SpeakTrue

SpeakTrue Docs

Current Status

• Current durable project guidance lives in tracked docs, README.md, and AGENTS.md.
• Legacy web deploy/app startup uses wsgi:app. gb_tts_app.py is hard-deprecated; src/legacy_runtime.py remains the transitional compatibility surface.
• After legacy adapter cleanup, no production route imports legacy_runtime_adapter; remaining legacy-web work is compatibility namespace retirement for src/legacy_runtime.py, src/services/legacy_runtime_adapter.py, and monkeypatch-heavy tests.
• Turn-Based STS (live_interpreter) is the primary account-synced STS workflow across iOS, Android, and web. Batch and realtime compatibility modes remain available.
• The current web speech studios include compact TTS/STT/STS workspaces, guided voice cloning, configurable STS turn completion, STT append/replace result modes, and local STT draft restoration.
• The iOS Voice Cloning surface includes an experimental physical-device Local Qwen Lab for downloadable MLX Qwen3-TTS models, persistent on-device voice-reference profiles, and long-form local generation.
• The iOS and Android Local TTS labs have a compact voice → text → generate flow with a single readiness action, direct model/voice/transcript recovery, and explicit local export or Soundboard upload. Device listening and thermal acceptance remain separate.
• Strict-mode web drafts are local-only until server-side Supabase draft persistence is explicitly migrated.
• Eleven v4 compatibility defines model-specific controls, text limits, and the live dialogue token/frame contract; application delivery and device acceptance remain separate.

Planning Status

Active execution and release work

• Web and iOS Security Remediation — active remediation of the August 31 audit; recording provenance and rollout evidence tracked separately.
• SpeakTrue LA/Canada Load Balancer Plan — production LA-first failover to the Canada warm standby is active and failure-drill verified; dual-site deployment automation remains a follow-up.
• Android Play Release Checklist and evidence template — source readiness exists; signed upload, Play Console, screenshots, and live-smoke evidence remain pending.
• Legacy Runtime Inventory — production routes are migrated; compatibility namespace and test-surface retirement remains.

• Native Speech Parity Anchor — local contracts are implemented; deployed generate-to-Soundboard smoke evidence remains.

• Published Obsidian project notes — dated project overview, planning, work log, and codebase-map snapshot.

Future or decision-gated work

• Local Qwen Generation Settings Plan — proposed iOS controls, per-voice persistence, paragraph handling, speed processing and device/listening gates.

• Next Product Direction: 90-Day Roadmap — proposed sequence for Conversation Workspace, Unified Speech Library, and Quick Speak.
• Conversation Workspace Plan — proposed durable multi-turn STS continuity across web, iOS, and Android.
• Unified Speech Library Plan — proposed global search, recent, favorites, tags, and source filters over Soundboard.
• Quick Speak Accessibility Mode Plan — proposed native-first fast speech with local device-voice fallback and privacy-safe shortcuts.
• Supabase to GCS Backup Future Direction — design-only future direction; no worker, scheduler, or restore drill exists yet.
• iOS Subscription Reintroduction Plan — decision-gated; the current product remains free/open-access.
• Server-side persistence for strict-mode web drafts — intentionally deferred; local drafts remain authoritative.

Implemented or historical references

• Local TTS UI Simplification Plan — implemented in the iOS and Android working tree; verification and device acceptance are recorded separately.
• Generate Speech to Soundboard Workflow Target — implemented contract and regression reference.
• Realtime STS Workflow Plan — implemented realtime checkpoint, superseded as the default by Turn-Based STS.
• Monorepo Migration Inventory — completed structural migration record.
• Android Visual Refresh References — completed historical design evidence, not an active queue.

Product

• iOS Local Qwen Prototype Guide
• Next Product Direction: 90-Day Roadmap
• Conversation Workspace Plan
• Unified Speech Library Plan
• Quick Speak Accessibility Mode Plan
• SpeakTrue unified engineering manual
• SpeakTrue repository ultimate engineering bible
• SpeakTrue canonical onboarding atlas
• SpeakTrue total engineering atlas
• Repository Full Deep Reference
• Repository Onboarding Reference
• Repository Operator Manual (2026-06-29)
• iOS File Inventory
• Android File Inventory
• Web File Inventory
• SpeakTrue extreme technical atlas
• Backend Function Detailed Map
• Engineer Change Cookbook
• SpeakTrue repository extremely detailed engineer guide
• SpeakTrue repository extreme deep behavior map
• Speech Workflow Contract
• Native Speech Parity Anchor
• Generate Speech to Soundboard Workflow Target — implemented reference
• Realtime STS Workflow Plan — historical implemented checkpoint
• Usage Limits Matrix

Operations

• Android Play Release Checklist
• Android Play Release Evidence Template
• Android Play Store Listing Packet
• Obsidian Documentation Sync
• Legacy Runtime Inventory
• Legacy Runtime Compatibility Removal
• Legacy Runtime Seam Deletion
• Grafana/Loki Query Catalog
• SpeakTrue LA/Canada Load Balancer Plan — implemented
• Supabase to GCS Backup Future Direction — future only

Design

• Android Visual Refresh References — historical

Legal

• Privacy Policy
• Terms of Use
• Account Deletion