fix: guard against non-numeric speaker label crashing TranscriptSegment

[krushnarout]

Summary

• Line of Code:

omi/app/lib/backend/schema/transcript_segment.dart

Line 58 in 96da1b0

speakerId = speaker != null ? int.parse(speaker!.split('_')[1]) : 0;

• Cause: speaker from the API can arrive as "SPEAKER_None" (Python's None serialized into a string). int.parse("None") throws a FormatException, crashing the app on conversation load.
• Fix: Replaced int.parse with int.tryParse(...) ?? 0 and guarded the split index — same pattern already used for start/end in the same file.

🤖 Generated with Claude Code

[greptile-apps]

Greptile Summary

This PR fixes a FormatException crash in TranscriptSegment's constructor that occurred when the API returned "SPEAKER_None" as a speaker label (Python's None embedded in the speaker string format). The fix replaces int.parse(speaker!.split('_')[1]) with a null-safe split guarded by a length check and int.tryParse(...) ?? 0, matching the defensive approach already used for start/end parsing in fromJson.

• Bug fixed: "SPEAKER_None" from the API no longer crashes the app with a FormatException on conversation load.
• Approach: Two-step guard — first checks that split('_') produced at least two parts, then uses int.tryParse to gracefully handle non-numeric suffixes, defaulting to 0.
• Consistency: The pattern matches how start and end are already parsed in fromJson (lines 81–82), keeping the defensive style uniform across the file.
• Note: Segments with an unresolvable speaker label (e.g. "SPEAKER_None") are silently assigned speakerId = 0, which is the same ID as a legitimate SPEAKER_00. In conversations that also contain a real SPEAKER_00, those segments will be visually grouped together in the UI — an acceptable trade-off over crashing.

Confidence Score: 5/5

• This PR is safe to merge — it is a minimal, targeted crash fix with no regressions introduced.
• The change is a single-line logic fix in a well-understood code path. All edge cases (null speaker, no underscore, numeric suffix, non-numeric suffix like "None") are handled correctly. The pattern is already established in the same file for other fields, and no new logic or dependencies are introduced.
• No files require special attention.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A["TranscriptSegment constructor<br/>called with speaker value"] --> B{"speaker == null?"}
    B -- "Yes" --> E["parts = []"]
    B -- "No" --> C["parts = speaker.split('_')"]
    C --> D{"parts.length > 1?"}
    D -- "No<br/>e.g. 'SPEAKER' or empty" --> F["speakerId = 0"]
    D -- "Yes<br/>e.g. 'SPEAKER_00' or 'SPEAKER_None'" --> G["int.tryParse(parts[1])"]
    G --> H{"Parse succeeded?"}
    H -- "Yes<br/>e.g. parts[1] = '00'" --> I["speakerId = parsed int"]
    H -- "No<br/>e.g. parts[1] = 'None'" --> J["speakerId = 0 (fallback)"]
    E --> F

Loading

_{Last reviewed commit: 0af437f}