Docs cleanup: remove .cursor, sync call-first docs

README.md

@@ -3,68 +3,64 @@
 OpenLive is an open, model-neutral runtime for continuous voice agents. It separates deadline-sensitive interaction continuity from slower model cognition and preserves native duplex provider capabilities instead of forcing every model through a text-chat abstraction.
 
 > [!IMPORTANT]
-> **Disclaimer:** This is an independent open-source project. It is **not an official OpenAI project** and has no affiliation with OpenAI.
+> **Disclaimer:** This is an independent open-source project. It is **not an official OpenAI project** and has no affiliation with OpenAI or xAI.
 
 ## Current status
 
-**Version 26.7.16** (`v26.7.16`) targets a **GPT-Live-comparable** experience: polished live voice UI, open neural speech (Piper), client-side audio intelligence, WebRTC session path, semantic endpointing, **real tools + multi-agent sandbox**, and durable profile/memory — with original visuals and model neutrality intact.
+**Version 26.7.16** targets a **GPT-Live-comparable** experience: a call-first live surface, open neural speech (Piper), client-side audio intelligence, WebRTC session path, semantic endpointing, **real tools + multi-agent sandbox**, and durable profile/memory — with original visuals and model neutrality intact.
 
 Full parity matrix: [`docs/gpt-live-parity.md`](docs/gpt-live-parity.md) · Architecture roadmap: [`docs/architecture-roadmap.md`](docs/architecture-roadmap.md) · Open stack guide: [`docs/open-source-stack.md`](docs/open-source-stack.md) · Credits: [`THIRD_PARTY_NOTICES.md`](THIRD_PARTY_NOTICES.md) · Release notes: [`docs/release-26.7.16.md`](docs/release-26.7.16.md)
 
 ### What 26.7.16 ships
 
-**Voice surface**
+**Call-first voice surface**
 
-- Minimal black live surface + setup wizard; Live Presence theme tokens.
-- Full-screen voice presence with 11 named modes and multi-layer procedural orb.
-- Inline layout toggle, live dual transcript, conversation modes, speaking-style axes.
-- Push-to-talk, barge-in with local duck, camera/screen share affordances, visual cards.
-- **Piper-first voice roster** (Lessac, Amy, Ryan, …) plus API-compatible fallbacks.
-- **Boot splash** with live status and animated page-load entrance.
-- **Ripple click feedback** and enhanced hover/lift micro-interactions across the UI.
-- Brand chrome and package version: **26.7.16**.
+- First paint is a live call: **Bloub** face on a black stage + **Mute / End / Settings** (Morphicons).
+- Session **auto-joins**. If the browser needs a gesture for the mic, the surface says **Tap to talk**.
+- Lab chrome (camera, screen share, sandbox, multi-agent, diagnostics) lives in **Settings → Advanced**.
+- Optional setup wizard remains available from Settings; it is not forced on launch.
+- Full-screen mode, live transcript, conversation modes, and speaking-style axes stay available behind Settings / shortcuts.
+- **Demo TTS policy: neural or silence.** Auto never pads with formant/mock. Formant only when you explicitly pick it.
+- Package version: **26.7.16**.
 
 **Open AI voice**
 
 - Production path: cascade → OpenAI-compatible **Piper** TTS (via LocalAI, openedai-speech, or gateway-local Piper).
-- Gateway: `GET /v1/tts/status`, `POST /v1/tts/speak`; formant fallback for demos.
-- **Graceful TTS fallback chain**: gateway TTS (Piper/formant) → browser TTS → text-only, so the conversation never hangs.
+- Gateway: `GET /v1/tts/status`, `POST /v1/tts/speak`.
+- Auto engine = Piper or silence until ready; browser / formant only when selected in Settings.
 - Licenses and attribution in `THIRD_PARTY_NOTICES.md`.
 
 **Client audio intelligence**
 
 - RNNoise-style noise suppression worklet (10 ms frames).
 - Silero-style VAD worklet + energy blend.
 - NLMS adaptive echo cancellation + windowed-sinc resampler.
+- Local-first barge-in: `local duck → soft_duck → hard_yield → cancel`.
 
 **Agent, tools & sandbox**
 
 - Internal agent (no OpenCode): search, deep research pool, calculator, time, identity, profile.
 - Path-safe sandbox file I/O + optional Chrome/Edge headless browse / screenshot / PDF.
 - Multi-agent pool (≤50) with SSE progress, agent classes, and destructive-action confirms.
-- Durable user profile (facts editor, drag-and-drop reorder) + session memory export.
+- Durable user profile (facts editor) + session memory export.
 - See `sandbox/README.md` and `docs/architecture-roadmap.md`.
 
 **Transport, providers & tasks**
 
 - Binary WebSocket PCM + **gateway-native WebRTC** (DTLS data channels for events/PCM).
 - Provider-edge WebRTC (OpenAI Realtime SDP) when secrets are available.
-- **Coordinated WebRTC → WebSocket fallback**: guarded re-entry, retry limit, and clean audio/TTS state reset so a failed peer connection recovers without dropping the conversation.
+- **Coordinated WebRTC → WebSocket fallback** with clean audio/TTS state reset.
 - `POST /v1/webrtc/offer` answers browser offers; `POST /v1/realtime/session` for edge secrets.
 - **Moshi** native duplex: `--provider moshi --moshi-url ws://127.0.0.1:8998/api/chat`.
-- **Built-in LLM provider catalog**: NVIDIA NIM, Groq, OpenRouter, Together, DeepSeek, Fireworks, Mistral, Ollama, OpenAI, Cerebras, SambaNova, and Custom — selectable in setup/settings even before the gateway is ready.
+- Built-in LLM provider catalog (NVIDIA NIM, Groq, OpenRouter, Ollama, …) in Settings.
 - Semantic endpointing (transcript-aware early end ~200 ms).
-- Visual cards + live translation demo (mock) / language-mode instructions.
 - Task lifecycle, evidence links, resume with dedup.
-- Configurable `--task-deadline-ms`.
-- Developer API: `GET /health`, `/v1/meta`, `/v1/sessions`, `/v1/agent/*`, `/v1/sandbox/*`, `/v1/profile`, MCP tools (+ optional API key).
-- Session persistence (JSONL under `data/openlive-sessions`), streaming safety holdback, MCP HTTP client.
+- Developer API: `GET /health`, `/v1/meta`, `/v1/sessions`, `/v1/agent/*`, `/v1/sandbox/*`, `/v1/profile`, MCP tools.
 
-**Desktop & fullscreen**
+**Desktop**
 
-- **Tauri-based desktop shells** for Windows (MSI) and macOS (DMG/App) in `apps/openlive-desktop/`.
-- **Full-screen voice mode** with hidden chrome, hover-to-reveal controls, and a dedicated exit button / `F` shortcut.
-- Keyboard shortcuts for layout, fullscreen, camera, screen share, voice picker, and more.
+- **Tauri** shells for Windows (MSI) and macOS (DMG/App) in `apps/openlive-desktop/`.
+- Local listening-orb splash first; gateway spawn must not block first paint.
 
 ### Still missing (vs full GPT-Live)
 
@@ -74,7 +70,7 @@ Full parity matrix: [`docs/gpt-live-parity.md`](docs/gpt-live-parity.md) · Arch
 
 ## Requirements
 
-- Rust 1.83 or newer.
+- Rust 1.83 or newer (CI / lockfile may require newer for builds; prefer a recent stable).
 - A modern Chromium, Firefox, or Safari browser.
 - Microphone permission.
 
@@ -84,7 +80,7 @@ Full parity matrix: [`docs/gpt-live-parity.md`](docs/gpt-live-parity.md) · Arch
 cargo run -p openlive-gateway --release
 ```
 
-Open `http://127.0.0.1:8787` and select **Start** (or press `Space` in push-to-talk mode). The mock speaks with a lightweight formant voice so you can exercise barge-in, transcript, and tasks without external services.
+Open `http://127.0.0.1:8787`. You should land on the listening call surface immediately. Talk (or tap the orb / Mute if the browser gated the mic). Prefer installing Piper for real voice; without it the demo path stays silent rather than faking a formant pad.
 
 ## Run open-source neural voice (recommended)
 
@@ -94,11 +90,8 @@ Use any OpenAI-compatible stack that exposes:
 - `POST /v1/chat/completions`
 - `POST /v1/audio/speech` with `response_format: "pcm"` (24 kHz mono PCM16 preferred)
 
-Example with Piper-style voice ids:
-
 ```bash
 # API keys: set in the environment only — never commit keys into this repo.
-# omit or leave empty for local unauthenticated servers
 export OPENLIVE_MODEL_API_KEY
 
 cargo run -p openlive-gateway --release -- \
@@ -115,7 +108,6 @@ See [`docs/open-source-stack.md`](docs/open-source-stack.md) for LocalAI / opene
 ## Run a native realtime speech endpoint
 
 ```bash
-# Read the key from your shell environment (do not put keys in project files)
 export OPENLIVE_MODEL_API_KEY
 
 cargo run -p openlive-gateway --release -- \
@@ -127,20 +119,13 @@ cargo run -p openlive-gateway --release -- \
 
 ## Desktop app (Windows / macOS)
 
-OpenLive ships a Tauri v2 desktop shell in `apps/openlive-desktop`. The shell
-loads the same web UI as the browser and can spawn the gateway as a child
-process.
-
 ```bash
-# Build the gateway first, then the desktop app
 cargo build -p openlive-gateway --release
 cd apps/openlive-desktop
 cargo tauri build
 ```
 
-Supported bundles: Windows MSI, macOS DMG/App. See
-[`apps/openlive-desktop/README.md`](apps/openlive-desktop/README.md) for dev
-mode and platform prerequisites.
+See [`apps/openlive-desktop/README.md`](apps/openlive-desktop/README.md).
 
 ## Deterministic replay
 
@@ -152,37 +137,30 @@ cargo run -p openlive-runtime --bin openlive-replay -- \
 ## Persistence, safety & MCP
 
 ```bash
-# Default: write session events/tasks under data/openlive-sessions
 cargo run -p openlive-gateway --release
 
-# Disable durability or safety:
 cargo run -p openlive-gateway --release -- --no-persist --safety false
 
-# Attach a remote MCP tool host:
 cargo run -p openlive-gateway --release -- --mcp-url http://127.0.0.1:3100/mcp
 
-# Deep model + local knowledge notes for complex turns:
 cargo run -p openlive-gateway --release -- \
   --provider openai-compatible \
   --model-base-url http://127.0.0.1:8000/v1 \
   --llm-model llama3.2 \
   --deep-llm-model qwen2.5-32b \
   --knowledge-dir ./knowledge
 
-# Hybrid: fast local duplex + deep cascade for hard turns
 cargo run -p openlive-gateway --release -- \
   --provider hybrid \
   --model-base-url http://127.0.0.1:8000/v1
 
-# Local Chronos full-duplex latency gate
 cargo run -p openlive-runtime --release --bin openlive-full-duplex-bench -- --turns 50
 ```
 
 ## Tests
 
 ```bash
 cargo test --workspace --release
-# Integration tests need a debug binary:
 cargo build -p openlive-gateway && cargo test -p openlive-gateway --test task_lifecycle
 node --test apps/openlive-gateway/web/tests/*.test.js
 ```

THIRD_PARTY_NOTICES.md

@@ -84,9 +84,8 @@ Morphicons are **not** used as the speaking face.
 
 ## Fonts (web UI)
 
-| Family | Source | License |
-|--------|--------|---------|
-| **DM Sans**, **Manrope**, **Space Mono**, **Inter** (if loaded) | [Google Fonts](https://fonts.google.com/) | OFL / respective font licenses |
+The default call surface uses **system UI fonts** (`system-ui`, `-apple-system`,
+`Segoe UI`, etc.). No Google Fonts CDN is loaded on the demo path.
 
 ---
 

apps/openlive-desktop/README.md

@@ -49,12 +49,11 @@ cargo tauri build
 
 ## Notes
 
-- The desktop shell loads a local listening-orb splash immediately, then
-  navigates to the gateway UI once `http://127.0.0.1:12345/health` is up.
-  Spawning the gateway must not block first paint.
+- First paint is a local **listening-orb** splash (`splash/`), then the shell
+  navigates to the gateway call surface once `http://127.0.0.1:12345/health`
+  is up. Spawning the gateway must not block that first paint.
 - `OPENLIVE_SKIP_GATEWAY_BUILD=1` skips copying the real gateway binary
   (used by clippy). A placeholder file is written so Tauri's resource
   check still passes.
-- Replace `icons/icon.ico` and `icons/icon.icns` with branded assets before
-  publishing. Regenerate placeholders with
-  `python3 scripts/generate-icons.py`.
+- Placeholder icons live under `icons/`. Regenerate with
+  `python3 scripts/generate-icons.py` before publishing branded assets.

docs/architecture-roadmap.md

@@ -7,7 +7,7 @@ This document captures the target architecture beyond the current voice+tools MV
 
 - Deterministic tools: `web_search`, `deep_search`, `research_pool`, `calculator`, `get_time`, `identity`, sandbox `list/read/write_file`
 - Typo correction for ASR/search
-- Open-source **Piper TTS** status + install command + speak endpoint (formant fallback)
+- Open-source **Piper TTS** status + install command + speak endpoint (demo path: neural or silence)
 - Session **memory** JSON store + export API + inject into LLM context
 - Thought depth setting (voice / balanced / deep) — drives reply length + deep research pool
 - Agent routing that does **not** force-search every sentence
@@ -18,7 +18,8 @@ This document captures the target architecture beyond the current voice+tools MV
 - **Coordinated WebRTC → WebSocket fallback** with re-entry guard and retry cap
 - **Desktop shell** (Tauri v2) for Windows MSI and macOS DMG/App
 - **Full-screen voice mode** with hover-to-reveal chrome and `F` shortcut
-- **Boot splash + ripple feedback + enhanced motion** across the voice surface
+- **Call-first Bloub surface** + Morphicons controls + Advanced lab disclosure
+- Ripple feedback and sheet motion without a forced branded splash
 
 ## Next
 

docs/architecture.md

@@ -3,7 +3,8 @@
 Living companion: [`architecture-roadmap.md`](architecture-roadmap.md) · Release: [`release-26.7.16.md`](release-26.7.16.md)
 
 v26.7.16 adds a Tauri v2 desktop shell (`apps/openlive-desktop`), a built-in
-LLM provider catalog, full-screen voice mode, a boot splash, and a coordinated
+LLM provider catalog, full-screen voice mode, a **call-first** Bloub surface
+(Mute / End / Settings), neural-or-silence demo TTS, and a coordinated
 WebRTC → WebSocket fallback path. The sections below describe the runtime
 architecture that underlies those features.
 

docs/evaluation.md

@@ -1,6 +1,8 @@
 # Evaluation and latency telemetry (v26.7.16)
 
-Openlive does not claim experiential parity without measurements. Protocol 1.0 emits generation-scoped `latency_mark` events from the gateway's monotonic clock:
+OpenLive does not claim experiential parity with GPT-Live without measurements.
+Call-first UX (listening orb, barge-in, neural-or-silence TTS) still needs the
+instrumented latency marks below for honest comparison. Protocol 1.0 emits generation-scoped `latency_mark` events from the gateway's monotonic clock:
 
 - `response_committed`;
 - `first_provider_event`;

docs/gpt-live-parity.md

@@ -57,8 +57,8 @@ Legend: ✅ CLONE · 🟡 DIFFERENT · ❌ GAP
 
 | # | Feature | GPT-Live / AVM behavior | Openlive 26.7.16 status | Category |
 |---|---------|-------------------------|------------------------|----------|
-| 1 | Signature voice orb | Blue animated orb, state-driven | Bloub face (x.ai-inspired, MIT, not affiliated) on a black call stage | ✅ CLONE (original visual) |
-| 2 | State-driven orb color | Blue / cyan / violet / red shifts | 11 named modes, each with its own palette | ✅ CLONE |
+| 1 | Signature voice orb | Blue animated orb, state-driven | Bloub face (x.ai-inspired, MIT, not affiliated) on a black call stage; auto-listening first paint | ✅ CLONE (original visual) |
+| 2 | State-driven orb color | Blue / cyan / violet / red shifts | Bloub listen/speak/barge-in states; optional presence palettes behind Settings | ✅ CLONE |
 | 3 | Live dual transcript | Inline user + assistant bubbles | Persistent scrolling transcript with role-differentiated bubbles + system channel | ✅ CLONE |
 | 4 | Inline-in-chat voice mode | Late-2025 redesign moved voice inline with chat | Layout toggle: focused vs inline | ✅ CLONE |
 | 5 | Voice picker | Named voices + personality descriptors | Piper-first roster + provider manifest + offline fallback | ✅ CLONE |
@@ -71,7 +71,7 @@ Legend: ✅ CLONE · 🟡 DIFFERENT · ❌ GAP
 | 12 | Backchanneling ("mhmm") | Native GPT-Live behavior | UI affordance + event handler; badge near orb | ✅ CLONE (UI; provider must emit) |
 | 13 | Camera input | Camera button streams frames | UI + `C` shortcut; truthful media lifecycle; visual-input negotiation | ✅ CLONE |
 | 14 | Screen sharing | Available alongside camera | UI + `Shift+C`; truthful media lifecycle | ✅ CLONE |
-| 15 | Mute / end-call / camera controls | Persistent in-call controls | Mute, End, Camera, Screen, Voice, Mode, Instructions — richer than AVM | ✅ CLONE (extended) |
+| 15 | Mute / end-call / camera controls | Persistent in-call controls | Default dock: Mute / End / Settings (Morphicons). Camera / screen / modes in Settings → Advanced | ✅ CLONE |
 | 16 | Daily/session quota indicator | Plus: 1 hr/day; free: 15 min preview | Operator-configured cap; soft warning at 80% | ✅ CLONE |
 | 17 | Latency display | Not surfaced in AVM UI | Latency pill + diagnostics p50/p95/jitter/loss | 🟡 DIFFERENT (Openlive exposes it) |
 | 18 | Live translation | Built-in GPT-Live feature | VisualCard + language chip instructions; cascade hop for production | 🟡 DIFFERENT (partial) |
@@ -94,7 +94,7 @@ Legend: ✅ CLONE · 🟡 DIFFERENT · ❌ GAP
 | 35 | Diagnostics on-demand | Hidden in AVM | Diagnostics drawer + LiveBench | 🟡 DIFFERENT |
 | 36 | Model neutrality | AVM locked to OpenAI models | Mock / cascade / realtime / moshi / hybrid | 🟡 DIFFERENT (Openlive beats AVM) |
 | 37 | Theme customization | AVM orb is fixed | Live Presence / Graphite / Signal + motion intensity | 🟡 DIFFERENT |
-| 38 | Onboarding overlay | (Not in AVM) | First-run onboarding + setup wizard | 🟡 DIFFERENT (Openlive original) |
+| 38 | Onboarding overlay | (Not in AVM) | Optional setup wizard / shortcuts sheet from Settings — not forced on launch | 🟡 DIFFERENT (Openlive original) |
 | 39 | Settings persistence | Account-scoped prefs | `localStorage` + durable server profile | ✅ CLONE (extended) |
 | 40 | Keyboard shortcuts | (Not in AVM) | 10+ shortcuts | 🟡 DIFFERENT (Openlive beats AVM) |
 | 41 | Task acknowledgement lifecycle | (Not in AVM) | Full lifecycle; p50 = 2 ms | 🟡 DIFFERENT (Openlive original) |
@@ -131,9 +131,9 @@ enforce p50 ≤ 50 ms and p95 ≤ 200 ms.
 ## What "open-source clone" means here
 
 Openlive 26.7.16 is a **behavioral clone** of gpt-live's voice surface,
-not a **visual clone**. The orb, palettes, copy, layout, and animation
-are original Openlive geometry — they do not reproduce any proprietary
-interface or its assets.
+not a **visual clone**. The default face uses the MIT **Bloub** engine
+(x.ai-inspired; not affiliated with xAI or OpenAI). Layout, chrome, and
+copy are OpenLive originals — they do not reproduce proprietary assets.
 
 The clone contract is: a user who is familiar with AVM should be able
 to use Openlive without relearning anything, and an operator comparing