add custom handler

handler.py

@@ -1,25 +1,24 @@
-from typing import Dict, List, Any
-from transformers import AutoModelForCausalLM, AutoTokenizer
+from typing import Dict, Any
+from transformers import AutoTokenizer
 
 class EndpointHandler:
     def __init__(self, path="prashanthbsp/reasoning-cpg-entity-v1"):
-        # Standard HF model loading - compatible with TGI
+        # Only load the tokenizer - the model is loaded by TGI
         self.tokenizer = AutoTokenizer.from_pretrained(path)
-        # Model is loaded by the TGI server, not by the handler
 
     def __call__(self, data: Dict[str, Any]) -> Dict[str, Any]:
         """
         data args:
-            inputs: text or dict containing text
+            inputs: Text or dict containing text
         Return:
-            A dict with the model's response
+            Dict with prompt and generation parameters
         """
-        # Extract inputs
+        # Extract the input text
         inputs = data.pop("inputs", data)
         context = inputs.pop("context", inputs)
         
-        # Format prompt according to your requirements
-        prompt = f"""Below is an instruction that describes a task, paired with an input that provides further context.
+        # Format the prompt
+        prompt = """Below is an instruction that describes a task, paired with an input that provides further context.
             Write a response that appropriately completes the request.
             Before answering, think carefully about the task to ensure a logical and accurate response.
 
@@ -54,15 +53,16 @@ class EndpointHandler:
             }}
 
             ### Social Media Post:
-            {context}
+            {0}
             ### Response:
-            <think>"""
+            <think>""".format(context)
         
-        # For TGI, we return a dict with the prompt and generation params
+        # Return the formatted prompt and generation parameters for TGI
         return {
             "inputs": prompt,
             "parameters": {
                 "max_new_tokens": 1200,
+                "temperature": 0.01,  # Low temperature for more deterministic outputs
                 "do_sample": False,
                 "return_full_text": False  # Only return the generated text, not the prompt
             }