docs(stt): remove emotion + gender detection from non-model-card surfaces#252
Open
abhishekmishragithub wants to merge 3 commits into
Open
docs(stt): remove emotion + gender detection from non-model-card surfaces#252abhishekmishragithub wants to merge 3 commits into
abhishekmishragithub wants to merge 3 commits into
Conversation
… atoms broken link Three independent fixes bundled (post-#249 audit pass): **Pulse STT language list — drift cleanup.** Streaming WebSocket spec enum had picked up 24 extra codes that don't work on streaming (pre-recorded-only codes like `uk`/`pl`/`cs`/..., unsupported extras like `pa`/`ta`/`te`/..., and `multi-eu` which is non-streaming). Realigned the streaming enum to the canonical 19 entries: 17 single-language codes + `north_indic` + `multi-asian`. Also fixed `multi-indic` references across docs — that aggregator doesn't exist on either mode. Updated: - `speech-to-text/quickstart.mdx` — replaced single combined list with mode-specific aggregator note (pre-recorded: multi-eu / multi-asian; streaming: north_indic / multi-asian US-region-only). - `speech-to-text/pre-recorded/quickstart.mdx` (×2) — same. - `speech-to-text/features/language-detection.mdx` — converted the bullet list into a per-mode aggregator availability table, removed `multi-indic` from copy, switched the WS code sample from `language=multi-eu` to `language=north_indic`. - `speech-to-text/realtime/response-format.mdx` — `multi` / `multi-eu` replaced with `north_indic` / `multi-asian` for the streaming context. - `api-references/pulse-stt-ws.mdx` — language param table + Python example + footer note all updated. - `integrations/livekit.mdx` — example `STT(language="multi-indic")` switched to `north_indic`; param table description fixed. - `integrations/vercel-ai-sdk.mdx` — enum comment updated. Pulse model card itself already correct; pre-recorded HTTP enum already correct — both untouched. **Lightning v3.1 Pro language default — visibility.** OpenAPI `language` param now leads with "Default on `lightning_v3.1_pro`: when omitted, defaults to `en + hi`". Same param's `requestBody` example was a single object; converted to four named examples covering Pro default (omit language → en + hi), Pro explicit `en`, Pro explicit `hi`, and standard `lightning_v3.1` with explicit language — Fern's docs UI will render them as selectable example variants so the Pro-default behavior is visible from the request playground. **Atoms broken link.** `fern/products/atoms/pages/dev/integrate/websocket-sdk.mdx` linked to `/atoms/api-reference/api-reference/logs/get-conversation-log-by-id`, which 404s — the endpoint moved to `/atoms/api-reference/api-reference/calls/get` (page title still "Get conversation log by ID"). Updated the link. Verified via `fern check` (0 errors) and `fern docs dev` render — target pages return 200, no MDX errors.
…aces Per Harshita's review (model-card audit), emotion and gender detection should be de-emphasized as user-facing features. Hashita is editing the model cards directly. This PR covers the parallel cleanup everywhere else those features were surfaced, **without touching any file under `/model-cards/`**. What's gone: - Standalone feature pages `speech-to-text/features/age-and-gender-detection.mdx` and `speech-to-text/features/emotion-detection.mdx` deleted (the age-detection portion was already removed from the platform per the 2026-05-06 changelog; only gender remained and is now also gone from docs). - Removed both pages from the nav YAML (`fern/products/waves/versions/v4.0.0.yml`). - Removed both pages from `fern/llms.txt`. - Removed the legacy `age-and-gender-detection` → `gender-detection` redirect from `fern/docs.yml` (target page no longer exists). What changed on rendered docs (non-model-card pages only): - STT `overview.mdx`: dropped Gender + Emotion accordion entries; page description now ends with "PII / PCI redaction" instead of "emotion detection". - STT `quickstart.mdx`: "Explore Features" card row — Emotion Detection card replaced with Language Detection (more useful for the quickstart audience). - Pre-recorded `code-examples.mdx`: advanced Pulse example renamed from "gender, emotion, diarization, utterances" to "diarization, utterances, redaction". Python code drops `gender_detection` and `emotion_detection` query params; adds `redact=pii,pci`. Comparison table and "Expected output" updated to match. - Pre-recorded `features.mdx`: Gender + Emotion cards removed from the card group. - Integrations `n8n.mdx`: Two option tables updated — drops `gender_detection` and `emotion_detection`; replaces with `diarize` and `word_timestamps` (parameters n8n nodes actually surface today). - Integrations `openclaw.mdx`: feature bullet changed from "Speaker diarization and emotion detection (STT)" to "Speaker diarization and PII / PCI redaction (STT)". - Atoms `mcp/tools.mdx`: `transcribe_audio` description drops "emotion detection"; adds "PII / PCI redaction". - Atoms `mcp/prompt-cookbook.mdx`: removed the emotion-detection example prompt. What was deliberately NOT touched: - All files under `/model-cards/` — Hashita is editing these. - The OpenAPI spec (`fern/apis/waves/openapi/pulse-stt-openapi.yaml`, `stt-openapi.yaml`) — `gender_detection` and `emotion_detection` query params still exist server-side. Removing them from the spec is a platform-team decision (deprecation pathway), not a docs-only one. Leaving the spec untouched means existing clients keep working; we're just no longer marketing these features in docs. - The orphan `api-references/pulse-stt.mdx` page (already not in any nav YAML; not rendered). - Frozen `versions/v4.0.0/` mirrors — per project policy. - Changelog entries — historical record, never edit. Verified via `fern check` (0 errors) and `fern docs dev` — `overview`, `pre-recorded/features` render 200; the two deleted feature pages return 404 as expected. Hashita's model-card edits land independently; after they merge we can pair on a follow-up to align cross-card structure (FAQ / Pricing / Best Practices / How-to-use placement, etc.).
Sample called the unified STT endpoint with 'redact: pii,pci' — but 'redact' is not a recognised query param on the Pulse STT API. The canonical params are 'redact_pii' and 'redact_pci' (separate booleans, each enum 'true'/'false'). Customers copy-pasting this example would silently get back unredacted transcripts. Verified against fern/apis/waves/openapi/stt-openapi.yaml:205,218. Fixed by splitting into the two real params with the actual string-enum values the spec accepts. Caught while validating the diarize-can-combine claim on PR #254 (diarization.mdx) — same drift, both fixed now.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What's gone
speech-to-text/features/age-and-gender-detection.mdxandspeech-to-text/features/emotion-detection.mdx(age-detection was already removed from the platform per the 2026-05-06 changelog; only gender remained and is now also gone from docs).fern/products/waves/versions/v4.0.0.yml.age-and-gender-detection→gender-detectionremoved fromfern/docs.yml(target page no longer exists).Rendered-page changes (non-model-card only)
overview.mdx— Gender + Emotion accordion entries removed; page description updated.quickstart.mdx— Emotion Detection card swapped for Language Detection (more relevant to the quickstart audience).code-examples.mdx— Advanced Pulse Python example renamed "diarization, utterances, redaction"; code dropsgender_detection/emotion_detectionquery params, addsredact=pii,pci; comparison table + Expected output updated.features.mdx— Gender + Emotion cards removed from the card group.n8n.mdx— Two option tables updated: dropsgender_detection/emotion_detection; addsdiarize+word_timestamps.openclaw.mdx— Feature bullet now reads "Speaker diarization and PII / PCI redaction (STT)".mcp/tools.mdx—transcribe_audiodescription drops "emotion detection"; adds "PII / PCI redaction".mcp/prompt-cookbook.mdx— Removed the emotion-detection example prompt.What was deliberately NOT touched
/model-cards/*.mdxpulse-stt-openapi.yaml,stt-openapi.yaml)gender_detection/emotion_detectionparams still exist server-side; removing them is a platform-team deprecation decision, not docs-onlyapi-references/pulse-stt.mdxversions/v4.0.0/mirrorsVerification
fern check→ 0 errors, 24 pre-existing warnings (unrelated).fern docs devboots clean; no MDX errors./waves/documentation/speech-to-text-pulse/overview,/waves/documentation/speech-to-text-pulse/pre-recorded/features..../features/gender-detection,.../features/emotion-detection.Test plan
/waves/documentation/speech-to-text-pulse/features/gender-detectionreturns 404 (no broken inbound nav).🤖 Generated with Claude Code