- Adding multiple reasoning models

Config_files/message_system_config.json

@@ -21,7 +21,10 @@
   "AI_phrases_singeo": ["your voice deserves more"],
   "header_limit": 30,
   "message_limit": 110,
-  "LLM_models": ["4o-mini", "gpt-4o", "gpt-4.1-nano", "gpt-4.1-mini", "gpt-3.5-turbo", "o4-mini", "o1"]
+  "LLM_models": ["gpt-4o-mini", "gpt-4o", "gpt-4.1-mini", "gpt-3.5-turbo", "o1", "o4-mini", "o1-mini", "o3-mini"],
+  "openai_models": ["gpt-4o-mini", "gpt-4o", "gpt-4.1-nano", "gpt-3.5-turbo", "gpt-4.1-mini"],
+  "reasoning": ["o1", "o4-mini", "o1-mini", "o3-mini"],
+  "ollama_models": ["deepseek-r1:1.5b", "gemma3:4b", "deepseek-r1:7b", "gemma3:4b"]
 }
 
 

Messaging_system/CoreConfig.py

@@ -23,6 +23,7 @@ class CoreConfig:
         self.api_key = None  # will be set by user
         self.model = "gpt-4o" # default -> will be set by user
         self.temperature = 0.7
+        self.reasoning_model=False
 
         # will be set by user
         self.CTA = None
@@ -84,6 +85,8 @@ class CoreConfig:
 
     def set_llm_model(self, model):
         self.model = model
+        if self.model in self.config_file["reasoning"]:
+            self.reasoning_model = True
 
     # --------------------------------------------------------------
     # --------------------------------------------------------------

Messaging_system/LLM.py

@@ -15,13 +15,10 @@ import re
 class LLM:
     def __init__(self, Core):
         self.Core = Core
-
-
         self.model = None
         self.model_type = "openai" # valid values -> ["openai", "ollama"]
         self.client = None
         self.connect_to_llm()
-        self.reasoning = {}
 
 
     def get_response(self, prompt, instructions):
@@ -39,98 +36,19 @@ class LLM:
         connect to selected llm -> ollama or openai connection
         :return:
         """
-        openai_models = ["4o-mini", "gpt-4o", "gpt-4.1-nano", "gpt-3.5-turbo", "gpt-4.1-mini"]
-        reasoning = ["o1", "o4-mini"]
-        ollama_models = ["deepseek-r1:1.5b", "gemma3:4b", "deepseek-r1:7b", "gemma3:4b"]
 
-        if self.Core.model in openai_models:
+        if self.Core.model in self.Core.config_file["openai_models"]:
             self.model_type = "openai"
-            if self.Core.model in reasoning:
-                self.reasoning= {"effort": "medium"}
 
-        if self.Core.model in ollama_models:
+        if self.Core.model in self.Core.config_file["ollama_models"]:
             self.model_type = "ollama"
             self.client = ollama.Client()
 
         self.model = self.Core.model
 
 
-    def get_message_openai(self, prompt, instructions, max_retries=4):
-
-
-        openai.api_key = self.Core.api_key
-        client = OpenAI(api_key=self.Core.api_key)
-
-        for attempt in range(max_retries):
-            try:
-                response = client.responses.create(
-                    model=self.Core.model,
-                    input=[{"role": "system", "content": instructions},
-                            {"role": "user", "content": prompt}],
-                    text={
-                        "format": {
-                            "type": "json_object"
-                        }
-                    },
-                    reasoning=self.reasoning,
-                    max_output_tokens=500,
-                    temperature=self.Core.temperature,
-                    top_p=1,
-                    tools=[],
-                )
-
-                tokens = {
-                    'prompt_tokens': response.usage.prompt_tokens,
-                    'completion_tokens': response.usage.completion_tokens,
-                    'total_tokens': response.usage.total_tokens
-                }
-
-                try:
-                    content = response.choices[0].message.content
-
-                    # Extract JSON code block
-
-                    output = json.loads(content)
-
-                    if 'message' not in output or 'header' not in output:
-                        print(f"'message' or 'header' is missing in response on attempt {attempt + 1}. Retrying...")
-                        continue  # Continue to next attempt
-
-                    else:
-                        if len(output["header"].strip()) > self.Core.config_file["header_limit"] or len(
-                                output["message"].strip()) > self.Core.config_file["message_limit"]:
-                            print(
-                                f"'header' or 'message' is more than specified characters in response on attempt {attempt + 1}. Retrying...")
-                            continue
-
-                    # validating the JSON
-                    self.Core.total_tokens['prompt_tokens'] += tokens['prompt_tokens']
-                    self.Core.total_tokens['completion_tokens'] += tokens['completion_tokens']
-                    self.Core.temp_token_counter += tokens['total_tokens']
-                    return output
-
-                except json.JSONDecodeError:
-                    print(f"Invalid JSON from LLM on attempt {attempt + 1}. Retrying...")
-
-            except openai.APIConnectionError as e:
-                print("The server could not be reached")
-                print(e.__cause__)  # an underlying Exception, likely raised within httpx.
-            except openai.RateLimitError as e:
-                print("A 429 status code was received; we should back off a bit.")
-            except openai.APIStatusError as e:
-                print("Another non-200-range status code was received")
-                print(e.status_code)
-                print(e.response)
-
-        print("Max retries exceeded. Returning empty response.")
-        return None
-
-
-
     # def get_message_openai(self, prompt, instructions, max_retries=4):
-    #     """
-    #     sending the prompt to openai LLM and get back the response
-    #     """
+    #
     #
     #     openai.api_key = self.Core.api_key
     #     client = OpenAI(api_key=self.Core.api_key)
@@ -139,14 +57,12 @@ class LLM:
     #         try:
     #             response = client.chat.completions.create(
     #                 model=self.Core.model,
+    #                 messages= [{"role": "system", "content": instructions},{"role": "user", "content": prompt}],
     #                 response_format={"type": "json_object"},
-    #                 messages=[
-    #                     {"role": "system", "content": instructions},
-    #                     {"role": "user", "content": prompt}
-    #                 ],
-    #                 max_tokens=500,
-    #                 n=1,
+    #                 reasoning_effort=self.reasoning,
+    #                 max_tokens=1000,
     #                 temperature=self.Core.temperature,
+    #                 tools=[]
     #             )
     #
     #             tokens = {
@@ -195,6 +111,88 @@ class LLM:
     #     print("Max retries exceeded. Returning empty response.")
     #     return None
 
+
+
+    def get_message_openai(self, prompt, instructions, max_retries=4):
+        """
+        sending the prompt to openai LLM and get back the response
+        """
+
+        openai.api_key = self.Core.api_key
+        client = OpenAI(api_key=self.Core.api_key)
+
+        for attempt in range(max_retries):
+            try:
+                if self.Core.reasoning_model:
+                    response = client.chat.completions.create(
+                        model=self.Core.model,
+                        response_format={"type": "json_object"},
+                        messages=[
+                            {"role": "system", "content": instructions},
+                            {"role": "user", "content": prompt}
+                        ],
+                        reasoning_effort="medium",
+                        n=1,
+                    )
+
+                else:
+                    response = client.chat.completions.create(
+                        model=self.Core.model,
+                        response_format={"type": "json_object"},
+                        messages=[
+                            {"role": "system", "content": instructions},
+                            {"role": "user", "content": prompt}
+                        ],
+                        n=1,
+                        temperature=self.Core.temperature
+                    )
+
+                tokens = {
+                    'prompt_tokens': response.usage.prompt_tokens,
+                    'completion_tokens': response.usage.completion_tokens,
+                    'total_tokens': response.usage.total_tokens
+                }
+
+                try:
+                    content = response.choices[0].message.content
+
+                    # Extract JSON code block
+
+                    output = json.loads(content)
+
+                    if 'message' not in output or 'header' not in output:
+                        print(f"'message' or 'header' is missing in response on attempt {attempt + 1}. Retrying...")
+                        continue  # Continue to next attempt
+
+                    else:
+                        if len(output["header"].strip()) > self.Core.config_file["header_limit"] or len(
+                                output["message"].strip()) > self.Core.config_file["message_limit"]:
+                            print(
+                                f"'header' or 'message' is more than specified characters in response on attempt {attempt + 1}. Retrying...")
+                            continue
+
+                    # validating the JSON
+                    self.Core.total_tokens['prompt_tokens'] += tokens['prompt_tokens']
+                    self.Core.total_tokens['completion_tokens'] += tokens['completion_tokens']
+                    self.Core.temp_token_counter += tokens['total_tokens']
+                    return output
+
+                except json.JSONDecodeError:
+                    print(f"Invalid JSON from LLM on attempt {attempt + 1}. Retrying...")
+
+            except openai.APIConnectionError as e:
+                print("The server could not be reached")
+                print(e.__cause__)  # an underlying Exception, likely raised within httpx.
+            except openai.RateLimitError as e:
+                print("A 429 status code was received; we should back off a bit.")
+            except openai.APIStatusError as e:
+                print("Another non-200-range status code was received")
+                print(e.status_code)
+                print(e.response)
+
+        print("Max retries exceeded. Returning empty response.")
+        return None
+
     # ======================================================================
 
     def get_message_ollama(self, prompt, instructions, max_retries=10):
@@ -255,24 +253,6 @@ class LLM:
 
     # ======================================================================
 
-    # def preprocess_and_parse_json(self, response):
-    #     # Remove any leading/trailing whitespace and newlines
-    #     if response.startswith('```json') and response.endswith('```'):
-    #         response = response[len('```json'):-len('```')].strip()
-    #
-    #     # Parse the cleaned response into a JSON object
-    #     try:
-    #         json_object = json.loads(response)
-    #         return json_object
-    #     except json.JSONDecodeError as e:
-    #         print(f"Failed to parse JSON: {e}")
-    #         return None
-
-    # =====================================================================
-
-    import re
-    import json
-
     def preprocess_and_parse_json(self, response: str):
         """
         Cleans an LLM response by removing <think> tags and extracting JSON

Messaging_system/PromptEng.py

@@ -2,6 +2,7 @@
 This is the prompt engineering layer to modifty the prompt for better perfromance
 """
 import openai
+from fontTools.ttLib.tables.ttProgram import instructions
 from openai import OpenAI
 
 
@@ -18,8 +19,25 @@ class PromptEngine:
         :return:
         """
 
-        new_prompt = self.get_llm_response(prompt)
-        return new_prompt
+        new_prompt = f"""
+
+Modify below prompt following best prompt engineering methods. return only the new prompt as a text.
+modify the prompt and instructions in <original_prompt> tag to maximimize better results by providing the new prompt.
+
+### Original prompt
+
+<original_prompt>
+
+{prompt}
+
+</original_prompt>
+
+output the new prompt as text without any additional information.
+        
+        """
+
+        final_prompt = self.get_llm_response(new_prompt)
+        return final_prompt
 
     # ============================================================
     def llm_instructions(self):
@@ -39,20 +57,33 @@ class PromptEngine:
 
         openai.api_key = self.Core.api_key
         client = OpenAI(api_key=self.Core.api_key)
+        reasoning = self.Core.reasoning_model
+        system_prompt = self.llm_instructions()
 
         for attempt in range(max_retries):
             try:
-                response = client.chat.completions.create(
-                    model=self.Core.model,
-                    response_format={"type": "text"},
-                    messages=[
-                        {"role": "system", "content": self.llm_instructions()},
-                        {"role": "user", "content": prompt}
-                    ],
-                    max_tokens=1000,
-                    n=1,
-                    temperature=self.Core.temperature,
-                )
+                if reasoning:
+                    response = client.chat.completions.create(
+                        model=self.Core.model,
+                        response_format={"type": "text"},
+                        messages=[
+                            {"role": "system", "content": system_prompt},
+                            {"role": "user", "content": prompt}
+                        ],
+                        reasoning_effort="medium",
+                        n=1,
+                    )
+                else:
+                    response = client.chat.completions.create(
+                        model=self.Core.model,
+                        response_format={"type": "text"},
+                        messages=[
+                            {"role": "system", "content": system_prompt},
+                            {"role": "user", "content": prompt}
+                        ],
+                        n=1,
+                        temperature=self.Core.temperature
+                    )
 
                 tokens = {
                     'prompt_tokens': response.usage.prompt_tokens,

messaging_main_test.py

@@ -164,8 +164,10 @@ if __name__ == "__main__":
     segment_name = "no_recent_activity"
     permes = Permes()
 
+    # o3-mini o1-mini o4-mini o1
+
     users_message = permes.create_personalize_messages(session=session,
-                                                model="gemma3:4b",
+                                                model="o4-mini",
                                                 users=users,
                                                 brand=brand,
                                                 config_file=config_file,