Upgrade pyannote community dependency stack

Reapply pyannote downgrade changes

Refactor pyannote token handling

Make pyannote Pipeline import module-scoped

Revert repository state to c7d2aa0

Revert repository state to e4b57c8

Remove main-process CUDA checks from CPU path

Add CPU-only Parakeet debug route

Replicate working ZeroGPU Parakeet reference space

Add detailed Parakeet ZeroGPU stage logging

Cap ZeroGPU window and add Parakeet GPU diagnostics

Reset repo for phased rebuild and scaffold Phase 1 Parakeet API

Disable Lhotse in Parakeet transcribe flow

fix(pyannote): support older auth arg name

feat(api): add pyannote-only diarize endpoint

feat(api): add parakeet-only transcribe endpoint

Restore app to c7d2aa0 and only update dependencies

Restore Parakeet-aligned deps and add pipeline stage logging

Pin runtime dependencies for stable Space rebuilds

Remove Unused Code From OpenAI Pipeline Stage

Use Chunking For OpenAI API Calls

Fix OpenAI Script Bug

Add Combined Pipeline For Audio Transcription

Match Parakeet Setup With Official Space

Fix Broken Build

Fix Parakeet Call Error And Remove Unused API Parameters

Fix Broken Parakeet Code And Add 2 Hour Limit On Pyannote

Code Fix Pyannote Bug #2

Codex Fix Pyannote AudioDecoder Bug

Fix Pyannote Setup Bug

Add Pyannote Diarization API

Remove Initial Prompt From Whisper Turbo

Switch To Preload Models And Other Optimizations

Codex Turbo Fix

Fix Turbo Mode Prompt Input

Fix Whisper Turbo Prompt Input

Refactor To Only Use Whisper Turbo And Parakeet

Remove Combined Benchmark API And Deprecated Code

Fix Bug: Remove Audio File Limit And Update Code

Add Multi-Model API For Transcription