pygpt-net 2.5.17__py3-none-any.whl → 2.5.18__py3-none-any.whl

pygpt_net/CHANGELOG.txt

@@ -1,3 +1,10 @@
+2.5.18 (2025-06-26)
+
+- Non-GPT models are now available in standard Chat mode.
+- Added a new remote tool: `image_generation` in Responses API -> disabled by default, enable in `Config -> Settings -> Remote Tools`. Enables native image generation and editing of uploaded images in Chat mode.
+- Added a new model `gpt-image-1` and improved image generation.
+- Other small fixes.
+
 2.5.17 (2025-06-25)
 
 - Added settings for enable/disable Remote Tools via Responses API in Chat mode: Config -> Settings -> Remote tools. Currently only web-search-preview tool is available, rest of tools coming soon.

pygpt_net/__init__.py

@@ -6,15 +6,15 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025-06-25 02:00:00                  #
+# Updated Date: 2025-06-26 02:00:00                  #
 # ================================================== #
 
 __author__ = "Marcin Szczygliński"
 __copyright__ = "Copyright 2025, Marcin Szczygliński"
 __credits__ = ["Marcin Szczygliński"]
 __license__ = "MIT"
-__version__ = "2.5.17"
-__build__ = "2025-06-25"
+__version__ = "2.5.18"
+__build__ = "2025-06-26"
 __maintainer__ = "Marcin Szczygliński"
 __github__ = "https://github.com/szczyglis-dev/py-gpt"
 __report__ = "https://github.com/szczyglis-dev/py-gpt/issues"

pygpt_net/controller/chat/common.py

@@ -282,15 +282,17 @@ class Common:
         if not exit:
             self.window.dispatch(AppEvent(AppEvent.INPUT_STOPPED))  # app event
 
-    def check_api_key(self) -> bool:
+    def check_api_key(self, monit: bool = True) -> bool:
         """
         Check if API KEY is set
 
+        :param monit: True if monitor should be shown
         :return: True if API KEY is set, False otherwise
         """
         result = True
         if self.window.core.config.get('api_key') is None or self.window.core.config.get('api_key') == '':
-            self.window.controller.launcher.show_api_monit()
+            if monit:
+                self.window.controller.launcher.show_api_monit()
             self.window.update_status("Missing API KEY!")
             result = False
         return result

pygpt_net/controller/chat/input.py

@@ -6,8 +6,9 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.06.24 16:00:00                  #
+# Updated Date: 2025.06.26 16:00:00                  #
 # ================================================== #
+
 import os
 from typing import Optional, Any, Dict
 
@@ -20,6 +21,7 @@ from pygpt_net.core.types import (
     MODE_LLAMA_INDEX,
     MODE_ASSISTANT,
     MODE_IMAGE,
+    MODE_CHAT,
 )
 from pygpt_net.item.ctx import CtxItem
 from pygpt_net.utils import trans
@@ -76,27 +78,29 @@ class Input:
 
             # check ollama model
             model = self.window.core.config.get('model')
-            if mode == MODE_LLAMA_INDEX and model is not None:
+            if model:
                 model_data = self.window.core.models.get(model)
                 if model_data is not None and model_data.is_ollama():
-                    model_id = model_data.get_ollama_model()
-                    # load ENV vars first
-                    if ('env' in model_data.llama_index
-                            and model_data.llama_index['env'] is not None):
-                        for item in model_data.llama_index['env']:
-                            key = item.get('name', '').strip()
-                            value = item.get('value', '').strip()
-                            os.environ[key] = value
-                    status = self.window.core.models.ollama.check_model(model_id)
-                    is_installed = status.get('is_installed', False)
-                    is_model = status.get('is_model', False)
-                    if not is_installed:
-                        self.window.ui.dialogs.alert(trans("dialog.ollama.not_installed"))
-                        return
-                    if not is_model:
-                        self.window.ui.dialogs.alert(
-                            trans("dialog.ollama.model_not_found").replace("{model}", model_id))
-                        return
+                    if (mode == MODE_LLAMA_INDEX or
+                            (mode == MODE_CHAT and not model_data.is_openai() and model_data.is_ollama())):
+                        model_id = model_data.get_ollama_model()
+                        # load ENV vars first
+                        if ('env' in model_data.llama_index
+                                and model_data.llama_index['env'] is not None):
+                            for item in model_data.llama_index['env']:
+                                key = item.get('name', '').strip()
+                                value = item.get('value', '').strip()
+                                os.environ[key] = value
+                        status = self.window.core.models.ollama.check_model(model_id)
+                        is_installed = status.get('is_installed', False)
+                        is_model = status.get('is_model', False)
+                        if not is_installed:
+                            self.window.ui.dialogs.alert(trans("dialog.ollama.not_installed"))
+                            return
+                        if not is_model:
+                            self.window.ui.dialogs.alert(
+                                trans("dialog.ollama.model_not_found").replace("{model}", model_id))
+                            return
 
         # listen for stop command
         if self.generating \
@@ -247,14 +251,19 @@ class Input:
             self.generating = False  # unlock as not generating
             return
 
-        # check API key, show monit if no API key
+        # check OpenAI API key, show monit if no API key
         if mode not in self.window.controller.launcher.no_api_key_allowed:
-            if not self.window.controller.chat.common.check_api_key():
-                self.generating = False
-                self.window.dispatch(KernelEvent(KernelEvent.STATE_ERROR, {
-                    "id": "chat",
-                }))
-                return
+            if not self.window.controller.chat.common.check_api_key(monit=False):
+                model = self.window.core.config.get('model')
+                if model:
+                    model_data = self.window.core.models.get(model)
+                    if model_data is not None and model_data.is_openai():
+                        self.window.controller.chat.common.check_api_key(monit=True)
+                        self.generating = False
+                        self.window.dispatch(KernelEvent(KernelEvent.STATE_ERROR, {
+                            "id": "chat",
+                        }))
+                        return
 
         # set state to: busy
         self.window.dispatch(KernelEvent(KernelEvent.STATE_BUSY, {

pygpt_net/controller/chat/stream.py

@@ -8,7 +8,8 @@
 # Created By  : Marcin Szczygliński                  #
 # Updated Date: 2025.06.25 02:00:00                  #
 # ================================================== #
-
+import base64
+import uuid
 from typing import Any
 
 from PySide6.QtWidgets import QApplication
@@ -39,6 +40,8 @@ class Stream:
         tool_calls = []
         fn_args_buffers = {}
         citations = []
+        img_path = self.window.core.image.gen_unique_path(ctx)
+        is_image = False
 
         # chunks: stream begin
         data = {
@@ -121,8 +124,8 @@ class Stream:
                                 if tool_chunk.function.arguments:
                                     tool_call["function"]["arguments"] += tool_chunk.function.arguments
 
+                    # OpenAI Responses API
                     elif chunk_type == "api_chat_responses":
-
                         if etype == "response.output_text.delta":
                             response = chunk.delta
 
@@ -154,6 +157,19 @@ class Stream:
                                 citations.append(url_citation)
                                 ctx.urls = citations
 
+                        # ---------- image gen ----------
+                        elif etype == "response.image_generation_call.partial_image":
+                            idx = chunk.partial_image_index
+                            image_base64 = chunk.partial_image_b64
+                            image_bytes = base64.b64decode(image_base64)
+                            with open(img_path, "wb") as f:
+                                f.write(image_bytes)
+                            is_image = True
+
+                        # ---------- response ID ----------
+                        elif etype == "response.created":
+                            ctx.msg_id = chunk.response.id # store previous response ID
+
                         # ---------- end / error ----------
                         elif etype in {"response.done", "response.failed", "error"}:
                             pass
@@ -214,6 +230,10 @@ class Stream:
                 if tool_calls:
                     self.window.core.command.unpack_tool_calls_chunks(ctx, tool_calls)
 
+                # append images
+                if is_image:
+                    ctx.images = [img_path]  # save image path to ctx
+
         except Exception as e:
             self.window.core.debug.log(e)
             error = e

pygpt_net/controller/config/placeholder.py

@@ -270,7 +270,7 @@ class Placeholder:
             model = models[id]
             suffix = ""
             if "provider" in model.llama_index and model.llama_index["provider"] == "ollama":
-                suffix = " [Ollama]"
+                suffix = " (Ollama)"
             name = model.name + suffix
             data.append({id: name})
         return data

pygpt_net/controller/model/__init__.py

@@ -161,7 +161,7 @@ class Model:
         for k in data:
             suffix = ""
             if "provider" in data[k].llama_index and data[k].llama_index["provider"] == "ollama":
-                suffix = " [Ollama]"
+                suffix = " (Ollama)"
             items[k] = data[k].name + suffix
         items = dict(sorted(items.items(), key=lambda item: item[1]))  # sort items by name
         self.window.ui.nodes["prompt.model"].set_keys(items)

pygpt_net/controller/model/editor.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2024.12.14 08:00:00                  #
+# Updated Date: 2025.06.26 16:00:00                  #
 # ================================================== #
 
 import copy
@@ -56,6 +56,11 @@ class Editor:
                 "type": "bool",
                 "label": "model.default",
             },
+            "openai": {
+                "type": "bool",
+                "label": "model.openai",
+                "description": "model.openai.desc",
+            },
             "langchain.provider": {
                 "type": "combo",
                 "use": "langchain_providers",

pygpt_net/controller/model/importer.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.06.24 02:00:00                  #
+# Updated Date: 2025.06.26 16:00:00                  #
 # ================================================== #
 
 import copy
@@ -200,7 +200,7 @@ class Importer:
         """
         Get current ollama models
 
-        :return: ollama models dictionary
+        :return: PyGPT ollama models dictionary
         """
         items = copy.deepcopy(self.window.core.models.items)
         for key in list(items.keys()):
@@ -214,7 +214,7 @@ class Importer:
         """
         Get available ollama models
 
-        :return: ollama models dictionary
+        :return: Ollama API models dictionary
         """
         models = {}
         status = self.window.core.models.ollama.get_status()
@@ -233,6 +233,7 @@ class Importer:
                     m.id = name
                     m.name = name
                     m.mode = [
+                        "chat",
                         "llama_index",
                         "agent",
                         "agent_llama",

pygpt_net/core/bridge/__init__.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.03.02 19:00:00                  #
+# Updated Date: 2025.06.26 16:00:00                  #
 # ================================================== #
 
 import time
@@ -93,9 +93,11 @@ class Bridge:
         if model is not None:
             if not model.is_supported(mode):  # check selected mode
                 mode = self.window.core.models.get_supported_mode(model, mode)  # switch
+                if base_mode == MODE_CHAT and mode == MODE_LLAMA_INDEX:
+                    context.idx = None # disable index if in Chat mode and switch to Llama Index
 
         if mode == MODE_LLAMA_INDEX and base_mode != MODE_LLAMA_INDEX:
-            context.idx_mode = MODE_CHAT
+            context.idx_mode = MODE_CHAT  # default in sub-mode
 
         if is_virtual:
             if mode == MODE_LLAMA_INDEX:  # after switch
@@ -192,11 +194,13 @@ class Bridge:
                 debug = {k: str(v) for k, v in context.to_dict().items()}
                 self.window.core.debug.debug(str(debug))
 
+        # --- DEBUG ONLY ---
         self.last_context_quick = context  # store last context for quick call (debug)
 
         if context.model is not None:
-            # check if model is supported by chat API, if not then try to use llama-index or langchain call
-            if not context.model.is_supported(MODE_CHAT) and not context.model.is_supported(MODE_RESEARCH):
+            # check if model is supported by OpenAI API, if not then try to use llama-index or langchain call
+            if ((not context.model.is_supported(MODE_CHAT) or not context.model.is_openai())
+                    and not context.model.is_supported(MODE_RESEARCH)):
 
                 # tmp switch to: llama-index
                 if context.model.is_supported(MODE_LLAMA_INDEX):

pygpt_net/core/command/__init__.py

@@ -241,6 +241,8 @@ class Command:
         parsed = []
         for tool_call in tool_calls:
             try:
+                if not hasattr(tool_call, 'name'):
+                    continue
                 parsed.append(
                     {
                         "id": tool_call.id,
@@ -615,7 +617,7 @@ class Command:
         :return: True if enabled
         """
         disabled_modes = [
-            MODE_LLAMA_INDEX,
+            #MODE_LLAMA_INDEX,
             MODE_LANGCHAIN,
             MODE_COMPLETION,
         ]
@@ -624,6 +626,13 @@ class Command:
             return False  # disabled for specific modes
         if self.window.controller.agent.legacy.enabled() or self.window.controller.agent.experts.enabled():
             return False
+        model = self.window.core.config.get('model')
+        if model:
+            model_data = self.window.core.models.get(model)
+            if model_data:
+                llama_provider = model_data.get_llama_provider()
+                if llama_provider in self.window.core.idx.chat.tool_calls_not_allowed_providers:
+                    return False
         return self.window.core.config.get('func_call.native', False)  # otherwise check config
 
     def is_enabled(self, cmd: str) -> bool:

pygpt_net/core/idx/chat.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.06.21 23:00:00                  #
+# Updated Date: 2025.06.26 16:00:00                  #
 # ================================================== #
 
 import json
@@ -39,6 +39,7 @@ class Chat:
         self.context = Context(window)
         self.tool_calls_not_allowed_providers = [
             "ollama",
+            "google",
             "hugging_face_api",
             "deepseek_api",
         ]
@@ -237,6 +238,7 @@ class Chat:
         use_index = True
         verbose = self.window.core.config.get("log.llama", False)
         allow_native_tool_calls = True
+        response = None
         if ('provider' in model.llama_index
                 and model.llama_index['provider'] in self.tool_calls_not_allowed_providers):
             allow_native_tool_calls = False
@@ -246,10 +248,12 @@ class Chat:
             use_index = False
 
         # disable index if no api key
+        """
         if self.window.core.config.get("api_key") == "" and self.window.core.config.get("llama.idx.embeddings.provider") == "openai":
             print("Warning: no api key! Disabling index...")
             chat_mode = "simple"  # do not use query engine if no index
             use_index = False
+        """
 
         if model is None or not isinstance(model, ModelItem):
             raise Exception("Model config not provided")
@@ -322,6 +326,7 @@ class Chat:
             if stream:
                 # IMPORTANT: stream chat with tools not supported by all providers
                 if allow_native_tool_calls and hasattr(llm, "stream_chat_with_tools"):
+                    self.log("Using with tools...")
                     response = llm.stream_chat_with_tools(
                         tools=tools,
                         messages=history,

pygpt_net/core/image/__init__.py

@@ -10,6 +10,8 @@
 # ================================================== #
 
 import os
+import uuid
+from time import strftime
 from typing import List
 
 from PySide6.QtCore import Slot, QObject
@@ -126,3 +128,16 @@ class Image(QObject):
             else:
                 return "_"
         return "".join(safe_char(c) for c in name).rstrip("_")[:30]
+
+    def gen_unique_path(self, ctx: CtxItem):
+        """
+        Generate unique image path based on context
+
+        :param ctx: CtxItem
+        :return: unique image path
+        """
+        img_id = uuid.uuid4()
+        dt_prefix = strftime("%Y%m%d_%H%M%S")
+        img_dir = self.window.core.config.get_user_dir("img")
+        filename = f"{dt_prefix}_{img_id}.png"
+        return os.path.join(img_dir, filename)

pygpt_net/core/models/__init__.py

@@ -6,7 +6,7 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.06.24 02:00:00                  #
+# Updated Date: 2025.06.26 16:00:00                  #
 # ================================================== #
 
 import copy
@@ -385,18 +385,26 @@ class Models:
         :param mode: mode (initial)
         :return: mode (supported)
         """
-        if model.is_supported(MODE_CHAT):
-            self.window.core.debug.info(
-                "WARNING: Switching to chat mode (model not supported in: {})".format(mode))
-            mode = MODE_CHAT
-        elif model.is_supported(MODE_RESEARCH):
+        # if OpenAI API model and not llama_index mode, switch to Chat mode
+        if model.is_openai():
+            if model.is_supported(MODE_CHAT) and mode != MODE_LLAMA_INDEX:  # do not switch if llama_index mode!
+                self.window.core.debug.info(
+                    "WARNING: Switching to chat mode (model not supported in: {})".format(mode))
+                return MODE_CHAT
+
+        # Research / Perplexity
+        if model.is_supported(MODE_RESEARCH):
             self.window.core.debug.info(
                 "WARNING: Switching to research mode (model not supported in: {})".format(mode))
             mode = MODE_RESEARCH
+
+        # Llama Index / Chat with Files
         elif model.is_supported(MODE_LLAMA_INDEX):
             self.window.core.debug.info(
                 "WARNING: Switching to llama_index mode (model not supported in: {})".format(mode))
             mode = MODE_LLAMA_INDEX
+
+        # LangChain
         elif model.is_supported(MODE_LANGCHAIN):
             self.window.core.debug.info(
                 "WARNING: Switching to langchain mode (model not supported in: {})".format(mode))

pygpt_net/core/models/ollama.py

@@ -6,10 +6,10 @@
 # GitHub:  https://github.com/szczyglis-dev/py-gpt   #
 # MIT License                                        #
 # Created By  : Marcin Szczygliński                  #
-# Updated Date: 2025.06.24 16:00:00                  #
+# Updated Date: 2025.06.26 16:00:00                  #
 # ================================================== #
-import os
 
+import os
 import requests
 
 class Ollama:
@@ -31,7 +31,8 @@ class Ollama:
         api_base = "http://localhost:11434"
         if 'OLLAMA_API_BASE' in os.environ:
             api_base = os.environ['OLLAMA_API_BASE']
-        print("Using Ollama base URL:", api_base)
+        self.window.core.idx.log("Using Ollama base URL: {}".format(api_base))
+
         url = api_base + "/api/tags"
         try:
             response = requests.get(url, timeout=2)

pygpt_net/data/config/config.json

@@ -1,8 +1,8 @@
 {
   "__meta__": {
-    "version": "2.5.17",
-    "app.version": "2.5.17",
-    "updated_at": "2025-06-25T00:00:00"
+    "version": "2.5.18",
+    "app.version": "2.5.18",
+    "updated_at": "2025-06-26T00:00:00"
   },
   "access.audio.event.speech": false,
   "access.audio.event.speech.disabled": [],
@@ -308,6 +308,7 @@
   "render.plain": false,
   "render.code_syntax": "github-dark",
   "remote_tools.web_search": true,
+  "remote_tools.image": false,
   "send_clear": true,
   "send_mode": 2,
   "store_history": true,