bootstrapping v0.0.7

Author: kspviswa

README.md

@@ -22,6 +22,10 @@ MacOS DMGs are available in [Releases](https://github.com/kspviswa/pyOllaMx/rele
 
 [PyOMlx](https://github.com/kspviswa/PyOMlx) : A Macos App capable of discovering, loading & serving Apple MlX models downloaded from [Apple MLX Community repo in hugging face](https://huggingface.co/mlx-community) 🤗
 
+## Star History
+
+[![Star History Chart](https://api.star-history.com/svg?repos=kspviswa/pyOllaMx&type=Date)](https://star-history.com/#kspviswa/pyOllaMx&Date)
+
 ## How to use?
 
 1) Install [Ollama Application](https://ollama.ai/download) & use Ollama CLI to download your desired models

buildflet.sh

@@ -1,2 +1,2 @@
 #!/bin/sh
-flet build macos --build-version '0.0.6' --copyright 'Viswa Kumar 2025 ©' --project 'PyOllaMx' --company 'PyOllaMx'
+flet build macos --build-version '0.0.7' --copyright 'Viswa Kumar 2025 ©' --project 'PyOllaMx' --company 'PyOllaMx'

main.py

@@ -136,10 +136,35 @@ def open_url(e):
     def updateChat(message: Message, controlHandle: ft.Column, ai_response: bool = False):
         #print(f'message {message} ai_response {ai_response} controlHandle {controlHandle}')
         if ai_response:
+            ai_thinking_text = re.sub(r"</think>.*$", "", message.text, flags=re.DOTALL) if "</think>" in message.text else ""
+            ai_non_thinking_text = message.text
+
+            #print(f'### {ai_thinking_text}')
             ai_message_container = ft.Container(width=550)
             ai_message_md = ft.Markdown(value="", extension_set="gitHubWeb", code_theme='obsidian', code_style=ft.TextStyle(font_family='Roboto Mono'), on_tap_link=open_url, auto_follow_links=True)
             ai_message_md_selectable = ft.SelectionArea(content=ai_message_md)
-            ai_message_container.content = ai_message_md_selectable
+            ai_message_thinking_md = ft.ExpansionTile(
+            title=ft.Text("Thinking tokens 🤔", font_family="RobotoSlab"),
+            subtitle=ft.Text("Expand to reveal the model's thinking tokens", theme_style=ft.TextThemeStyle.BODY_SMALL, font_family="RobotoSlab"),
+            affinity=ft.TileAffinity.LEADING,
+            initially_expanded=False,
+            collapsed_text_color=ft.Colors.BLUE,
+            text_color=ft.Colors.BLUE,
+            controls=[
+                ft.ListTile(title=ft.Text(
+                    theme_style=ft.TextThemeStyle.BODY_SMALL,
+                    font_family="RobotoSlab",
+                )),
+            ],
+        )
+            if ai_thinking_text :
+                ai_message_container.content = ft.Column([ai_message_thinking_md,
+                                                      ai_message_md_selectable,
+                                                      ])
+                ai_non_thinking_text = ''.join(message.text.split("</think>")[1:])
+            else:
+                ai_message_container.content = ai_message_md_selectable
+
             controlHandle.controls.append(
                 ft.Row([
                     ft.Image(src=getAILogo(page.session.get('isMlx')),
@@ -151,16 +176,17 @@ def updateChat(message: Message, controlHandle: ft.Column, ai_response: bool = F
                 width=500, vertical_alignment=ft.CrossAxisAlignment.START,
                 )
             )
+            ai_message_thinking_md.controls[0].title.value = ai_thinking_text
             if enable_streaming.value:
                 full_r = ""
-                for chunk in message.text.split(sep=" "):
+                for chunk in ai_non_thinking_text.split(sep=" "):
                     full_r += chunk + " "
                     ai_message_md.value = full_r
                     # controlHandle.scroll_to(offset=-1, duration=100, curve=ft.AnimationCurve.EASE_IN_OUT)
                     page.update()
                     time.sleep(0.05)
             else:
-                ai_message_md.value = message.text
+                ai_message_md.value = ai_non_thinking_text
         else:
             controlHandle.controls.append(
                 ft.Row([

mlxClient.py

@@ -14,7 +14,7 @@ def append_history(self, message):
         self.messages.append(message)
 
     def chat(self, prompt:str, model: str, temp: float, system:str = 'default') -> str:
-        print('Entering Chat for Mlx' + model)
+        #print('Entering Chat for Mlx' + model)
         message = {}
         message['role'] = 'user'
         message['content'] = prompt
@@ -26,7 +26,7 @@ def chat(self, prompt:str, model: str, temp: float, system:str = 'default') -> s
             response = self.client.chat.completions.create(model=model, 
                                           messages=self.messages)
             response = response.choices[0].message.content
-            print(response)
+            # print(response)
         except Exception as e:
             raise ValueError(e)
         ai_message = dict({'role' : 'assistant', 'content' : response})

model_hub.py

@@ -171,8 +171,9 @@ def download_from_ollama(e: ft.ControlEvent):
     ollama_download_pbar,
     restart_required_container,
     ft.Column([
+        ft.Text(),
         ollama_models_table
-    ], scroll=ft.ScrollMode.ADAPTIVE, height=350),
+    ], scroll=ft.ScrollMode.ADAPTIVE, height=300),
 ],alignment=ft.MainAxisAlignment.START, horizontal_alignment=ft.CrossAxisAlignment.STRETCH, spacing=10)
 
 coming_soon_view = ft.Column([

pyomlx_test.py

@@ -0,0 +1,14 @@
+from mlx_lm import load, generate
+
+model, tokenizer = load("mlx-community/Phi-4-mini-instruct-8bit")
+
+prompt="Hi how are you?"
+
+if hasattr(tokenizer, "apply_chat_template") and tokenizer.chat_template is not None:
+    messages = [{"role": "user", "content": prompt}]
+    prompt = tokenizer.apply_chat_template(
+        messages, tokenize=False, add_generation_prompt=True
+    )
+
+response = generate(model, tokenizer, prompt=prompt, verbose=True)
+print(response)