bool | None = None
     # OpenRouter-specific features (ignored by other providers)
     response_healing: bool = False
     fallback_models: list[str] = field(default_factory=list)
@@ -44,6 +43,11 @@ class ModelConfig:
     model_id: str
     supports_strict_json: bool = True
     supports_reasoning: bool = False
+    # Whether the API accepts a ``temperature`` parameter. Independent of
+    # ``supports_reasoning``: gpt-oss reasons and honours temperature, while
+    # OpenAI's gpt-5 family rejects any value other than the default (1).
+    # When False, a configured temperature is dropped from the request.
+    supports_temperature: bool = True
     requires_cache_control: bool = False
     input_token_cost: float | None = None
     output_token_cost: float | None = None
@@ -75,6 +79,9 @@ class ModelConfig:
             supports_reasoning=pmc.supports_reasoning
             if pmc.supports_reasoning is not None
             else self.supports_reasoning,
+            supports_temperature=pmc.supports_temperature
+            if pmc.supports_temperature is not None
+            else self.supports_temperature,
             requires_cache_control=self.requires_cache_control,
             input_token_cost=pmc.input_token_cost
             if pmc.input_token_cost is not None
@@ -98,6 +105,7 @@ class ResolvedModelConfig:
     api_model_name: str
     supports_strict_json: bool = True
     supports_reasoning: bool = False
+    supports_temperature: bool = True
     requires_cache_control: bool = False
     input_token_cost: float | None = None
     output_token_cost: float | None = None
@@ -130,6 +138,7 @@ def _parse_models_yaml(raw: dict[str, Any]) -> dict[str, ModelConfig]:
                 input_token_cost=_to_float(prov_cfg.get("input_token_cost")),
                 output_token_cost=_to_float(prov_cfg.get("output_token_cost")),
                 supports_reasoning=prov_cfg.get("supports_reasoning"),
+                supports_temperature=prov_cfg.get("supports_temperature"),
                 response_healing=features.get("response_healing", False),
                 fallback_models=features.get("fallback_models", []),
                 provider_routing=features.get("provider_routing"),
@@ -139,6 +148,7 @@ def _parse_models_yaml(raw: dict[str, Any]) -> dict[str, ModelConfig]:
             model_id=model_id,
             supports_strict_json=cfg.get("supports_strict_json", True),
             supports_reasoning=cfg.get("supports_reasoning", False),
+            supports_temperature=cfg.get("supports_temperature", True),
             requires_cache_control=cfg.get("requires_cache_control", False),
             input_token_cost=_to_float(cfg.get("input_token_cost")),
             output_token_cost=_to_float(cfg.get("output_token_cost")),

{dot_inference-2.2.0 → dot_inference-3.0.0}/src/dot_inference/_local_check.py

@@ -1,6 +1,3 @@
-# SPDX-FileCopyrightText: Kannon For Deep Tech
-# SPDX-License-Identifier: AGPL-3.0-or-later
-
 """Sanity check for the ``local`` provider endpoint.
 
 Single-model OpenAI-compatible servers (SGLang, vLLM in single-model mode)

{dot_inference-2.2.0 → dot_inference-3.0.0}/src/dot_inference/_mistral_compat.py

@@ -1,6 +1,3 @@
-# SPDX-FileCopyrightText: Kannon For Deep Tech
-# SPDX-License-Identifier: AGPL-3.0-or-later
-
 """Shared Mistral-native client.
 
 Provides ``MistralClient`` — a class-based LLM client using the native
@@ -161,9 +158,13 @@ class MistralClient(LLMClient):
             # Mistral reasoning models (Medium 3.5, Small) take reasoning_effort
             # natively on the chat-completion request — La Plateforme accepts
             # "none" (minimal thinking, thinking chunk omitted) and "high"
-            # (full thinking chunk). Temperature is ignored for these.
+            # (full thinking chunk).
             create_kwargs["reasoning_effort"] = reasoning_effort
-        elif not is_reasoning and temperature is not None:
+        # Temperature is independent of reasoning: sent whenever it is set,
+        # unless the registry says the model rejects it (unknown models get it).
+        if temperature is not None and (
+            self._model_config is None or self._model_config.supports_temperature
+        ):
             create_kwargs["temperature"] = temperature
 
         response_model: type[BaseModel] | None = None

{dot_inference-2.2.0 → dot_inference-3.0.0}/src/dot_inference/_openai_compat.py

@@ -1,6 +1,3 @@
-# SPDX-FileCopyrightText: Kannon For Deep Tech
-# SPDX-License-Identifier: AGPL-3.0-or-later
-
 """Shared OpenAI-compatible client.
 
 Provides ``OpenAICompatClient`` — a class-based LLM client for any
@@ -187,11 +184,14 @@ class OpenAICompatClient(LLMClient):
             "messages": [m.model_dump(exclude_none=True) for m in messages],
         }
 
-        # Reasoning models use reasoning_effort instead of temperature
         is_reasoning = self._model_config.supports_reasoning if self._model_config else False
         if is_reasoning and reasoning_effort is not None:
             create_kwargs["reasoning_effort"] = reasoning_effort
-        elif not is_reasoning and temperature is not None:
+        # Temperature is independent of reasoning: sent whenever it is set,
+        # unless the registry says the model rejects it (unknown models get it).
+        if temperature is not None and (
+            self._model_config is None or self._model_config.supports_temperature
+        ):
             create_kwargs["temperature"] = temperature
 
         response_model: type[BaseModel] | None = None

{dot_inference-2.2.0 → dot_inference-3.0.0}/src/dot_inference/_rate_limit.py

@@ -1,6 +1,3 @@
-# SPDX-FileCopyrightText: Kannon For Deep Tech
-# SPDX-License-Identifier: AGPL-3.0-or-later
-
 """Async rate-limiting wrapper for ``LLMCallable``."""
 
 from __future__ import annotations

{dot_inference-2.2.0 → dot_inference-3.0.0}/src/dot_inference/_retry.py

@@ -1,6 +1,3 @@
-# SPDX-FileCopyrightText: Kannon For Deep Tech
-# SPDX-License-Identifier: AGPL-3.0-or-later
-
 """Decides how long to wait before retrying a transient provider error.
 
 Shared by the OpenAI and Mistral clients so both honor ``Retry-After`` the

{dot_inference-2.2.0 → dot_inference-3.0.0}/src/dot_inference/_sync.py

@@ -1,6 +1,3 @@
-# SPDX-FileCopyrightText: Kannon For Deep Tech
-# SPDX-License-Identifier: AGPL-3.0-or-later
-
 """Sync helpers — run async code from synchronous contexts.
 
 Uses a single persistent background event loop in a daemon thread so that:

{dot_inference-2.2.0 → dot_inference-3.0.0}/src/dot_inference/_types.py

@@ -1,6 +1,3 @@
-# SPDX-FileCopyrightText: Kannon For Deep Tech
-# SPDX-License-Identifier: AGPL-3.0-or-later
-
 """Shared types for dot-llm."""
 
 from __future__ import annotations

{dot_inference-2.2.0 → dot_inference-3.0.0}/src/dot_inference/mock.py

@@ -1,6 +1,3 @@
-# SPDX-FileCopyrightText: Kannon For Deep Tech
-# SPDX-License-Identifier: AGPL-3.0-or-later
-
 """Mock clients for testing without API calls."""
 
 from __future__ import annotations

{dot_inference-2.2.0 → dot_inference-3.0.0}/src/dot_inference/models.yaml

@@ -38,6 +38,13 @@
 #     - Can be overridden per provider (see below) when reasoning is
 #       available through one provider but not another.
 #
+#   supports_temperature: true/false (default true)
+#     - Whether the API accepts a temperature parameter. Independent of
+#       supports_reasoning: gpt-oss reasons and honours temperature, while
+#       OpenAI's gpt-5 family accepts only its default (1). When false, a
+#       configured temperature is not sent.
+#     - Can be overridden per provider (see below).
+#
 #   reasoning_effort_options: list[str] (optional)
 #     - The reasoning_effort values this model actually accepts. Lets a
 #       client offer only the valid choices. Defaults to
@@ -78,6 +85,10 @@
 #     - Useful when the model supports reasoning via one provider (e.g.
 #       OpenRouter) but not another (e.g. native SDK lacks the param).
 #
+#   supports_temperature: true/false (optional)
+#     - Per-provider override for temperature support; inherits the
+#       model-level value when omitted.
+#
 #   features (optional, provider-specific):
 #     response_healing: true/false
 #       - OpenRouter plugin that post-processes malformed JSON responses.
@@ -95,6 +106,7 @@
 openai/gpt-5:
   supports_strict_json: true
   supports_reasoning: true
+  supports_temperature: false   # the OpenAI API accepts only its default (1) for this family
   input_token_cost: 1.25e-6   # $1.25/M
   output_token_cost: 1.e-5    # $10/M
   context_window: 400000
@@ -110,6 +122,7 @@ openai/gpt-5:
 openai/gpt-5.1:
   supports_strict_json: true
   supports_reasoning: true
+  supports_temperature: false   # the OpenAI API accepts only its default (1) for this family
   input_token_cost: 1.25e-6   # $1.25/M
   output_token_cost: 1.e-5    # $10/M
   context_window: 400000
@@ -125,6 +138,7 @@ openai/gpt-5.1:
 openai/gpt-5.2:
   supports_strict_json: true
   supports_reasoning: true
+  supports_temperature: false   # the OpenAI API accepts only its default (1) for this family
   input_token_cost: 1.75e-6   # $1.75/M
   output_token_cost: 1.4e-5   # $14/M
   context_window: 400000
@@ -140,6 +154,7 @@ openai/gpt-5.2:
 openai/gpt-5-mini:
   supports_strict_json: true
   supports_reasoning: true
+  supports_temperature: false   # the OpenAI API accepts only its default (1) for this family
   input_token_cost: 2.5e-7    # $0.25/M
   output_token_cost: 2.e-6    # $2/M
   context_window: 400000
@@ -179,6 +194,7 @@ openai/gpt-oss-20b-local:
 openai/gpt-5.4-nano:
   supports_strict_json: true
   supports_reasoning: true
+  supports_temperature: false   # the OpenAI API accepts only its default (1) for this family
   input_token_cost: 2e-7      # $0.20/M
   output_token_cost: 1.25e-6  # $1.25/M
   context_window: 400000
@@ -194,6 +210,7 @@ openai/gpt-5.4-nano:
 openai/gpt-5.4-mini:
   supports_strict_json: true
   supports_reasoning: true
+  supports_temperature: false   # the OpenAI API accepts only its default (1) for this family
   input_token_cost: 7.5e-7    # $0.75/M
   output_token_cost: 4.5e-6   # $4.50/M
   context_window: 400000
@@ -209,6 +226,7 @@ openai/gpt-5.4-mini:
 openai/gpt-5.4:
   supports_strict_json: true
   supports_reasoning: true
+  supports_temperature: false   # the OpenAI API accepts only its default (1) for this family
   input_token_cost: 2.5e-6    # $2.50/M
   output_token_cost: 1.5e-5   # $15/M
   context_window: 1050000
@@ -225,6 +243,7 @@ openai/gpt-5.4:
 openai/gpt-6-luna:
   supports_strict_json: true
   supports_reasoning: true
+  supports_temperature: false   # the OpenAI API accepts only its default (1) for this family
   input_token_cost: 1e-7      # $0.10/M
   output_token_cost: 5e-7     # $0.50/M
   context_window: 1050000
@@ -242,6 +261,7 @@ openai/gpt-6-luna:
 openai/gpt-6.1-sol:
   supports_strict_json: true
   supports_reasoning: true
+  supports_temperature: false   # the OpenAI API accepts only its default (1) for this family
   input_token_cost: 2e-6      # $2/M
   output_token_cost: 1e-5     # $10/M
   context_window: 1050000

{dot_inference-2.2.0 → dot_inference-3.0.0}/src/dot_inference/settings.py

@@ -1,6 +1,3 @@
-# SPDX-FileCopyrightText: Kannon For Deep Tech
-# SPDX-License-Identifier: AGPL-3.0-or-later
-
 """Settings for dot-inference configuration via environment variables.
 
 All environment variables are prefixed with ``DOTI_`` and use ``__`` as
@@ -85,7 +82,8 @@ class GenerationSettings(BaseModel):
 
     mock: bool = False
     provider: Provider = Provider.OPENROUTER
-    temperature: float = 0
+    # None: no temperature is sent, the provider applies its own default.
+    temperature: float | None = None
     max_tokens: int | None = None
     reasoning_effort: ReasoningEffort | None = ReasoningEffort.MEDIUM
     calls_per_minute_limit: int = 0  # no limit by default

{dot_inference-2.2.0 → dot_inference-3.0.0}/tests/test_cost_tracker.py

@@ -1,6 +1,3 @@
-# SPDX-FileCopyrightText: Kannon For Deep Tech
-# SPDX-License-Identifier: AGPL-3.0-or-later
-
 """Tests for dot_inference._cost_tracker."""
 
 import pytest

{dot_inference-2.2.0 → dot_inference-3.0.0}/tests/test_factories.py

@@ -1,6 +1,3 @@
-# SPDX-FileCopyrightText: Kannon For Deep Tech
-# SPDX-License-Identifier: AGPL-3.0-or-later
-
 """Tests for get_llm_client and get_embedding_client factories."""
 
 import pytest