Compares two text strings and calculates their similarity ratio. This class provides methods to calculate the similarity between two texts

Provides a unified interface for interacting with various LLM backends. Supports Ollama (via direct HTTP), OpenAI API, and LMStudio (via Ope

Represents a request to be processed by the SpeechPipelineManager's request queue. Holds information about the action to perform (e.g., 'pre

Manages audio input, processes it for transcription, and handles related callbacks. This class receives raw audio chunks, resamples them to

Manages Text-to-Speech (TTS) synthesis using various engines via RealtimeTTS. This class initializes a chosen TTS engine (Coqui, Kokoro, or

A logging Formatter that displays timestamps as MM:SS.cs using local time. Inherits from `logging.Formatter` and overrides the `formatTime`

A context manager for safely handling LLM generation streams. Ensures that the underlying generation stream is properly requested for cancel

Serves static files without allowing client-side caching. Overrides the default Starlette StaticFiles to add 'Cache-Control' headers tha

Holds the state and resources for a single, ongoing text-to-speech generation process. This includes the generation ID, input text, the LLM

Orchestrates the text-to-speech pipeline, managing LLM and TTS workers. This class handles incoming text requests, manages the lifecycle of

Extracts meaningful text segments (contexts) from a given string. This class identifies substrings that end with predefined split tokens,

Manages state and callbacks for a single WebSocket connection's transcription lifecycle. This class holds connection-specific state flags (l

Manages audio transcription using RealtimeSTT, handling real-time and final transcription callbacks, silence detection, turn detection (optio

Manages turn detection logic based on text input and sentence completion model. This class receives text segments, uses a transformer model

Manages chunk-wise audio upsampling with overlap handling. This class processes sequential audio chunks, upsamples them from 24kHz to 48kHz