ollama provider fixes, custom model fixes #168

src/pygpt_net/controller/chat/common.py

@@ -343,6 +343,9 @@ def check_api_key(
         """
         if model is None:
             return True
+        # Ollama and other local providers do not require an API key
+        if model.is_ollama():
+            return True
 
         config = self.window.core.config
         provider_keys = {

src/pygpt_net/controller/model/model.py

@@ -58,7 +58,7 @@ def handle(self, event: BaseEvent):
                                         mode == MODE_CHAT and not model_data.is_openai_supported() and model_data.is_ollama()
                                 )
                         ):
-                            model_id = model_data.get_ollama_model()
+                            model_id = (model_data.get_ollama_model() or "").strip()
 
                             # load ENV vars first
                             if ('env' in model_data.llama_index
@@ -76,8 +76,9 @@ def handle(self, event: BaseEvent):
                                 return
                             if not is_model:
                                 event.data["stop"] = True  # stop flow
+                                display_name = model_id or model_data.id or model or ""
                                 self.window.ui.dialogs.alert(
-                                    trans("dialog.ollama.model_not_found").replace("{model}", model_id))
+                                    trans("dialog.ollama.model_not_found").replace("{model}", display_name))
                                 return
 
         # on input before

src/pygpt_net/core/idx/response.py

@@ -125,13 +125,14 @@ def from_llm(
         :param llm: LLM instance
         :param response: Response data
         """
-        output = response.message.content
+        msg = getattr(response, "message", None)
+        output = (getattr(msg, "content", None) if msg else None) or ""
+        if isinstance(output, str):
+            output = output.strip() or output
         tool_calls = llm.get_tool_calls_from_response(
             response,
             error_on_no_tool_call=False,
         )
-        if output is None:
-            output = ""
         ctx.set_output(output, "")
         ctx.tool_calls = self.window.core.command.unpack_tool_calls_from_llama(tool_calls)
 

src/pygpt_net/core/models/ollama.py

@@ -73,6 +73,9 @@ def check_model(self, model: str) -> dict:
             'is_installed': False,
             'is_model': False,
         }
+        if not model:
+            return result
+        model = model.strip()
         if model in self.available_models:
             return {
                 'is_installed': True,
@@ -84,18 +87,20 @@ def check_model(self, model: str) -> dict:
         if "models" not in status:
             return result
         for item in status.get('models', []):
-            model_id = item.get('name')
+            raw_name = item.get('name') or ''
+            model_id = raw_name.strip()
             if model_id not in self.available_models:
                 self.available_models.append(model_id)
             if model_id == model:
                 return {
                     'is_installed': True,
                     'is_model': True,
                 }
-            model_id = item.get('name').replace(":latest", "")  # handle latest tag
-            if model_id not in self.available_models:
-                self.available_models.append(model_id)
-            if model_id == model:
+            # handle :latest tag (compare without it)
+            model_id_no_latest = model_id.replace(":latest", "").strip()
+            if model_id_no_latest not in self.available_models:
+                self.available_models.append(model_id_no_latest)
+            if model_id_no_latest == model or model_id == model:
                 return {
                     'is_installed': True,
                     'is_model': True,

src/pygpt_net/item/model.py

@@ -231,15 +231,15 @@ def get_provider(self) -> str:
 
     def get_ollama_model(self) -> str:
         """
-        Get Ollama model ID
+        Get Ollama model ID (name used for ollama pull / API).
 
-        :return: model ID
+        :return: model ID, or self.id when no "model" arg is set (e.g. custom/non-default list)
         """
         if "args" in self.llama_index:
             for arg in self.llama_index["args"]:
-                if arg["name"] == "model":
-                    return arg["value"]
-        return ""
+                if arg["name"] == "model" and arg.get("value"):
+                    return (arg["value"] or "").strip()
+        return (self.id or "").strip()
 
     def has_mode(self, mode: str) -> bool:
         """

src/pygpt_net/provider/api/openai/__init__.py

@@ -332,10 +332,15 @@ def quick_call(
         if len(tools) > 0 and "disable_tools" not in extra:
             additional_kwargs["tools"] = tools
 
+        model_id = (model.get_ollama_model() or model.id or "").strip() if model.is_ollama() else (model.id or "")
+        if not model_id:
+            self.locked = False
+            return None
+
         try:
             response = client.chat.completions.create(
                 messages=messages,
-                model=model.id,
+                model=model_id,
                 temperature=temperature,
                 top_p=1.0,
                 frequency_penalty=0.0,

src/pygpt_net/provider/api/openai/chat.py

@@ -146,9 +146,14 @@ def send(
             response_kwargs['stream_options'] = {"include_usage": True}
 
         # OpenRouter: add web search remote tool (if enabled)
-        model_id = model.id
+        # Ollama: use actual Ollama model name (get_ollama_model), not just list key
+        model_id = model.id or ""
         if model.provider == "open_router":
             model_id = self.window.core.models.get_openrouter_model(model)
+        elif model.is_ollama():
+            model_id = (model.get_ollama_model() or model.id or "").strip()
+        if not model_id:
+            raise ValueError("Model name is required for the API request. Check that the selected model has a valid model ID.")
 
         response = client.chat.completions.create(
             messages=messages,

src/pygpt_net/provider/api/openai/responses.py

@@ -169,9 +169,12 @@ def send(
                 "summary": "concise",
             }
 
+        model_id = (model.get_ollama_model() or model.id or "").strip() if model.is_ollama() else (model.id or "")
+        if not model_id:
+            raise ValueError("Model name is required for the API request.")
         response = client.responses.create(
             input=messages,
-            model=model.id,
+            model=model_id,
             stream=stream,
             **response_kwargs,
         )

src/pygpt_net/provider/llms/ollama.py

@@ -92,9 +92,24 @@ def llama(
             if "base_url" not in args:
                 args["base_url"] = os.environ['OLLAMA_API_BASE']
         if "model" not in args:
-            args["model"] = model.id
+            args["model"] = (model.id or "").strip()
+        else:
+            args["model"] = (args["model"] or "").strip()
         if "request_timeout" not in args:
             args["request_timeout"] = 300
+        # Apply context and output token settings from model/config
+        if window:
+            ctx_size = window.core.models.get_num_ctx(model.id) if model.id else 0
+            if ctx_size <= 0:
+                ctx_size = window.core.config.get("max_total_tokens") or 0
+            if ctx_size > 0:
+                args["context_window"] = int(ctx_size)
+            max_out = window.core.models.get_tokens(model.id) if model.id else 0
+            if max_out <= 0:
+                max_out = window.core.config.get("max_output_tokens") or 0
+            if max_out > 0:
+                extra_opts = args.get("additional_kwargs") or {}
+                args["additional_kwargs"] = {**extra_opts, "num_predict": int(max_out)}
         return Ollama(**args)
 
     def get_embeddings_model(