Speechmatics-flow

Added

- `tools` parameter: Introduced in the `client.run()` function to enable custom tool functionality.
- `ToolFunctionParam` class: Added for enhanced type-checking when building client functions.
- New message types: `ToolInvoke` and `ToolResult` messages are now supported for handling function calling.

Changed

- StartConversation message: Updated to include the `tools` parameter.

Added

- Added the option to change the assistant from CLI
- Added the option to load conversation_config from a config file
- Added client handling of unexpected messages from the server

Changed

- Allow versions of websockets from `10.0` up to and including `13.1` to mitigate extra_headers compatibility issue
with websockets `14.0`
- Improved documentation for Interaction class

Added

- `ResponseStarted`: Indicates the start of TTS audio streaming from the server.
The message contains the textual content of the utterance to be spoken.
- `ResponseInterrupted`: Indicates an interruption in the TTS audio stream from the server.
The message contains the textual content up to the point where the utterance was stopped.
- `ResponseCompleted`: Indicates the completion of TTS audio transmission from the server.
The message includes the textual content of the utterance just spoken.
- `ConversationEnding`: Indicates the session will continue in one-sided mode during TTS playback of the final words.
- `AddAudio`: Implicit name for all inbound binary messages.
The client confirms receipt by sending an `ServerMessageType.AudioReceived` message.
- `AudioReceived`: Response to `ServerMessageType.AddAudio`, indicating that audio has been added successfully.
- Deprecation warning for `audio` (replaced by AddAudio) and `prompt` (replaced by Response*) messages

Removed

- Unused `EndOfTranscript` server message

Changed

- PyAudio class is instantiated only when the client is started directly from the CLI.
- Simplified microphone example

Fixed

- Choppy audio playback on some systems using Python 3.12+
- Latency issues on some systems using Python 3.12+

Added

- Improved handling of the AudioEnded which caused the client to abruptly close the connection.
The client now waits up to 5 seconds for a ConversationEnded message from the server before closing the connection.

Changed

- Do not generate JWT when connecting to a local Flow server.
- `TranscriptionError` is now `ConversationError`

Fixed

- CLI usage example from README using `-` which caused an `unrecognized arguments` error.
- Stream from microphone example using ssl_context=None

Removed

- `EndOfTranscriptException` from exceptions

Added

- Add speechmatics-flow client