Model Replacement

Replaced Mistral with Nano(gpt-4.1-nano) for suggestions and the Mistral Generate, uses the gpt-4.1-mini for generating the business plan.

app/main.py

@@ -172,12 +172,19 @@ def get_llm(model_name: str):
         logging.error("HUGGINGFACEHUB_API_TOKEN environment variable not set")
         raise ValueError("HUGGINGFACEHUB_API_TOKEN environment variable is required for Hugging Face models")
         
-    if model_name == "GPT":
+    if model_name == "Mistral": #Front end fix, Mistral Calls GPT
         openai_api_key = os.getenv("OPENAI_API_KEY")
         if not openai_api_key:
             logging.error("OPENAI_API_KEY environment variable not set")
             raise ValueError("OPENAI_API_KEY environment variable is required for GPT models")
-        return ChatOpenAI(model_name="gpt-4o-mini-2024-07-18", temperature=0.7, max_tokens=4000)
+        return ChatOpenAI(model_name="gpt-4.1-mini", temperature=0.7, max_tokens=6000)
+    # ──────────────────────────────────────────────────────────────────
+    elif model_name.lower() == "gpt-4.1-nano":   #𝙉𝙀𝙒: support GPT-4.1 Nano
+        openai_api_key = os.getenv("OPENAI_API_KEY")
+        if not openai_api_key:
+            raise ValueError("OPENAI_API_KEY is required for GPT models")
+        return ChatOpenAI(model_name="gpt-4.1-nano", temperature=0.7, max_tokens=4000)
+    # ──────────────────────────────────────────────────────────────────
     elif model_name == "Llama":
         logging.info("Initializing Llama model - note this is a gated model and requires special access")
         try:
@@ -213,33 +220,6 @@ def get_llm(model_name: str):
                 raise ValueError("The Qwen-2.5-7B model (15GB) exceeds the free tier limit (10GB) for Hugging Face Inference API. You need to upgrade to Hugging Face Pro to use this model, or use a smaller model like Mistral.")
             else:
                 raise ValueError(f"Failed to initialize Qwen model: {error_msg}")
-    elif model_name == "Mistral":
-        logging.info("Initializing Mistral model")
-        try:
-            return HFInferenceLLM(
-                model="mistralai/Mistral-7B-Instruct-v0.3",
-                provider="together",
-                api_key=hf_token,
-                max_tokens=4000
-            )
-        except Exception as e:
-            error_msg = str(e)
-            logging.error(f"Failed to initialize Mistral model: {error_msg}")
-            
-            if "too large to be loaded automatically" in error_msg:
-                # Try a smaller fallback model
-                logging.info("Trying smaller Mistral model...")
-                try:
-                    return HFInferenceLLM(
-                        model="mistralai/Mistral-7B-Instruct-v0.1",
-                        provider="hf-inference",
-                        api_key=hf_token,
-                        max_tokens=4000
-                    )
-                except Exception:
-                    pass
-            
-            raise ValueError(f"Failed to initialize Mistral model: {error_msg}")
     elif model_name == "Gemma":
         logging.info("Initializing Gemma model")
         try:
@@ -319,7 +299,8 @@ async def get_suggestions(
     prompt += "\nSuggestions:\n-"
 
     try:
-        llm = get_llm(model)  # your existing get_llm
+        #llm = get_llm(model)  # your existing get_llm which would now call gpt-4.1-mini
+        llm = get_llm("gpt-4.1-nano") #now we force it to use gpt-4.1-nano
         suggestion_prompt = PromptTemplate(
             input_variables=["prompt"],
             template=prompt
@@ -464,7 +445,7 @@ def health_check():
     for model_name in models_to_check:
         try:
             # Just initialize the model without running inference
-            if model_name == "Mistral":
+            if model_name == "":
                 client = InferenceClient(provider="hf-inference", api_key=hf_token)
                 # Just a simple check that the model exists
                 client.model_info("mistralai/Mistral-7B-Instruct-v0.3")