Update llm_func.py

llm_func.py

@@ -1,24 +1,154 @@
 import time
-from llama_cpp import Llama
-#model = Llama.from_pretrained(repo_id="tensorblock/SmolLM-135M-Instruct-GGUF",filename="*SmolLM-135M-Instruct-Q6_K.gguf",verbose=False,n_ctx=1000) 
-model = Llama.from_pretrained(repo_id="HuggingFaceTB/SmolLM2-360M-Instruct-GGUF",filename="*smollm2-360m-instruct-q8_0.gguf",verbose=False,n_ctx=500) 
-#model = Llama.from_pretrained(repo_id="tensorblock/SmolLM2-360M-GGUF",filename="*SmolLM2-360M-Q4_K_M.gguf",verbose=False,n_ctx=500)
-#model = Llama.from_pretrained(repo_id="bartowski/SmolLM2-135M-Instruct-GGUF",filename="*SmolLM2-135M-Instruct-Q4_K_M.gguf",verbose=False,n_ctx=500)
-#model = Llama.from_pretrained(repo_id="bartowski/granite-3.0-1b-a400m-instruct-GGUF",filename="*granite-3.0-1b-a400m-instruct-Q5_K_S.gguf",verbose=False,n_ctx=1000) 
-
-
-def mdl(input):
-    print(input)
-    start = time.time()
-    output = model(prompt=f"""<|start_of_role|>system<|end_of_role|>You are a helpful chatbot<|end_of_text|>
-    <|start_of_role|>user<|end_of_role|>{input}<|end_of_text|>
-    <|start_of_role|>assistant<|end_of_role|>""",
-    max_tokens=200,
-    temperature=0.1,
-    top_p=0.1,
-    echo=False,
-    stop=["<|"])
-    end = time.time()
-    total_time = end - start
-    print(f"Execution time: {total_time:.2f} seconds")
-    return output["choices"][0]["text"].replace('\nassistant\n','')  
+import json
+
+def llm_normal(question,llm):
+	print(question)
+	start = time.time()
+	answer = llm.create_chat_completion(
+	        messages= [
+    			{
+        		"role": "user",
+        		"content": question
+    			}
+			]		
+	    )
+	end = time.time()
+	total_time = end - start
+	print(answer["choices"][0]["message"]['content'])
+	print(f"Execution time: {total_time:.2f} seconds")
+	return answer["choices"][0]["message"]['content']
+
+
+def llm_agent(question,llm):
+	print(question)
+	start = time.time()
+	answer = llm.create_chat_completion(
+	        messages= [
+    			{
+        		"role": "system",
+        		"content": """
+                You are a json formatter assistant, 
+                specialized in converting user-supplied raw text into json format
+                """
+    		},{
+        		"role": "user",
+        		"content": question+"""
+				format in JSON: """
+    			}
+			],response_format = {"type":"json_object"}		
+	    )
+	end = time.time()
+	total_time = end - start
+	try:
+		args = json.loads(answer["choices"][0]["message"]['content'])
+		print(args)
+		print(f"Execution time: {total_time:.2f} seconds")
+	except:
+		print(answer["choices"][0]["message"]['content'])
+		print(f"Execution time: {total_time:.2f} seconds")
+	return answer["choices"][0]["message"]['content']
+
+
+def name_age(name,age):
+    return str(name) + ' - '+str(age)
+
+def llm_functioncalling(question,llm):
+	print(question)
+	start = time.time()
+	answer = llm.create_chat_completion(
+    messages = [
+            {
+            "role": "system",
+            "content": """
+                    You are a json formatter assistant, 
+                    specialized in converting user-supplied raw text into json format
+                    """
+            },{
+            "role": "user",
+            "content": question +"""
+            format in JSON: """
+            }
+			],
+	tools=[{
+		    "type": "function",
+		    "function": {
+		      "name": "name_age",
+		    "description": """return the name and age, if not informed, 
+		                      set name None and age 0""",
+		      "parameters": {
+		        "type": "object",
+		        "title": "return the name and age",
+		        "properties": {
+		          "name": {
+		            "title": "Name",
+		            "type": "string"
+		          },
+		          "age": {
+		            "title": "Age",
+		            "type": "integer"
+		          }
+		        },
+		        "required": [ "name", "age" ]
+		      }
+		    }
+		  }
+	],response_format = {"type":"json_object"},
+    )
+	end = time.time()
+	total_time = end - start
+	try:
+		args = json.loads(answer["choices"][0]["message"]['content'])
+		print(name_age(args['name'],args['age']))
+		print(f"Execution time: {total_time:.2f} seconds")
+		return name_age(args['name'],args['age'])
+	except:
+		print(answer["choices"][0]["message"]['content'])
+		print(f"Execution time: {total_time:.2f} seconds.")
+		return answer["choices"][0]["message"]['content']
+
+
+
+import requests 
+from bs4 import BeautifulSoup
+
+def search(question):
+	payload = {'q': question}
+	request = requests.get('https://www.bing.com/search?&cc=US', params=payload,
+		headers = {
+		'User-Agent':'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/110.0.0.0 Safari/537.36'
+		})
+	soup = BeautifulSoup(request.text, 'html.parser')
+	try:	 
+		text = 'context : '+ str(soup.text.split('·')[1].split('https://')[0])
+		return text
+	except:
+		return ' '
+
+
+
+
+def llm_search(question,llm):
+	print(question)
+	start = time.time()
+	messages = [
+	    {
+	        "role": "system",
+	        "content": """You are a search assistant, gives helpful, 
+            detailed, and polite answers to the user's questions
+	        """
+	    },
+	    {
+	        "role": "user",
+	        "content": question+"""
+	        """+search(question)
+	    },
+	]
+	answer = llm.create_chat_completion(
+		messages = messages
+	)
+	end = time.time()
+	total_time = end - start
+	print(answer["choices"][0]["message"]['content'])
+	print(f"Execution time: {total_time:.2f} seconds")
+	return answer["choices"][0]["message"]['content']
+