Releases: lofcz/LlmTornado
Release v3.5.1
What's Changed
This release contains a fix for file uploading to Google's API.
- google: fix file upload
- Update README.md
- Update README.md
- Update README.md
- Update README.md
- Update project versions to v3.5.0
Updated Projects
LlmTornado.Demo.csproj -> 0.0.1
LlmTornado.csproj -> 3.5.1
Release v3.5.0
What's Changed
This release adds LlmTornado.Toolkit package.
- toolkit branding
- toolkit: chat
- Update README.md
- Update project versions to v3.4.22
Updated Projects
LlmTornado.Toolkit.csproj -> 0.1.0
LlmTornado.Demo.csproj -> 0.1.0
LlmTornado.csproj -> 3.5.0
Release v3.4.22
What's Changed
This release adds support for gemini-2.0-flash-preview-image-generation, mistral-medium-2505, gemini-2.5-pro-preview-03-25, and the newly introduced web_search_20250305 tool in Anthropic's API.
- add gemini-2.0-flash-preview-image-generation to image gen models
- google: gemini-2.0-flash-preview-image-generation
- anthropic: web_search_20250305
- mistral: mistral-medium-2505
- google: gemini-2.5-pro-preview-03-25
- Update project versions to v3.4.21
Updated Projects
LlmTornado.Demo.csproj -> 0.0.1
LlmTornado.csproj -> 3.4.22
Release v3.4.21
What's Changed
This release adds two observability improvements:
Conversation.Serialize(options? = null) enables inspecting the Conversation requests before firing them.
Additionally, ChatRichResponse now contains Request property, after firing the request, the outbound JSON can be observed via response.Request?.Body
- improve Conversation / ChatRichResponse observability
- Update project versions to v3.4.20
Updated Projects
LlmTornado.csproj -> 3.4.21
LlmTornado.Demo.csproj -> 0.0.1
Release v3.4.20
What's Changed
This release improves the serialization of Anthropic tool calls, passing information about tool invocation success back to the model on subsequent requests.
- improve anthropic tool calls
- Update project versions to v3.4.19
Updated Projects
LlmTornado.csproj -> 3.4.20
Release v3.4.19
What's Changed
This release adds full support for gpt-image-1, including (masked) image edits.
Generation:
Edits:
- oai: gpt-image-1
- Update project versions to v3.4.18
Updated Projects
LlmTornado.csproj -> 3.4.19
LlmTornado.Demo.csproj -> 0.0.1
Release v3.4.18
What's Changed
This release adds support for new Perplexity extensions, including search_after_date_filter, search_before_date_filter, search_recency_filter, return_related_questions, return_images, and search_domain_filter. ReasoningBudget is now safe to pass to all non-reasoning models, where the argument is ignored (previously, this would cause the request to be rejected by certain providers).
- google: harmonize thinkingConfig
- perplexity: introduce extensions
- remove deprecated mixtral-8x7b-32768
- Update README.md
- Update project versions to v3.4.17
Updated Projects
LlmTornado.csproj -> 3.4.18
LlmTornado.Demo.csproj -> 0.0.1
Release v3.4.17
What's Changed
This release adds support for Cohere's embed-v4.0. A new handler for rich streaming OnFinished is added, providing convenient access to usage & finish reason. finish_reason is now strongly typed.
- cohere: embed-v4.0
- add "OnFinished" handler to streaming responses
- make finish_reason strongly typed
- Update project versions to v3.4.16
Updated Projects
LlmTornado.csproj -> 3.4.17
LlmTornado.Demo.csproj -> 0.0.1
Release v3.4.16
What's Changed
This release adds support for o4-mini, o3, gemini-2.5-flash-preview-04-17, and new features in the APIs: adaptive thinking (ReasoningBudget), service tiers, and predictions. Memory allocations are reduced in this release. Stream argument can now be explicitly set, but is still managed by the library. Chat results now expose RawResponse for observability. ChatRichResponse got new getters for quick access to Usage and RawResponse.
- google: gemini-2.5-flash-preview-04-17, adaptive thinking
- oai: o4-mini, o3, service_tier, prediction
- Update README.md
- Update project versions to v3.4.15
Updated Projects
LlmTornado.csproj -> 3.4.16
LlmTornado.Demo.csproj -> 0.0.1