Add split metadata attribute to audio_files

[gianpaj]

Track whether an audio file was generated via the split-text flag by
accepting a split param in the generate-voice API and persisting it in
the audio_files usage metadata. The frontend now sends split: true for
segment generations and retries.

https://claude.ai/code/session_01RBKEv4AdxCJPX3rCP9B7b6

[bolt-new-by-stackblitz]

Run & review this pull request in StackBlitz Codeflow.

[pullfrog]

lifecycle hook 'setup' failed with exit code 127. output: bash: line 1: pnpm: command not found. retry the operation if the failure looks flaky (network blips, transient rate limits). do NOT retry if the script is broken (missing commands, syntax errors) or the error is persistent.

  ^{｜ Rerun failed job ➔ ｜ View workflow run ｜ via Pullfrog ｜ 𝕏}

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
sexyvoice	Ready	Preview, Comment	May 14, 2026 9:46pm
sexyvoice-docs	Ready	Preview, Comment	May 14, 2026 9:46pm

[gemini-code-assist]

Code Review

This pull request introduces a split flag to the voice generation workflow, enabling the API to track segmented text processing. The API route now parses the split parameter and includes it in usage metadata, while the AudioGenerator component has been updated to pass this flag. Feedback suggests extending this tracking to usage_events and PostHog analytics for better feature visibility and consistency.

[gemini-code-assist]

The split attribute is correctly added to the audio_files usage metadata here. However, for complete feature tracking and consistent analytics, this flag should also be included in the usage_events metadata (line 633) and the PostHog event tracking (lines 210 and 647). Ensuring consistency across all tracking mechanisms is recommended for better visibility into the adoption of the split-text feature.

[claude]

Claude finished @gianpaj's task in 2m 55s —— View job

---

Claude Code Review

• Read changed files
• Check existing usage_events / PostHog tracking
• Analyze code quality, bugs, security, performance
• Post review feedback

---

Summary

The PR cleanly threads a split boolean from the frontend (segment generations + retries) through the /api/generate-voice route into the audio_files.usage JSON. Implementation is small and focused, the server-side body.split === true check is appropriately defensive, and the frontend default param (split = false) keeps existing call sites unchanged.

Two things to address before merging — one definitely-breaking, one analytics-consistency:

🛑 Tests will fail on this branch

apps/web/tests/generate-voice.test.ts asserts the exact usage shape passed to saveAudioFile via toHaveBeenCalledWith (deep equality, no extra keys allowed):

Line	Path	Current expectation
454	Replicate (xtts-v2)	usage: { userHasPaid: false }
572	Grok (xAI)	usage: { userHasPaid: false }
754	Gemini paid (pro)	usage: { promptTokenCount, candidatesTokenCount, totalTokenCount, userHasPaid: true }
851	Gemini free (flash)	same shape, userHasPaid: false
945	Gemini fallback	same shape, userHasPaid: true

After this PR, every call site adds split: false (or true) to that object → all five assertions will mismatch. Please update them and add at least one new case that posts split: true and asserts the flag is persisted.

⚠️ Analytics surfaces are inconsistent (also raised by gemini-code-assist)

split is recorded only in audio_files.usage. The other two tracking surfaces in the same handler are missing it:

• insertUsageEvent (route.ts:633) — canonical credit/usage analytics. Without split in metadata, downstream usage_events queries can't slice by mode. This matters more than audio_files, because:
• Cache-hit path (route.ts:204–220) skips saveAudioFile entirely. So if the same text was first generated single-mode and then re-requested via split, the cached audio_files row keeps its original split: false. usage_events is always written, so putting split there closes the gap.
• sendPosthogEvent (route.ts:210 and 647, defined at 761) — needs a split?: boolean parameter threaded into properties so product analytics can measure adoption.

Minor / non-blocking

• audio-generator.tsx:260 — ...(split ? { split: true } : {}) is functionally fine; you could just send split unconditionally since the server validates === true. Stylistic only.
• Naming: server uses isSplit, request body / DB uses split — slight asymmetry but harmless.
• No security or performance concerns. Payload addition is one boolean. body.split === true rejects non-boolean truthy values, which is good.

Security & performance

• ✅ No injection vector — split is coerced to a strict boolean before use.
• ✅ No PII or auth concerns.
• ✅ No measurable perf impact (one extra JSON key).

Inline comments left on route.ts:601-604 (analytics + tests) and audio-generator.tsx:260 (style nit).
• Branch: claude/add-split-metadata-attribute-eBhV9

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 56f176f4a3

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[Copilot]

Pull request overview

Adds split-generation metadata tracking for audio files so split-text segment generations can be identified after persistence.

Changes:

• Adds a split flag to frontend generate-voice requests for split segment generation and retries.
• Parses the split request field in the generate-voice API.
• Persists split into audio_files.usage metadata.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
apps/web/components/audio-generator.tsx	Sends split: true for split segment generation and retry requests.
apps/web/app/api/generate-voice/route.ts	Reads the split request field and stores it in audio file usage metadata.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[argos-ci]

The latest updates on your projects. Learn more about Argos notifications ↗︎

Build	Status	Details	Updated (UTC)
default (Inspect)	⚠️ Changes detected (Review)	5 changed	May 14, 2026, 9:48 PM