v2.9.2

Major Changes

• None

New Features

• Adds reasoning_effort parameter to chat.completions

Bug Fixes and Patches

• None

v.2.9.1

Major Changes

• None

New Features

• Adds support for setting request timeouts, either through a TIMEOUT env variable or setting timeout when initializing the PredictionGuard class.

Bug Fixes and Patches

• None

v2.9.0

Major Changes

• Cleans up and updates examples in code and notebooks

New Features

• Adds support for new /detokenize endpoint
• Adds support for timestamps and speaker diarization to audio transcriptions (not currently supported in multi-tenant environments).
• Adds notebooks for new multiple new endpoints that were already supported.

Bug Fixes and Patches

• Fixes incorrect max_token parameter deprecation in /completions functionality.

v2.8.3

Major Changes

• Adds deprecation message for translate function.

New Features

• Adds support for enable_ocr parameter in document.extract function.

Bug Fixes and Patches

• None

v2.8.2

Major Changes

• None

New Features

• Adding support for new /documents parameters.

Bug Fixes and Patches

• None

v2.8.1

Major Changes

• None

New Features

• Adds support for new audio parameters, language, prompt, and temperature.

Bug Fixes and Patches

• Fixes error in vision calls related to tool calling error even when not using tool calls.

v2.8.0

Major Changes

• None

New Features

• Adds support for streaming in /completions endpoint.
• Adds support for new /documents endpoint, which currently allows for the extraction of data contained in documents.
• Adds support for new /audio endpoint, which currently allows for the transcription of audio files.
• Adds support for tool calling in /chat/completions using the new tools, tool_choice, and parallel_tool_calls parameters.
• Adds support for new parameters added in new model server, including stop tokens, frequency_penalty, presence_penalty, and logit_bias.

Bug Fixes and Patches

• None

v2.7.1

Major Changes

• None

New Features

• Adds support for max_completion_tokens parameter in chat.completions function

Bug Fixes and Patches

• None

v2.7.0

Major Changes

• None

New Features

• Adds support for new rerank function, allowing for the use of reranker models from the Prediction Guard API.

Bug Fixes and Patchs

• None

v2.6.0

Major Changes

• None

New Features

• Adding models function to allow for fetching all models available in Prediction Guard platform.
• Adding capability parameter to individual endpoint model listing functions to allow for sorting by capabilities.
• Adding list_models function to tokenize endpoint

Bug Fixes and Patched

• None