W3: adicionada ClientWrapperFactory ao lado da classe AI, correcao em model_name.setter"

.gitignore

@@ -159,4 +159,4 @@ cython_debug/
 #  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
 #  and can be added to the global gitignore or merged into this file.  For a more nuclear
 #  option (not recommended) you can uncomment the following to ignore the entire idea folder.
-#.idea/
+.idea/

README.md

@@ -2,7 +2,7 @@
 
 Inspired by [tinygrad](https://github.com/tinygrad/tinygrad) and [simpleaichat](https://github.com/minimaxir/simpleaichat/tree/main/simpleaichat), `tiny-ai-client` is the easiest way to use and switch LLMs with vision and tool usage support. It works because it is `tiny`, `simple` and most importantly `fun` to develop.
 
-I want to change LLMs with ease, while knowing what is happening under the hood. Langchain is cool, but became bloated, complicated there is just too much chaos going on. I want to keep it simple, easy to understand and easy to use. If you want to use a LLM and have an API key, you should not need to read a 1000 lines of code and write `response.choices[0].message.content` to get the response.
+I want to change LLMs with ease, while knowing what is happening under the hood. Langchain is cool, but became bloated, complicated there is just too much chaos going on. I want to keep it simple, easy to understand and easy to use. If you want to use a LLM and have an API key, you should not need to read 1000 lines of code and write `response.choices[0].message.content` to get the response.
 
 Simple and tiny, that's the goal.
 

examples/anthropic_.py

@@ -62,6 +62,6 @@ def main():
 
 
 if __name__ == "__main__":
-    os.environ["ANTHROPIC_API_KEY"] = None
+    os.environ["ANTHROPIC_API_KEY"] = ""
     main()
     asyncio.run(async_ai_main())

examples/gemini_.py

@@ -58,6 +58,6 @@ async def async_ai_main():
 
 
 if __name__ == "__main__":
-    os.environ["GOOGLE_API_KEY"] = None
+    os.environ["GOOGLE_API_KEY"] = ""
     main()
     asyncio.run(async_ai_main())

examples/openai_.py

@@ -66,6 +66,6 @@ def main():
 
 
 if __name__ == "__main__":
-    os.environ["OPENAI_API_KEY"] = None
+    os.environ["OPENAI_API_KEY"] = ""
     main()
     asyncio.run(async_ai_main())

requirements.txt

@@ -1,3 +1,5 @@
 pydantic==2.7.3
 openai==1.31.0
 anthropic==0.28.0
+pillow~=10.3.0
+google-generativeai

tiny_ai_client/anthropic_.py

@@ -101,7 +101,7 @@ def call_llm_provider(
         temperature: int | None,
         max_new_tokens: int | None,
         timeout: int,
-    ) -> str:
+    ) -> Message:
         kwargs = {}
         input_messages, system = model_input
         if temperature is not None:
@@ -139,7 +139,7 @@ async def async_call_llm_provider(
         temperature: int | None,
         max_new_tokens: int | None,
         timeout: int,
-    ) -> str:
+    ) -> Message:
         kwargs = {}
         input_messages, system = model_input
         if temperature is not None:

tiny_ai_client/gemini_.py

@@ -31,7 +31,6 @@ def build_model_input(self, messages: List["Message"]) -> Any:
         history = []
         local_messages = deepcopy(messages)
         system = None
-        message = None
 
         for message in local_messages:
             if message.role == "system":
@@ -45,6 +44,7 @@ def build_model_input(self, messages: List["Message"]) -> Any:
                 if message.text is not None:
                     parts.append(message.text)
                 if message.images is not None:
+                    # noinspection PyTypeChecker
                     parts.extend(message.images)
                 history.append(
                     {
@@ -53,15 +53,15 @@ def build_model_input(self, messages: List["Message"]) -> Any:
                     }
                 )
 
-        return (system, history)
+        return system, history
 
     def call_llm_provider(
         self,
         model_input: Any,
         temperature: int | None,
         max_new_tokens: int | None,
         timeout: int,
-    ) -> str:
+    ) -> Message:
         system, history = model_input
 
         generation_config_kwargs = {}
@@ -92,7 +92,7 @@ async def async_call_llm_provider(
         temperature: int | None,
         max_new_tokens: int | None,
         timeout: int,
-    ) -> str:
+    ) -> Message:
         system, history = model_input
 
         generation_config_kwargs = {}

tiny_ai_client/models.py

@@ -5,17 +5,41 @@
 
 from tiny_ai_client.tools import json_to_function_input
 
+"""
+Using the factory design pattern approach
+provides a way to encapsulate the instantiate-and-return 
+The logic is code easy to code and test, much cleaner
+Following the best object-oriented design practices (DRI, SRP, OCP, Factory).
+"""
+
+
+class ClientWrapperFactory:
+    _client_wrapper_mapping = {
+        "gpt": "tiny_ai_client.openai_.OpenAIClientWrapper",
+        "claude": "tiny_ai_client.anthropic_.AnthropicClientWrapper",
+        "gemini": "tiny_ai_client.gemini_.GeminiClientWrapper"
+    }
+
+    @staticmethod
+    def create(model_name: str, tools: List[Union[Callable, Dict]]) -> "LLMClientWrapper":
+        for key, wrapper_path in ClientWrapperFactory._client_wrapper_mapping.items():
+            if key in model_name:
+                module_path, wrapper_class = wrapper_path.rsplit('.', 1)
+                module = __import__(module_path, fromlist=[wrapper_class])
+                return getattr(module, wrapper_class)(model_name, tools)
+        raise NotImplementedError(f"{model_name=} not supported")
+
 
 class AI:
     def __init__(
-        self,
-        model_name: str,
-        system: str | None = None,
-        temperature: int = 1,
-        max_new_tokens: int | None = None,
-        timeout: int = 30,
-        tools: List[Union[Callable, Dict]] | None = None,
-        chat: List["Message"] | None = None,
+            self,
+            model_name: str,
+            system: str | None = None,
+            temperature: int = 1,
+            max_new_tokens: int | None = None,
+            timeout: int = 30,
+            tools: List[Union[Callable, Dict]] | None = None,
+            chat: List["Message"] | None = None,
     ):
         # llm sampling parameters
         self.temperature: int = temperature
@@ -36,12 +60,12 @@ def __init__(
         )
 
     def __call__(
-        self,
-        message: str | None = None,
-        max_new_tokens: int | None = None,
-        temperature: int | None = 1,
-        timeout: int | None = None,
-        images: List[PIL_Image.Image] | PIL_Image.Image | None = None,
+            self,
+            message: str | None = None,
+            max_new_tokens: int | None = None,
+            temperature: int | None = 1,
+            timeout: int | None = None,
+            images: List[PIL_Image.Image] | PIL_Image.Image | None = None,
     ) -> str:
         max_new_tokens = max_new_tokens or self.max_new_tokens
         temperature = temperature or self.temperature
@@ -66,42 +90,30 @@ def __call__(
             response_msg.tool_call.result if response_msg.tool_call else ""
         )
 
+    # noinspection PyMethodMayBeStatic
     def get_llm_client_wrapper(
-        self, model_name: str, tools: List[Union[Callable, Dict]]
+            self, model_name: str, tools: List[Union[Callable, Dict]]
     ) -> "LLMClientWrapper":
-        if "gpt" in model_name:
-            from tiny_ai_client.openai_ import OpenAIClientWrapper
-
-            return OpenAIClientWrapper(model_name, tools)
-        if "claude" in model_name:
-            from tiny_ai_client.anthropic_ import AnthropicClientWrapper
-
-            return AnthropicClientWrapper(model_name, tools)
-
-        if "gemini" in model_name:
-            from tiny_ai_client.gemini_ import GeminiClientWrapper
-
-            return GeminiClientWrapper(model_name, tools)
-
-        raise NotImplementedError(f"{model_name=} not supported")
+        return ClientWrapperFactory.create(model_name, tools)
 
     @property
     def model_name(self) -> str:
-        return self._model_name
+        return self.model_name
 
     @model_name.setter
     def model_name(self, value: str) -> None:
+        self.model_name = value
         self.client_wrapper = self.get_llm_client_wrapper(value, self.tools)
 
 
 class AsyncAI(AI):
     async def __call__(
-        self,
-        message: str | None = None,
-        max_new_tokens: int | None = None,
-        temperature: int | None = 1,
-        timeout: int | None = None,
-        images: List[PIL_Image.Image] | PIL_Image.Image | None = None,
+            self,
+            message: str | None = None,
+            max_new_tokens: int | None = None,
+            temperature: int | None = 1,
+            timeout: int | None = None,
+            images: List[PIL_Image.Image] | PIL_Image.Image | None = None,
     ) -> str:
         max_new_tokens = max_new_tokens or self.max_new_tokens
         temperature = temperature or self.temperature
@@ -148,25 +160,25 @@ class Config:
 
 
 class LLMClientWrapper:
-    def build_model_input(self, messages: List["Message"]) -> Any:
+    def build_model_input(self, messages: List[Message]) -> Any:
         raise NotImplementedError
 
     def call_llm_provider(
-        self,
-        model_input: Any,
-        temperature: int | None,
-        max_new_tokens: int | None,
-        timeout: int,
-    ) -> "Message":
+            self,
+            model_input: Any,
+            temperature: int | None,
+            max_new_tokens: int | None,
+            timeout: int,
+    ) -> Message:
         raise NotImplementedError
 
     def __call__(
-        self,
-        max_new_tokens: int | None,
-        temperature: int | None,
-        timeout: int,
-        chat: List["Message"],
-    ) -> "Message":
+            self,
+            max_new_tokens: int | None,
+            temperature: int | None,
+            timeout: int,
+            chat: List[Message],
+    ) -> Message:
         model_input = self.build_model_input(chat)
         return self.call_llm_provider(
             model_input,
@@ -176,21 +188,21 @@ def __call__(
         )
 
     async def async_call_llm_provider(
-        self,
-        model_input: Any,
-        temperature: int | None,
-        max_new_tokens: int | None,
-        timeout: int,
-    ) -> "Message":
+            self,
+            model_input: Any,
+            temperature: int | None,
+            max_new_tokens: int | None,
+            timeout: int,
+    ) -> Message:
         raise NotImplementedError
 
     async def acall(
-        self,
-        max_new_tokens: int | None,
-        temperature: int | None,
-        timeout: int,
-        chat: List["Message"],
-    ) -> "Message":
+            self,
+            max_new_tokens: int | None,
+            temperature: int | None,
+            timeout: int,
+            chat: List[Message],
+    ) -> Message:
         model_input = self.build_model_input(chat)
         return await self.async_call_llm_provider(
             model_input,

tiny_ai_client/openai_.py

@@ -68,7 +68,7 @@ def call_llm_provider(
         temperature: int | None,
         max_new_tokens: int | None,
         timeout: int,
-    ) -> str:
+    ) -> Message:
         kwargs = {}
         if temperature is not None:
             kwargs["temperature"] = temperature
@@ -103,7 +103,7 @@ async def async_call_llm_provider(
         temperature: int | None,
         max_new_tokens: int | None,
         timeout: int,
-    ) -> str:
+    ) -> Message:
         kwargs = {}
         if temperature is not None:
             kwargs["temperature"] = temperature