fixing mkdown copy bug and pyomlx client

Author: kspviswa

main.py

@@ -137,7 +137,7 @@ def updateChat(message: Message, controlHandle: ft.Column, ai_response: bool = F
         #print(f'message {message} ai_response {ai_response} controlHandle {controlHandle}')
         if ai_response:
             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'),selectable=True, on_tap_link=open_url, auto_follow_links=True)
+            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
             controlHandle.controls.append(

mlxClient.py

@@ -1,4 +1,5 @@
 from openai import OpenAI
+from typing import List
 
 class MlxClient():
 
@@ -13,6 +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)
         message = {}
         message['role'] = 'user'
         message['content'] = prompt
@@ -21,14 +23,22 @@ def chat(self, prompt:str, model: str, temp: float, system:str = 'default') -> s
         try:
             #response = requests.post(self.llmHost, data=json.dumps(data), headers={'Content-Type': 'application/json'})
             #print(f'response code {response.status_code}')
-            response = self.client.chat.completions.create(model=f"mlx-community/{model}", 
+            response = self.client.chat.completions.create(model=model, 
                                           messages=self.messages)
             response = response.choices[0].message.content
+            print(response)
         except Exception as e:
             raise ValueError(e)
         ai_message = dict({'role' : 'assistant', 'content' : response})
         self.messages.append(ai_message)
         return response
 
     def chat_stream(self, prompt:str, model: str, temp: float) -> str:
-        pass
+        pass
+
+    def list(self) -> List[str]:
+        response = self.client.models.list()
+        models = []
+        for m in response:
+            models.append(m.id)
+        return models

models.py

@@ -3,7 +3,8 @@
 from typing_extensions import List
 import flet as ft
 import re
-from ollama import list
+from ollamaOpenAIClient import OllamaOpenAPIClient
+from mlxClient import MlxClient
 
 DEFAULT_OLLAMA_MODEL_REGISTRY = Path("~/.ollama/models/manifests/registry.ollama.ai/library").expanduser()
 DEFAULT_HF_MLX_MODEL_REGISTRY = Path("~/.cache/huggingface/hub/").expanduser()
@@ -22,23 +23,12 @@
 
 def returnModels() -> List[str]:
     print('entering retModels')
-    response = list()
-    raw_models = response['models']
-    models = []
-    for model in raw_models:
-        models.append(model['name'])
+    models = OllamaOpenAPIClient().list()
     return models
 
 def returnMlxModels() -> List[str]:
-    print(DEFAULT_HF_MLX_MODEL_REGISTRY)
-    models = []
-    for d in os.listdir(DEFAULT_HF_MLX_MODEL_REGISTRY):
-        a_dir = os.path.join(DEFAULT_HF_MLX_MODEL_REGISTRY, d)
-        if os.path.isdir(a_dir) and not d.startswith('.'):
-            #print(f'a_dir is {a_dir}')
-            model = re.split('--', d)[-1]  
-            models.append(model)
-    #print(models)
+    print('entering retModels')
+    models = MlxClient().list()
     return models
 
 def retModelOptions(isMlx=False):

ollamaOpenAIClient.py

@@ -0,0 +1,43 @@
+from openai import OpenAI
+from typing import List
+
+class OllamaOpenAPIClient():
+
+    def __init__(self):
+        self.messages = []
+        self.client = OpenAI(base_url='http://127.0.0.1:11434/v1', api_key='pyomlx')
+    
+    def clear_history(self):
+        self.messages.clear()
+    
+    def append_history(self, message):
+        self.messages.append(message)
+    
+    def chat(self, prompt:str, model: str, temp: float, system:str = 'default') -> str:
+        message = {}
+        message['role'] = 'user'
+        message['content'] = prompt
+        self.messages.append(message)
+        try:
+            response = self.client.chat.completions.create(model=model, 
+                                          messages=self.messages)
+            response = response.choices[0].message.content
+        except Exception as e:
+            raise ValueError(e)
+        ai_message = dict({'role' : 'assistant', 'content' : response})
+        self.messages.append(ai_message)
+        return response
+
+    def chat_stream(self, prompt:str, model: str, temp: float) -> str:
+        pass
+
+    def list(self) -> List[str]:
+        response = self.client.models.list()
+        models = []
+        for m in response:
+            models.append(m.id)
+        return models
+
+
+if __name__ == '__main__':
+    print(OllamaOpenAPIClient().list())

prompt.py

@@ -1,12 +1,10 @@
-from mlxLLM import MlxLLM
 from search import *
-#from mlxLLM_local import MlxLLM_local
 import requests
-from ollamaClient import OllamaClient
 from mlxClient import MlxClient
+from ollamaOpenAIClient import OllamaOpenAPIClient
 
-oChatClient = OllamaClient()
-oSearchClient = OllamaClient()
+oChatClient = OllamaOpenAPIClient()
+oSearchClient = OllamaOpenAPIClient()
 
 mChatClient = MlxClient()
 mSearchClient = MlxClient()
@@ -70,39 +68,4 @@ def firePrompt(prompt: str, model: str="dolphin-mistral:latest", temp=0.4, isMlx
         except Exception as e:
             return f'Generic Error Occured {e}', ""
 
-    return response, keywords
-
-
-
-
-
-# def firePrompt1(prompt: str, model: str="dolphin-mistral:latest", temp=0.4, isMlx=False, chat_mode=False) -> str:
-#     res = ""
-#     keywords = ""
-#     if chat_mode:
-#         if isMlx:
-#             llm = MlxLLM(model=model, temp=temp)
-#             #llm = OllamaClient()
-#             #llm = MlxLLM_local(model=model, temp=temp)
-#         else:
-#             #llm = Ollama(model=model,
-#             #    temperature=temp
-#             #    )
-#             llm = OllamaClient()
-#         try:
-#             #res = llm.invoke(prompt)
-#             res = llm.chat(model=model, temp=temp, prompt=prompt)
-#         except requests.exceptions.ConnectionError as e:
-#             err_str = f'PyOMlX' if isMlx else f'Ollama'
-#             return f'Unable to connect to {err_str}. Is it running?🤔', ""
-#         except Exception as e:
-#             return f'Generic Error Occured {e}', ""
-#     else:
-#         try:
-#             if isMlx:
-#                 res = f'Search not implemented for MlX models. Stay tuned!'
-#             else:
-#                 res, keywords = retSearchResults(model=model, search_str=prompt, temp=temp)
-#         except Exception as e:
-#             return f'__Generic Error Occured__ {e}', ""
-#     return res, keywords
+    return response, keywords

requirements.txt

@@ -2,6 +2,7 @@ duckduckgo_search==6.1.7
 langchain==0.1.4
 langchain-community==0.0.16
 langchain-core==0.1.17
-typing_extensions==4.9.0
-flet==0.23.2
+typing_extensions==4.12.2
+flet==0.25.2
+flet-core==0.24.1
 openai==1.58.1

settings.py

@@ -1,6 +1,8 @@
 import flet as ft
 from models import *
 
+PYOLLAMX_VERSION = '0.0.5'
+
 def clearState(e: ft.ControlEvent) -> None:
     model_dropdown.value = "N/A"
     e.page.session.set('selected_model', 'N/A')
@@ -73,10 +75,12 @@ def updateModelInfo(e: ft.ControlEvent) -> None:
                       height=75,
                       fit=ft.ImageFit.CONTAIN,
                       )
+settings_pyollmx_version_text = ft.Text(value=f'v{PYOLLAMX_VERSION}', style=ft.TextStyle(font_family='CabinSketch-Bold'), size=10)
 
 settings_banner_view = ft.Row([
     settings_banner_image,
-    settings_banner_text
+    settings_banner_text,
+    settings_pyollmx_version_text
 ], alignment=ft.MainAxisAlignment.CENTER, vertical_alignment=ft.CrossAxisAlignment.CENTER)