voice-vibecoder 2.17.4__tar.gz → 2.17.6__tar.gz

{voice_vibecoder-2.17.4 → voice_vibecoder-2.17.6}/CHANGELOG.md

@@ -2,6 +2,22 @@
 
 <!-- version list -->
 
+## v2.17.6 (2026-02-23)
+
+### Bug Fixes
+
+- **server**: Make audio responses opt-in to reduce bandwidth
+  ([`5816035`](https://github.com/snokam/voice-vibecoder/commit/5816035f2590ed92297c83e1a59c65ec92578b07))
+
+
+## v2.17.5 (2026-02-23)
+
+### Bug Fixes
+
+- **server**: Make ui_command messages opt-in to reduce bandwidth
+  ([`d1e3589`](https://github.com/snokam/voice-vibecoder/commit/d1e358932f07a52feded5458996d2f51f9f7e4d9))
+
+
 ## v2.17.4 (2026-02-22)
 
 ### Bug Fixes

{voice_vibecoder-2.17.4 → voice_vibecoder-2.17.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: voice-vibecoder
-Version: 2.17.4
+Version: 2.17.6
 Summary: Olaf The Vibecoder — a voice-controlled coding assistant using OpenAI Realtime API + Claude Code
 Project-URL: Homepage, https://github.com/snokam/voice-vibecoder
 Project-URL: Repository, https://github.com/snokam/voice-vibecoder

{voice_vibecoder-2.17.4 → voice_vibecoder-2.17.6}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "voice-vibecoder"
-version = "2.17.4"
+version = "2.17.6"
 description = "Olaf The Vibecoder — a voice-controlled coding assistant using OpenAI Realtime API + Claude Code"
 readme = "README.md"
 requires-python = ">=3.10"

{voice_vibecoder-2.17.4 → voice_vibecoder-2.17.6}/src/voice_vibecoder/__init__.py

@@ -3,7 +3,7 @@
 A voice-controlled coding assistant using OpenAI Realtime API + Claude Code.
 """
 
-__version__ = "2.17.4"
+__version__ = "2.17.6"
 
 from voice_vibecoder.app_config import VoiceConfig
 from voice_vibecoder.ui.screen import VoiceCodingScreen

{voice_vibecoder-2.17.4 → voice_vibecoder-2.17.6}/src/voice_vibecoder/config.py

@@ -48,6 +48,10 @@ med beskrivende navn.
 - Si aldri "jeg sender dette til en agent" — si heller "Ser på det!" / "Et øyeblikk."
 - Videresend brukerens ord NØYAKTIG til agenten. Ikke omskriv.
 
+Hold samtalen i gang mens du venter — ikke la det bli stille. \
+Småprat, kommentarer, humor — hva som helst som føles naturlig. \
+Hvis en forespørsel tar tid, forklar kort hva du tenker eller gjør (uten å nevne agenter).
+
 Når agenten er ferdig, lever resultatet som om DU fant svaret. \
 Oppsummer kort i én-to setninger.
 
@@ -97,6 +101,10 @@ with a descriptive name.
 - Never say "I'm sending this to an agent" — say "Looking into it!" / "One moment."
 - Forward the user's words EXACTLY to the agent. Don't rewrite.
 
+Keep the conversation going while you wait — never let it go silent. \
+Small talk, comments, humor — whatever feels natural. \
+If a request takes time, briefly explain what you're thinking or doing (without mentioning agents).
+
 When the agent finishes, deliver the result as if YOU found the answer. \
 Summarize briefly in one or two sentences.
 
@@ -146,6 +154,10 @@ med beskrivande namn.
 - Säg aldrig "jag skickar detta till en agent" — säg hellre "Kollar på det!" / "Ett ögonblick."
 - Vidarebefordra användarens ord EXAKT till agenten. Skriv inte om.
 
+Håll samtalet igång medan du väntar — låt det aldrig bli tyst. \
+Småprat, kommentarer, humor — vad som helst som känns naturligt. \
+Om en förfrågan tar tid, förklara kort vad du tänker eller gör (utan att nämna agenter).
+
 När agenten är klar, leverera resultatet som om DU hittade svaret. \
 Sammanfatta kort i en till två meningar.
 

{voice_vibecoder-2.17.4 → voice_vibecoder-2.17.6}/src/voice_vibecoder/server.py

@@ -6,7 +6,7 @@ The browser sends/receives JSON frames — never talks to OpenAI directly.
 
 Protocol (client -> server):
   {"type": "auth", "token": "<JWT>"}           — must be the first message when auth is enabled
-  {"type": "sync_state"}                       — request full state sync from backend
+  {"type": "sync_state", "audio": true, "ui_commands": true}  — full state sync (audio/ui_commands opt in)
   {"type": "audio", "data": "<base64 PCM16 24kHz>"}
   {"type": "audio.commit"}
   {"type": "update_settings", "settings": {...}}  — partial settings update, reconnects provider
@@ -109,6 +109,8 @@ class WebSocketBridge:
         self._user_email = user_email
         self._user_jwt = user_jwt
         self._messages_task: asyncio.Task | None = None
+        self._ui_commands_enabled = False
+        self._audio_enabled = False
 
     async def run(self) -> None:
         provider = create_provider(self._settings)
@@ -179,7 +181,12 @@ class WebSocketBridge:
                 elif msg_type == "update_settings":
                     await self._handle_settings_update(msg.get("settings", {}))
                 elif msg_type == "sync_state":
-                    # Frontend requests full state sync from backend
+                    # Frontend requests full state sync from backend.
+                    # Opt-in to ui_command messages (file_tree, show_diff, etc.)
+                    if msg.get("ui_commands"):
+                        self._ui_commands_enabled = True
+                    if msg.get("audio"):
+                        self._audio_enabled = True
                     self._send({"type": "status", "status": "syncing"})
                     await self._sync_state_to_frontend()
                     self._send({"type": "status", "status": "ready"})
@@ -286,6 +293,8 @@ class WebSocketBridge:
             logger.warning("Failed to send WS frame", exc_info=True)
 
     def _on_audio(self, base64_chunk: str) -> None:
+        if not self._audio_enabled:
+            return
         self._send({"type": "audio", "data": base64_chunk})
 
     def _on_transcript(self, role: str, text: str) -> None:
@@ -315,6 +324,8 @@ class WebSocketBridge:
         self._send({"type": "tool_call", "name": name, "arguments": arguments})
 
     def _on_ui_command(self, instance_id: str, action: str, data: Any) -> None:
+        if not self._ui_commands_enabled:
+            return
         self._send({"type": "ui_command", "instanceId": instance_id, "action": action, "data": data})
 
     def _on_interrupt(self) -> None: