clawdex-mobile 1.3.2 → 2.0.0

package/docs/troubleshooting.md

@@ -0,0 +1,121 @@
+# Troubleshooting
+
+## Onboarding looks stuck before Expo logs appear
+
+- Expo startup can be slow on first launch.
+- You should see: `Waiting for Expo output ...`
+- Increase timeout if needed:
+
+```bash
+EXPO_OUTPUT_WAIT_SECS=180 clawdex init
+```
+
+- If Expo never emits logs:
+
+```bash
+tail -n 120 .expo.log
+```
+
+## Expo starts but QR/network is wrong
+
+- Re-run `npm run secure:setup`
+- Confirm `.env.secure` has correct `BRIDGE_HOST`
+- Restart `npm run mobile`
+
+## Stop all running services quickly
+
+Preferred:
+
+```bash
+clawdex stop
+```
+
+From repo checkout:
+
+```bash
+npm run stop:services
+```
+
+## Bridge auth errors (`401`, invalid token)
+
+- Ensure `BRIDGE_AUTH_TOKEN` in `.env.secure` matches `EXPO_PUBLIC_HOST_BRIDGE_TOKEN` in `apps/mobile/.env`
+- Restart bridge + Expo after token changes
+
+## Tailscale issues
+
+- Verify host and phone are on the same Tailscale network
+- Check host IP (`tailscale ip -4`) and mobile `.env` URL
+
+## `codex` not found
+
+- Ensure `codex` is in `PATH`
+- Or set `CODEX_CLI_BIN` explicitly
+
+## Bridge build fails with `linker 'cc' not found`
+
+Install C build tools:
+
+```bash
+sudo apt-get update && sudo apt-get install -y build-essential
+```
+
+Then retry `npm run secure:bridge`.
+
+## iOS bundling error: `Unable to resolve "./BoundingDimensions"`
+
+Manual recovery:
+
+```bash
+npm install --include=dev --force
+npm install --include=dev --force -w apps/mobile
+npm run -w apps/mobile start -- --clear
+```
+
+## Runtime errors: `[runtime not ready]` / `property is not writable`
+
+Manual recovery:
+
+```bash
+rm -rf node_modules apps/mobile/node_modules
+npm install --include=dev --force
+npm install --include=dev --force -w apps/mobile
+npm run -w apps/mobile start -- --clear
+```
+
+Also update Expo Go on your phone.
+
+## Git operations fail
+
+- Verify chat workspace is a valid git repo
+- Verify remote auth/access for push
+
+## Attachment upload issues
+
+- Ensure mobile app has file/photo permissions
+- File limit is `20 MB` per upload
+- Uploads persist under `BRIDGE_WORKDIR/.clawdex-mobile-attachments`
+- Ensure `BRIDGE_WORKDIR` is writable
+
+## Worklets/Reanimated mismatch
+
+```bash
+cd apps/mobile
+npx expo install --fix
+npm run start -- --clear
+```
+
+## Plan mode errors (`RPC-32600` invalid `collaborationMode`)
+
+- Restart Expo and reload app bundle
+- Ensure bridge/mobile revisions match
+- Run API test if needed:
+
+```bash
+npm run -w apps/mobile test -- --runInBand src/api/__tests__/client.test.ts
+```
+
+## Stop button does not interrupt a run
+
+- Ensure revision supports `turn/interrupt`
+- If run already finished, stop button disappears by design
+- Pull latest, restart bridge, reload Expo bundle

package/docs/voice-transcription.md

@@ -0,0 +1,87 @@
+# Push-to-Talk Voice Transcription
+
+Voice-to-text input for the mobile composer. Tap the mic button to record, tap again to stop — the audio is sent to the Rust bridge which calls OpenAI's transcription API and returns the text into the composer.
+
+## Architecture
+
+```
+Phone (expo-audio)        Rust Bridge              OpenAI
+  record 16kHz mono       --> base64 over WebSocket --> POST /v1/audio/transcriptions
+  (iOS WAV, Android M4A)
+  insert text in composer <-- { text: "..." }    <-- gpt-4o-transcribe response
+```
+
+The flow mirrors Codex CLI 0.105.0's TUI push-to-talk feature — same API, same model (`gpt-4o-transcribe`), same auth. The bridge acts as the HTTP proxy since the phone doesn't hold API keys directly.
+
+## Auth Resolution
+
+The bridge resolves transcription credentials in order:
+
+1. `OPENAI_API_KEY` env var → `https://api.openai.com/v1/audio/transcriptions` (model: `gpt-4o-transcribe`)
+2. `BRIDGE_CHATGPT_ACCESS_TOKEN` env var → `https://chatgpt.com/backend-api/transcribe` (no model param)
+3. `~/.codex/auth.json` fallback:
+   - `OPENAI_API_KEY` field present → same as path 1
+   - `auth_mode: "chatgpt"` with `tokens.access_token` → same as path 2
+
+## Files
+
+### Rust Bridge
+
+| File | What |
+|------|------|
+| `services/rust-bridge/Cargo.toml` | Added `reqwest` (multipart, json, rustls-tls) |
+| `services/rust-bridge/src/main.rs` | `bridge/voice/transcribe` JSON-RPC method, `transcribe_voice()`, `resolve_transcription_auth()`, `resolve_codex_auth_json_path()` |
+
+### Mobile
+
+| File | What |
+|------|------|
+| `apps/mobile/app.json` | `NSMicrophoneUsageDescription` (iOS), `RECORD_AUDIO` permission (Android), `expo-audio` plugin |
+| `apps/mobile/src/api/types.ts` | `VoiceTranscribeRequest`, `VoiceTranscribeResponse` |
+| `apps/mobile/src/api/client.ts` | `transcribeVoice()` method on `HostBridgeApiClient` |
+| `apps/mobile/src/hooks/useVoiceRecorder.ts` | Recording state machine hook using `expo-audio` |
+| `apps/mobile/src/components/ChatInput.tsx` | Mic button UI with three visual states |
+| `apps/mobile/src/screens/MainScreen.tsx` | Wires hook to ChatInput |
+
+## Recording Config
+
+- Sample rate: 16,000 Hz
+- Channels: 1 (mono)
+- Format:
+  - iOS: LINEARPCM 16-bit (`audio/wav`)
+  - Android: MPEG-4 AAC (`audio/mp4`, `.m4a`)
+- Minimum duration:
+  - Mobile guard: 1 second
+  - Bridge raw payload guard: ~0.5 seconds (16KB minimum)
+- Maximum payload size:
+  - Mobile guard: 20MB
+  - Bridge guard: 100MB by default (override with `BRIDGE_MAX_VOICE_TRANSCRIPTION_BYTES`)
+
+## UI States
+
+The mic button occupies the send button slot when the composer is empty and no turn is running:
+
+| State | Icon | Style |
+|-------|------|-------|
+| Idle | `mic-outline` | Muted color |
+| Recording | `mic` | Red icon, red border |
+| Transcribing | `ActivityIndicator` | Spinner |
+
+Interaction: tap to start recording, tap again to stop and transcribe. The transcribed text is appended to the current draft.
+
+Voice input is enabled on iOS and Android. The mic button is hidden on web.
+
+## Dependencies
+
+- **`expo-audio`** — Recording via `useAudioRecorder` hook. Works in Expo Go (unlike `expo-av` which requires a dev build in SDK 55).
+- **`reqwest`** (Rust) — HTTP client for the multipart POST to OpenAI. Uses `rustls-tls` to avoid native-tls linking issues on macOS.
+
+## Error Handling
+
+- Mic permission denied → error message, stays idle
+- Recording < 1 second → "Recording too short" error, discarded
+- Audio payload < 16KB → bridge rejects with `invalid_params`
+- Audio payload > 20MB → mobile rejects before upload
+- Audio payload > bridge max bytes (default 100MB) → bridge rejects with `invalid_params`
+- No credentials found → bridge returns error code `-32002`
+- Transcription API HTTP error → bridge returns status + body in error data

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "clawdex-mobile",
-  "version": "1.3.2",
+  "version": "2.0.0",
   "homepage": "https://github.com/Mohit-Patil/clawdex-mobile#readme",
   "repository": {
     "type": "git",
@@ -25,7 +25,9 @@
     "stop:services": "./scripts/stop-services.sh",
     "secure:setup": "./scripts/setup-secure-dev.sh",
     "secure:bridge": "./scripts/start-bridge-secure.sh",
+    "secure:bridge:dev": "BRIDGE_RUN_MODE=dev ./scripts/start-bridge-secure.sh",
     "bridge:ts": "BRIDGE_WORKDIR=$(pwd) npm run -w @codex/mac-bridge dev",
+    "version:sync": "node scripts/sync-versions.js",
     "build": "npm run --workspaces build",
     "typecheck": "npm run --workspaces typecheck",
     "lint": "npm run --workspaces lint",
@@ -33,20 +35,10 @@
     "teardown": "./scripts/teardown.sh"
   },
   "dependencies": {
-    "@types/react": "19.1.10",
-    "react": "19.1.0",
-    "react-native": "0.81.5",
-    "expo": "54.0.33"
-  },
-  "overrides": {
-    "minimatch": "10.2.2",
-    "@react-navigation/drawer": {
-      "react-native-reanimated": "4.1.1"
-    },
-    "react-native-drawer-layout": {
-      "react-native-reanimated": "4.1.1"
-    },
-    "react-native-reanimated": "4.1.1",
-    "react-native-worklets": "0.5.1"
+    "@types/react": "~19.2.10",
+    "react": "19.2.0",
+    "react-dom": "19.2.0",
+    "react-native": "0.83.2",
+    "expo": "^55.0.2"
   }
 }

package/scripts/setup-secure-dev.sh

@@ -77,6 +77,34 @@ ensure_tailscale_cli() {
   return 0
 }
 
+is_ipv4() {
+  local ip="$1"
+  local part=""
+  local -a octets=()
+
+  [[ "$ip" =~ ^([0-9]{1,3}\.){3}[0-9]{1,3}$ ]] || return 1
+  IFS='.' read -r -a octets <<<"$ip"
+  if [[ "${#octets[@]}" -ne 4 ]]; then
+    return 1
+  fi
+
+  for part in "${octets[@]}"; do
+    if (( part < 0 || part > 255 )); then
+      return 1
+    fi
+  done
+
+  return 0
+}
+
+is_non_loopback_ipv4() {
+  local ip="$1"
+  if ! is_ipv4 "$ip"; then
+    return 1
+  fi
+  [[ "$ip" != 127.* ]]
+}
+
 resolve_tailscale_ipv4() {
   local ip
   ip="$(tailscale ip -4 2>/dev/null | head -n1 | tr -d '[:space:]' || true)"
@@ -100,6 +128,77 @@ resolve_tailscale_ipv4() {
   printf '%s' "$ip"
 }
 
+current_local_ipv4() {
+  local candidate=""
+  local iface=""
+
+  if [[ "$(uname -s)" == "Darwin" ]] && command -v ipconfig >/dev/null 2>&1; then
+    for iface in en0 en1; do
+      candidate="$(ipconfig getifaddr "$iface" 2>/dev/null | tr -d '[:space:]' || true)"
+      if is_non_loopback_ipv4 "$candidate"; then
+        printf '%s' "$candidate"
+        return 0
+      fi
+    done
+  fi
+
+  if command -v hostname >/dev/null 2>&1; then
+    while IFS= read -r candidate; do
+      candidate="$(printf '%s' "$candidate" | tr -d '[:space:]')"
+      if is_non_loopback_ipv4 "$candidate"; then
+        printf '%s' "$candidate"
+        return 0
+      fi
+    done < <(hostname -I 2>/dev/null | tr ' ' '\n' || true)
+  fi
+
+  if command -v ip >/dev/null 2>&1; then
+    candidate="$(ip route get 1.1.1.1 2>/dev/null | awk '{ for (i=1; i<=NF; i++) if ($i=="src") { print $(i+1); exit } }')"
+    candidate="$(printf '%s' "$candidate" | tr -d '[:space:]')"
+    if is_non_loopback_ipv4 "$candidate"; then
+      printf '%s' "$candidate"
+      return 0
+    fi
+  fi
+
+  if command -v ifconfig >/dev/null 2>&1; then
+    while IFS= read -r candidate; do
+      candidate="$(printf '%s' "$candidate" | tr -d '[:space:]')"
+      if is_non_loopback_ipv4 "$candidate"; then
+        printf '%s' "$candidate"
+        return 0
+      fi
+    done < <(ifconfig 2>/dev/null | awk '/inet /{print $2}' || true)
+  fi
+
+  return 1
+}
+
+resolve_local_ipv4() {
+  local ip=""
+  ip="$(current_local_ipv4 || true)"
+
+  if [[ -n "$ip" ]]; then
+    printf '%s' "$ip"
+    return 0
+  fi
+
+  echo "No active local/LAN IPv4 found automatically."
+  if ! confirm_prompt "Enter bridge host IP manually?"; then
+    echo "error: unable to resolve LAN IPv4. Connect to LAN and retry." >&2
+    return 1
+  fi
+
+  read -r -p "Bridge host IP: " ip
+  ip="$(printf '%s' "$ip" | tr -d '[:space:]')"
+  if ! is_non_loopback_ipv4 "$ip"; then
+    echo "error: invalid IPv4 '$ip'." >&2
+    return 1
+  fi
+
+  printf '%s' "$ip"
+}
+
 if ! command -v openssl >/dev/null 2>&1; then
   echo "error: openssl not found. Install OpenSSL first." >&2
   exit 1
@@ -117,13 +216,28 @@ fi
 
 BRIDGE_HOST="${BRIDGE_HOST_OVERRIDE:-}"
 HOST_SOURCE=""
+BRIDGE_NETWORK_MODE="${BRIDGE_NETWORK_MODE:-tailscale}"
+
+case "$BRIDGE_NETWORK_MODE" in
+  tailscale|local)
+    ;;
+  *)
+    echo "error: BRIDGE_NETWORK_MODE must be 'tailscale' or 'local'." >&2
+    exit 1
+    ;;
+esac
 
 if [[ -n "$BRIDGE_HOST" ]]; then
   HOST_SOURCE="override"
 else
-  ensure_tailscale_cli
-  BRIDGE_HOST="$(resolve_tailscale_ipv4)"
-  HOST_SOURCE="tailscale"
+  if [[ "$BRIDGE_NETWORK_MODE" == "tailscale" ]]; then
+    ensure_tailscale_cli
+    BRIDGE_HOST="$(resolve_tailscale_ipv4)"
+    HOST_SOURCE="tailscale"
+  else
+    BRIDGE_HOST="$(resolve_local_ipv4)"
+    HOST_SOURCE="local"
+  fi
 fi
 
 BRIDGE_PORT="${BRIDGE_PORT_OVERRIDE:-8787}"
@@ -139,26 +253,26 @@ if [[ -z "$BRIDGE_TOKEN" ]]; then
 fi
 
 cat > "$SECURE_ENV_FILE" <<EOT
+BRIDGE_NETWORK_MODE=$BRIDGE_NETWORK_MODE
 BRIDGE_HOST=$BRIDGE_HOST
 BRIDGE_PORT=$BRIDGE_PORT
 BRIDGE_AUTH_TOKEN=$BRIDGE_TOKEN
-BRIDGE_ALLOW_QUERY_TOKEN_AUTH=false
+BRIDGE_ALLOW_QUERY_TOKEN_AUTH=true
 CODEX_CLI_BIN=codex
 BRIDGE_WORKDIR=$ROOT_DIR
 EOT
 
 chmod 600 "$SECURE_ENV_FILE"
 
-upsert_env_key "$MOBILE_ENV_FILE" "EXPO_PUBLIC_HOST_BRIDGE_URL" "http://$BRIDGE_HOST:$BRIDGE_PORT"
 upsert_env_key "$MOBILE_ENV_FILE" "EXPO_PUBLIC_HOST_BRIDGE_TOKEN" "$BRIDGE_TOKEN"
-# Backward compatibility for older app builds that still read MAC_BRIDGE keys.
-upsert_env_key "$MOBILE_ENV_FILE" "EXPO_PUBLIC_MAC_BRIDGE_URL" "http://$BRIDGE_HOST:$BRIDGE_PORT"
+# Backward compatibility for older app builds that still read MAC_BRIDGE token key.
 upsert_env_key "$MOBILE_ENV_FILE" "EXPO_PUBLIC_MAC_BRIDGE_TOKEN" "$BRIDGE_TOKEN"
-upsert_env_key "$MOBILE_ENV_FILE" "EXPO_PUBLIC_ALLOW_QUERY_TOKEN_AUTH" "false"
+upsert_env_key "$MOBILE_ENV_FILE" "EXPO_PUBLIC_ALLOW_QUERY_TOKEN_AUTH" "true"
 upsert_env_key "$MOBILE_ENV_FILE" "EXPO_PUBLIC_ALLOW_INSECURE_REMOTE_BRIDGE" "true"
 
 echo "Secure dev setup complete."
 echo ""
+echo "Bridge network mode: $BRIDGE_NETWORK_MODE"
 echo "Bridge host: $BRIDGE_HOST ($HOST_SOURCE)"
 echo "Bridge port: $BRIDGE_PORT"
 echo "Token source: $SECURE_ENV_FILE"