pygpt-net 2.7.8__py3-none-any.whl → 2.7.9__py3-none-any.whl

pygpt_net/CHANGELOG.txt

@@ -1,3 +1,8 @@
+2.7.9 (2026-01-08)
+
+- Improved realtime audio mode.
+- Added xAI provider and Grok support in realtime audio mode.
+
 2.7.8 (2026-01-06)
 
 - Added the xAI Collections remote tool and integrated collections management into the Remote Vector Stores tool.

pygpt_net/__init__.py

@@ -6,15 +6,15 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2026-01-06 00:00:00                  #
+# Updated Date: 2026-01-08 00:00:00                  #
 # ================================================== #
 
 __author__ = "Marcin Szczygliński"
 __copyright__ = "Copyright 2026, Marcin Szczygliński"
 __credits__ = ["Marcin Szczygliński"]
 __license__ = "MIT"
-__version__ = "2.7.8"
-__build__ = "2026-01-06"
+__version__ = "2.7.9"
+__build__ = "2026-01-08"
 __maintainer__ = "Marcin Szczygliński"
 __github__ = "https://github.com/szczyglis-dev/py-gpt"
 __report__ = "https://github.com/szczyglis-dev/py-gpt/issues"

pygpt_net/controller/realtime/realtime.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.09.17 07:00:00                  #
+# Updated Date: 2026.01.07 23:00:00                  #
 # ================================================== #
 
 from PySide6.QtCore import Slot, QTimer
@@ -87,6 +87,8 @@ class Realtime:
                 self.window.core.api.google.realtime.handle_audio_input(event)
             elif self.current_active == "openai":
                 self.window.core.api.openai.realtime.handle_audio_input(event)
+            elif self.current_active == "x_ai":
+                self.window.core.api.xai.realtime.handle_audio_input(event)
 
         # begin: first text chunk or audio chunk received, start rendering
         elif event.name == RealtimeEvent.RT_OUTPUT_READY:
@@ -216,6 +218,8 @@ class Realtime:
             self.window.core.api.google.realtime.manual_commit()
         elif self.current_active == "openai":
             self.window.core.api.openai.realtime.manual_commit()
+        elif self.current_active == "x_ai":
+            self.window.core.api.xai.realtime.manual_commit()
 
     def end_turn(self, ctx):
         """
@@ -252,6 +256,10 @@ class Realtime:
             self.window.core.api.google.realtime.shutdown()
         except Exception as e:
             self.window.core.debug.log(f"[google] Realtime shutdown error: {e}")
+        try:
+            self.window.core.api.xai.realtime.shutdown()
+        except Exception as e:
+            self.window.core.debug.log(f"[xAI] Realtime shutdown error: {e}")
         try:
             self.manager.shutdown()
         except Exception as e:
@@ -267,6 +275,10 @@ class Realtime:
             self.window.core.api.google.realtime.reset()
         except Exception as e:
             self.window.core.debug.log(f"[google] Realtime reset error: {e}")
+        try:
+            self.window.core.api.xai.realtime.reset()
+        except Exception as e:
+            self.window.core.debug.log(f"[xAI] Realtime reset error: {e}")
 
     def is_supported(self) -> bool:
         """

pygpt_net/data/config/config.json

@@ -1,8 +1,8 @@
 {
   "__meta__": {
-    "version": "2.7.8",
-    "app.version": "2.7.8",
-    "updated_at": "2026-01-06T00:00:00"
+    "version": "2.7.9",
+    "app.version": "2.7.9",
+    "updated_at": "2026-01-08T00:00:00"
   },
   "access.audio.event.speech": false,
   "access.audio.event.speech.disabled": [],
@@ -88,6 +88,7 @@
   "api_key_perplexity": "",
   "api_key_voyage": "",
   "api_key_xai": "",
+  "api_key_management_xai": "",
   "api_native_anthropic": true,
   "api_native_google": true,
   "api_native_google.app_credentials": "",

pygpt_net/data/config/models.json

@@ -1,8 +1,8 @@
 {
     "__meta__": {
-        "version": "2.7.8",
-        "app.version": "2.7.8",
-        "updated_at": "2026-01-06T00:00:00"
+        "version": "2.7.9",
+        "app.version": "2.7.9",
+        "updated_at": "2026-01-08T00:00:00"
     },
     "items": {
         "SpeakLeash/bielik-11b-v2.3-instruct:Q4_K_M": {
@@ -1073,9 +1073,9 @@
             "provider": "google",
             "tool_calls": true
         },
-        "gemini-2.5-flash-preview-native-audio-dialog": {
-            "id": "gemini-2.5-flash-preview-native-audio-dialog",
-            "name": "gemini-2.5-flash-preview-native-audio-dialog",
+        "gemini-2.5-flash-native-audio-latest": {
+            "id": "gemini-2.5-flash-native-audio-latest",
+            "name": "gemini-2.5-flash-native-audio-latest",
             "mode": [
                 "audio"
             ],
@@ -1083,7 +1083,7 @@
                 "args": [
                     {
                         "name": "model",
-                        "value": "models/gemini-2.5-flash-preview-native-audio-dialog",
+                        "value": "models/gemini-2.5-flash-native-audio-latest",
                         "type": "str"
                     }
                 ],
@@ -3170,7 +3170,8 @@
                 "agent",
                 "agent_llama",
                 "expert",
-                "agent_openai"
+                "agent_openai",
+                "audio"
             ],
             "llama_index": {
                 "args": [
@@ -3196,10 +3197,12 @@
             "default": false,
             "input": [
                 "text",
-                "image"
+                "image",
+                "audio"
             ],
             "output": [
-                "text"
+                "text",
+                "audio"
             ],
             "extra": {},
             "imported": false,
@@ -3215,7 +3218,8 @@
                 "agent_llama",
                 "agent_openai",
                 "agent",
-                "expert"
+                "expert",
+                "audio"
             ],
             "llama_index": {
                 "args": [
@@ -3243,10 +3247,12 @@
             "default": false,
             "input": [
                 "text",
-                "image"
+                "image",
+                "audio"
             ],
             "output": [
-                "text"
+                "text",
+                "audio"
             ],
             "extra": {},
             "imported": false,
@@ -3262,7 +3268,8 @@
                 "agent_llama",
                 "agent_openai",
                 "agent",
-                "expert"
+                "expert",
+                "audio"
             ],
             "llama_index": {
                 "args": [
@@ -3290,10 +3297,12 @@
             "default": false,
             "input": [
                 "text",
-                "image"
+                "image",
+                "audio"
             ],
             "output": [
-                "text"
+                "text",
+                "audio"
             ],
             "extra": {},
             "imported": false,
@@ -3309,7 +3318,8 @@
                 "agent_llama",
                 "agent_openai",
                 "agent",
-                "expert"
+                "expert",
+                "audio"
             ],
             "llama_index": {
                 "args": [
@@ -3337,10 +3347,12 @@
             "default": false,
             "input": [
                 "text",
-                "image"
+                "image",
+                "audio"
             ],
             "output": [
-                "text"
+                "text",
+                "audio"
             ],
             "extra": {},
             "imported": false,
@@ -3356,7 +3368,8 @@
                 "agent_llama",
                 "agent_openai",
                 "agent",
-                "expert"
+                "expert",
+                "audio"
             ],
             "llama_index": {
                 "args": [
@@ -3384,10 +3397,12 @@
             "default": false,
             "input": [
                 "text",
-                "image"
+                "image",
+                "audio"
             ],
             "output": [
-                "text"
+                "text",
+                "audio"
             ],
             "extra": {},
             "imported": false,

pygpt_net/provider/api/google/realtime/client.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.08.31 23:00:00                  #
+# Updated Date: 2026.01.07 23:00:00                  #
 # ================================================== #
 
 import asyncio
@@ -338,20 +338,25 @@ class GoogleLiveClient:
         if sys_prompt:
             live_cfg["system_instruction"] = str(sys_prompt)
 
-        # Session resumption: enable updates; resume when a different non-empty handle is given
+        # Save callbacks and ctx early so handle persistence can target the current context
+        self._on_text = on_text
+        self._on_audio = on_audio
+        self._should_stop = should_stop or (lambda: False)
+        self._ctx = ctx
+        self._last_opts = opts
+
+        # Session resumption: configure per docs; include handle when provided, otherwise None.
         try:
+            ph = None
             provided_handle = getattr(opts, "rt_session_id", None)
-            resume_handle = None
             if isinstance(provided_handle, str):
-                ph = provided_handle.strip()
-                if ph and ph != (self._rt_session_id or ""):
-                    resume_handle = ph
+                ph = provided_handle.strip() or None
 
-            live_cfg["session_resumption"] = gtypes.SessionResumptionConfig(handle=resume_handle)
+            sr_cfg = gtypes.SessionResumptionConfig(handle=ph)
+            live_cfg["session_resumption"] = sr_cfg
 
-            if resume_handle:
-                self._rt_session_id = resume_handle
-                set_ctx_rt_handle(self._ctx, resume_handle, self.window)
+            if ph:
+                self._persist_rt_handle(ph)
         except Exception:
             pass
 
@@ -360,13 +365,6 @@ class GoogleLiveClient:
         apply_turn_mode_google(live_cfg, turn_mode)
         self._tune_google_vad(live_cfg, opts)
 
-        # Save callbacks and ctx
-        self._on_text = on_text
-        self._on_audio = on_audio
-        self._should_stop = should_stop or (lambda: False)
-        self._ctx = ctx
-        self._last_opts = opts
-
         # Control primitives
         self._response_done = asyncio.Event()
         self._send_lock = asyncio.Lock()
@@ -407,7 +405,7 @@ class GoogleLiveClient:
         self._rt_state = None
         self._last_tool_calls = []
 
-        # Clear only in-memory handle; keep persisted ctx.extra["rt_session_id"]
+        # Clear in-memory handle as well to prevent unintended resumption
         self._rt_session_id = None
 
         # Clear cached tools signature
@@ -820,11 +818,10 @@ class GoogleLiveClient:
                 try:
                     sru = getattr(response, "session_resumption_update", None) or getattr(response, "sessionResumptionUpdate", None)
                     if sru:
-                        resumable = bool(getattr(sru, "resumable", None))
-                        new_handle = getattr(sru, "new_handle", None) or getattr(sru, "newHandle", None)
-                        if resumable and isinstance(new_handle, str) and new_handle.strip():
-                            self._rt_session_id = new_handle.strip()
-                            set_ctx_rt_handle(self._ctx, self._rt_session_id, self.window)
+                        # Prefer robustness: persist handle if present, regardless of 'resumable' flag inconsistencies
+                        new_handle = self._extract_sru_handle(sru)
+                        if isinstance(new_handle, str) and new_handle.strip():
+                            self._persist_rt_handle(new_handle.strip())
                             if self.debug:
                                 print(f"[google.live] session handle updated: {self._rt_session_id}")
                 except Exception:
@@ -1740,6 +1737,10 @@ class GoogleLiveClient:
         """
         self.debug = bool(enabled)
 
+    def is_session(self) -> bool:
+        """Check if the WS session is currently open."""
+        return self._session is not None
+
     def is_session_active(self) -> bool:
         """Check if the WS session is currently open."""
         return self._session is not None
@@ -1748,6 +1749,12 @@ class GoogleLiveClient:
         """Update the current CtxItem (for session handle persistence)."""
         self._ctx = ctx
 
+    def get_current_rt_session_id(self) -> Optional[str]:
+        """
+        Return the current resumable session handle if known.
+        """
+        return self._rt_session_id
+
     # -----------------------------
     # Internal: auto-turn receiver bootstrap
     # -----------------------------
@@ -1942,4 +1949,43 @@ class GoogleLiveClient:
         """
         Emit RT_OUTPUT_AUDIO_COMMIT on first sign of model output in auto-turn mode.
         """
-        self._emit_audio_commit_signal()
+        self._emit_audio_commit_signal()
+
+    # -----------------------------
+    # Internal: session handle helpers
+    # -----------------------------
+
+    def _persist_rt_handle(self, handle: str) -> None:
+        """
+        Persist current session handle in-memory, to ctx.extra and into last opts for future restarts.
+        """
+        try:
+            self._rt_session_id = handle
+            set_ctx_rt_handle(self._ctx, handle, self.window)
+        except Exception:
+            pass
+        try:
+            if self._last_opts is not None:
+                setattr(self._last_opts, "rt_session_id", handle)
+        except Exception:
+            pass
+
+    def _extract_sru_handle(self, sru: Any) -> Optional[str]:
+        """
+        Extract handle from SessionResumptionUpdate (supports snake_case and camelCase, and token alias).
+        """
+        # Objects (attrs)
+        for attr in ("new_handle", "newHandle", "token"):
+            try:
+                v = getattr(sru, attr, None)
+                if isinstance(v, str) and v.strip():
+                    return v.strip()
+            except Exception:
+                pass
+        # Dicts
+        if isinstance(sru, dict):
+            for k in ("new_handle", "newHandle", "token"):
+                v = sru.get(k)
+                if isinstance(v, str) and v.strip():
+                    return v.strip()
+        return None

pygpt_net/provider/api/google/realtime/realtime.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2026.01.02 19:00:00                  #
+# Updated Date: 2026.01.07 23:00:00                  #
 # ================================================== #
 
 import json
@@ -86,8 +86,55 @@ class Realtime:
                     self.handler.send_tool_results_sync({
                         tool_call_id: tool_results
                     })
+                    self.handler.update_ctx(context.ctx)
                     return True  # do not start new session, just send tool results
 
+        # Tools
+        tools = self.window.core.api.google.tools.prepare(model, context.external_functions)
+        remote_tools = self.window.core.api.google.remote_tools.build_remote_tools(model)
+        if tools:
+            remote_tools = []  # in Google, remote tools are not allowed if function calling is used
+
+        # Resolve last session ID, prefer history, then fallback to current ctx and in-memory handler handle
+        last_session_id = extract_last_session_id(context.history) if context.history else None
+        if not last_session_id:
+            try:
+                if context.ctx and isinstance(context.ctx.extra, dict):
+                    sid = context.ctx.extra.get("rt_session_id")
+                    if isinstance(sid, str) and sid.strip():
+                        last_session_id = sid.strip()
+            except Exception:
+                pass
+        if not last_session_id and self.handler.is_session_active():
+            try:
+                sid = self.handler.get_current_rt_session_id()
+                if isinstance(sid, str) and sid.strip():
+                    last_session_id = sid.strip()
+            except Exception:
+                pass
+
+        if is_debug:
+            print("[realtime session] Last ID", last_session_id)
+
+        # Enforce clean state rules:
+        # - No history: always reset to ensure a fresh server context.
+        # - If history exists, keep the current live session even if the resumable handle has not been captured yet.
+        #   Gemini Live can emit the handle slightly after the first turn starts; closing here would drop context continuity.
+        try:
+            history_len = len(context.history) if context.history else 0
+        except Exception:
+            history_len = 0
+
+        if history_len == 0:
+            if self.handler.is_session_active():
+                self.handler.close_session_sync()
+            try:
+                if context.ctx and isinstance(context.ctx.extra, dict):
+                    context.ctx.extra.pop("rt_session_id", None)
+            except Exception:
+                pass
+            last_session_id = None  # force new session
+
         # update auto-turn in active session
         if (self.handler.is_session_active()
                 and (auto_turn != self.prev_auto_turn
@@ -95,23 +142,12 @@ class Realtime:
                      or opt_vad_prefix != self.prev_vad_prefix)):
             self.handler.update_session_autoturn_sync(auto_turn, opt_vad_silence, opt_vad_prefix)
 
-        # Tools
-        tools = self.window.core.api.google.tools.prepare(model, context.external_functions)
-        remote_tools = self.window.core.api.google.remote_tools.build_remote_tools(model)
-        if tools:
-            remote_tools = []  # in Google, remote tools are not allowed if function calling is used
-
         # if auto-turn is enabled and prompt is empty, update session and context only
         if auto_turn and self.handler.is_session_active() and (context.prompt.strip() == "" or context.prompt == "..."):
             self.handler.update_session_tools_sync(tools, remote_tools)
             self.handler.update_ctx(context.ctx)
             return True  # do not send new request if session is active
 
-        # Last session ID
-        last_session_id = extract_last_session_id(context.history)
-        if is_debug:
-            print("[realtime session] Last ID", last_session_id)
-
         # Voice
         voice_name = "Kore"
         try:

pygpt_net/provider/api/openai/realtime/realtime.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.09.01 23:00:00                  #
+# Updated Date: 2026.01.07 23:00:00                  #
 # ================================================== #
 
 import json
@@ -102,6 +102,31 @@ class Realtime:
                     self.handler.update_ctx(context.ctx)
                     return True  # do not start new session, just send tool results
 
+        # Resolve last session ID from history only (do not fallback anywhere)
+        last_session_id = extract_last_session_id(context.history) if context.history else None
+        if is_debug:
+            print("[realtime session] Last ID", last_session_id)
+
+        # Enforce clean state rules before any live updates:
+        # - If there is no history at all: always reset live session to ensure a fresh context.
+        # - If there is history but it has no resumable session id: close any active session to avoid accidental continuation.
+        try:
+            history_len = len(context.history) if context.history else 0
+        except Exception:
+            history_len = 0
+
+        if history_len == 0:
+            if self.handler.is_session_active():
+                self.handler.close_session_sync()
+            try:
+                if context.ctx and isinstance(context.ctx.extra, dict):
+                    context.ctx.extra.pop("rt_session_id", None)
+            except Exception:
+                pass
+            last_session_id = None  # force new session
+        elif not last_session_id and self.handler.is_session_active():
+            self.handler.close_session_sync()
+
         # update auto-turn in active session
         if (self.handler.is_session_active()
                 and (auto_turn != self.prev_auto_turn
@@ -116,11 +141,6 @@ class Realtime:
             self.window.update_status(trans("speech.listening"))
             return True # do not send new request if session is active
 
-        # Last session ID
-        last_session_id = extract_last_session_id(context.history)
-        if is_debug:
-            print("[realtime session] Last ID", last_session_id)
-
         # Voice
         voice = "alloy"
         try:

pygpt_net/provider/api/x_ai/__init__.py

@@ -141,9 +141,6 @@ class ApiXAI:
                 MODE_RESEARCH,
                 MODE_AUDIO
         ):
-            if mode == MODE_AUDIO:
-                raise NotImplementedError("Not available. xAI realtime audio streaming coming soon!")
-
             if mode == MODE_AUDIO and stream:
                 # Realtime API for audio streaming
                 is_realtime = self.realtime.begin(