Problem

TranscriptionSpanData and SpeechSpanData both set and read instance attributes that are absent from their __slots__ declarations:

• TranscriptionSpanData: sets self.input_format (line 338), reads it in export() (line 352) — not in __slots__ (line 322–327)
• SpeechSpanData: sets self.output_format (line 379), reads it in export() (line 394) — not in __slots__ (line 366)

Because SpanData (the base class) doesn't define __slots__, instances have a __dict__ fallback, so the code works today — but:

1. The __slots__ declarations are incomplete and misleading
2. If SpanData ever gains __slots__ = (), both attributes become un-settable and __init__ raises AttributeError immediately
3. Any reflection-based code (e.g. vars(span), serializers walking __slots__) sees an inconsistent picture

Fix

Add the two missing entries — one line each:

# TranscriptionSpanData
__slots__ = ("input", "input_format", "output", "model", "model_config")

# SpeechSpanData  
__slots__ = ("input", "output", "output_format", "model", "model_config", "first_content_at")