docs: tighten privacy claims to reflect first-run model downloads

[lewiswigmore]

Small honesty pass on docs claims after a self-review of what the code actually does.

Why

A reader pointed out (correctly) that some privacy claims are slightly stronger than what the code does:

• faster-whisper, MLX Whisper and silero-vad all fetch model weights from Hugging Face / a remote URL on first run (dictate/asr.py:137, dictate/asr_mlx.py, dictate/vad.py:_download_model). So 'no required network calls' and 'nothing leaves your machine' without qualification aren't quite right.
• Apple's exact telemetry for built-in Dictation isn't publicly documented, so 'Apple analytics (opt-out)' in the compare table was a bit hand-wavy.
• FAQ said transcripts live in <repo>/history.jsonl, but the config key is history.path and the default is ~/dictate/history.jsonl — fine when the repo is cloned to ~/dictate, misleading otherwise.

The Apple SFSpeech backend claim ('no Apple cloud') is correct as-is — asr_apple.py forces setRequiresOnDeviceRecognition_(True) and fails closed if the device can't honour it.

Changes

• docs/index.md — 'no required network calls' → 'no network calls beyond first-run model downloads'; same qualifier added to the default-pipeline bullet.
• docs/recipes.md — qualify 'nothing leaves your machine'.
• docs/compare.md — soften 'Apple analytics (opt-out)' → 'Per Apple's privacy policy'.
• FAQ.md + docs/faq.md — clarify default history path and that it's configurable via history.path.

Pure copy edits, no code changes.

[sebastionai]

Pre-merge checks · ✅ 2 · ⚠ 0 · ❌ 0 · ⏭ 1

Check	Status	Reason
PR title	✅	The title clearly describes the change as tightening privacy claims to reflect first-run model downloads.
Description	✅	The body thoroughly explains what changed, why and how with specific file-level details.
Linked issue	⏭	No linked issues were provided.

[sebastionai]

Walkthrough

Documentation-only change that qualifies privacy claims to acknowledge first-run model downloads from remote sources. Also corrects the default history file path in the FAQ and softens an unverifiable claim about Apple telemetry in the comparison table.

Changes

File	Summary
Privacy claims docs/index.md, docs/recipes.md	Qualifies 'no network calls' and 'nothing leaves your machine' to note first-run model downloads.
Comparison table docs/compare.md	Replaces 'Apple analytics (opt-out)' with 'Per Apple's privacy policy' to avoid undocumented claims.
FAQ corrections FAQ.md, docs/faq.md	Fixes the default history path and notes it is configurable via history.path.

🎯 Effort: 1 (Trivial) · ⏱ ~3 minutes

_{Generated by Sebastion AI · docs}

[Copilot]

Pull request overview

Documentation-only edits to soften privacy claims so they match actual behavior (first-run model downloads from Hugging Face), correct the default history path, and reword the Apple telemetry cell to avoid an unsubstantiated specific claim.

Changes:

• Qualify "no network calls" / "nothing leaves your machine" with a first-run model download caveat in docs/index.md and docs/recipes.md.
• Update FAQ (FAQ.md, docs/faq.md) to show the correct default history path (~/dictate/history.jsonl) and the history.path config key.
• Soften the Apple telemetry cell in docs/compare.md from "Apple analytics (opt-out)" to "Per Apple's privacy policy".

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
FAQ.md	Corrects default transcripts path and mentions history.path config key.
docs/faq.md	Mirrors the FAQ.md correction.
docs/index.md	Adds first-run model download qualifier to the two no-network-calls claims.
docs/recipes.md	Adds the same qualifier to the "nothing leaves your machine" line.
docs/compare.md	Replaces specific Apple telemetry claim with a neutral reference to Apple's privacy policy.

---

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

[sebastiondev]

LGTM — docs honesty pass.

[sebastiondev]

Still LGTM after empty re-trigger commit.