Enhance project structure by adding new files and updating dependencies. Introduce model_factory.py for streamlined model creation using environment variables. Update .gitignore to exclude additional cache and model files. Modify main_v2.py to utilize ModelFactory for model instantiation and adjust API key loading. Expand requirements.txt with new dependencies for improved functionality. Add new notebooks for fine-tuning and model training, enhancing project documentation and usability.

.gitignore

@@ -1,4 +1,8 @@
 .env
+lora_model
+outputs
 __pycache__
 .pytest_cache
+unsloth_compiled_cache
 .venv
+wandb

main_v2.py

@@ -13,6 +13,7 @@ from smolagents import CodeAgent, LiteLLMModel
 from smolagents.monitoring import LogLevel
 
 from tools.smart_search.tool import SmartSearchTool
+from model_factory import ModelFactory
 
 _disable_debugging()
 
@@ -25,17 +26,14 @@ logging.basicConfig(
 )
 logger = logging.getLogger(__name__)
 
-load_dotenv(find_dotenv())
+# load_dotenv(find_dotenv())
 
-API_BASE = os.getenv("API_BASE")
-API_KEY = os.getenv("API_KEY")
-MODEL_ID = os.getenv("MODEL_ID")
+# API_BASE = os.getenv("API_BASE")
+# API_KEY = os.getenv("API_KEY")
+# MODEL_ID = os.getenv("MODEL_ID")
 
-model = LiteLLMModel(
-    api_base=API_BASE,
-    api_key=API_KEY,
-    model_id=MODEL_ID,
-)
+# Create model using the factory
+model = ModelFactory.create_model()
 
 # data_agent = create_data_agent(model)
 # media_agent = create_media_agent(model)
@@ -140,7 +138,7 @@ if __name__ == "__main__":
     response.raise_for_status()
     questions_data = response.json()
 
-    for question_data in questions_data[:2]:
+    for question_data in questions_data[:1]:
         file_name = question_data["file_name"]
         level = question_data["Level"]
         question = question_data["question"]

model_factory.py

@@ -0,0 +1,37 @@
+import os
+from dotenv import find_dotenv, load_dotenv
+from smolagents import LiteLLMModel, TransformersModel
+
+class ModelFactory:
+    @staticmethod
+    def create_model():
+        """
+        Creates and returns a LiteLLMModel instance configured with environment variables.
+        
+        Returns:
+            LiteLLMModel: A configured instance of LiteLLMModel
+        """
+        # Load environment variables
+        load_dotenv(find_dotenv())
+        
+        # Get configuration from environment variables
+        api_base = os.getenv("API_BASE")
+        api_key = os.getenv("API_KEY")
+        model_id = os.getenv("MODEL_ID")
+        
+        # Create and return the model
+        # return LiteLLMModel(
+        #     api_base=api_base,
+        #     api_key=api_key,
+        #     model_id=model_id,
+        # )
+
+        return TransformersModel(
+            # max_new_tokens=5000,
+            max_new_tokens=256,
+            model_id="HuggingFaceTB/SmolLM2-135M-Instruct",
+            # model_id="HuggingFaceTB/SmolLM2-360M-Instruct",
+            # model_id="HuggingFaceTB/SmolLM2-1.7B-Instruct",
+            # model_id="HuggingFaceTB/SmolVLM2-256M-Video-Instruct",
+            # model_id="Qwen/Qwen2.5-Coder-32B-Instruct",
+        )

requirements.txt

@@ -1,7 +1,12 @@
+accelerate>=1.6.0
 beautifulsoup4>=4.13.4
+bitsandbytes>=0.45.5
 duckduckgo-search>=8.0.1
 gradio[oauth]>=5.26.0
+hf-xet>=1.0.5
+ipywidgets>=8.1.6
 isort>=6.0.1
+jupyter>=1.1.1
 kagglehub>=0.3.12
 langchain>=0.1.0
 langchain-community>=0.0.10
@@ -12,15 +17,20 @@ llama-index-embeddings-huggingface>=0.5.3
 llama-index-readers-wikipedia>=0.3.0
 markdown>=3.8
 mlcroissant>=1.0.17
+num2words>=0.5.14
 numpy>=2.2.5
 pandas>=2.0.0
+peft>=0.15.2
 pytest>=8.3.5
 pytest-cov>=6.1.1
 python-dotenv>=1.0.0
 requests>=2.32.3
 sentence-transformers>=4.1.0
 smolagents[litellm,telemetry]>=1.14.0
+tensorboardX>=2.6.2.2
+trl>=0.17.0
 typing-extensions>=4.5.0
-hf-xet>=1.0.5
+unsloth>=2025.4.3
+wandb>=0.19.10
 wikipedia>=1.4.0
 wikipedia-api>=0.8.1