Use system instructions to steer the behavior of a model

lib/gpt_providers/text_generation/gemini_pro_text.py

@@ -2,7 +2,6 @@
 import os
 import sys
 from pathlib import Path
-import streamlit as st
 
 import google.generativeai as genai
 from dotenv import load_dotenv
@@ -21,7 +20,7 @@ from tenacity import (
 
 
 @retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(6))
-def gemini_text_response(prompt, temperature, top_p, n, max_tokens):
+def gemini_text_response(prompt, temperature, top_p, n, max_tokens, system_prompt):
     """ Common functiont to get response from gemini pro Text. """
     #FIXME: Include : https://github.com/google-gemini/cookbook/blob/main/quickstarts/rest/System_instructions_REST.ipynb
     try:
@@ -37,7 +36,9 @@ def gemini_text_response(prompt, temperature, top_p, n, max_tokens):
         "max_output_tokens": max_tokens,
     }
     # FIXME: Expose model_name in main_config
-    model = genai.GenerativeModel(model_name="gemini-1.5-pro-latest", generation_config=generation_config)
+    model = genai.GenerativeModel(model_name="gemini-1.5-pro-latest", 
+                                  generation_config=generation_config,
+                                  system_instruction=system_prompt)
     try:
         # text_response = []
         response = model.generate_content(prompt, stream=True)
@@ -45,7 +46,6 @@ def gemini_text_response(prompt, temperature, top_p, n, max_tokens):
             for chunk in response:
                 # text_response.append(chunk.text)
                 print(chunk.text)
-                #st.write(chunk.text)
         else:
             print(response)
         logger.info(f"Number of Token in Prompt Sent: {model.count_tokens(prompt)}")

lib/gpt_providers/text_generation/main_text_generation.py

@@ -27,7 +27,21 @@ def llm_text_gen(prompt):
         str: Generated text based on the prompt.
     """
     try:
+        # Read the config param to create system instruction for the LLM.
         gpt_provider, model, temperature, max_tokens, top_p, n, fp = read_return_config_section('llm_config')
+        blog_tone, blog_demographic, blog_type, blog_language, \
+            blog_output_format, blog_length = read_return_config_section('blog_characteristics')
+
+        # Construct the system prompt with the sidebar config params.
+        system_instructions = f"""
+        Below are the guidelines to follow:
+        1). You must respond in {blog_language} language.
+        2). Tone and Brand Alignment: Adjust your tone, voice, personality for {blog_tone} audience.
+        3). Make sure your response content length is of {blog_length} words.
+        4). The type of blog is {blog_type}, write accordingly.
+        5). The demographic for this content is {blog_demographic}.
+        6). Your response should be in {blog_output_format} format.
+        """
 
         #gpt_provider = check_gpt_provider(gpt_provider)
         # Check if API key is provided for the given gpt_provider
@@ -37,7 +51,7 @@ def llm_text_gen(prompt):
         if 'google' in gpt_provider.lower():
             try:
                 logger.info("Using Google Gemini Pro text generation model.")
-                response = gemini_text_response(prompt, temperature, top_p, n, max_tokens)
+                response = gemini_text_response(prompt, temperature, top_p, n, max_tokens, system_instructions)
                 return response
             except Exception as err:
                 logger.error(f"Failed to get response from gemini: {err}")
@@ -45,7 +59,7 @@ def llm_text_gen(prompt):
         elif 'openai' in gpt_provider.lower():
             try:
                 logger.info(f"Using OpenAI Model: {model} for text Generation.")
-                response = openai_chatgpt(prompt, model, temperature, max_tokens, top_p, n, fp)
+                response = openai_chatgpt(prompt, model, temperature, max_tokens, top_p, n, fp, system_instructions)
                 return response
             except Exception as err:
                 logger.error(f"Failed to get response from Openai: {err}")

lib/gpt_providers/text_generation/openai_text_gen.py

@@ -14,7 +14,7 @@ from tenacity import (
  
 
 @retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(6))
-def openai_chatgpt(prompt, model, temperature, max_tokens, top_p, n, fp):
+def openai_chatgpt(prompt, model, temperature, max_tokens, top_p, n, fp, system_prompt):
     """
     Wrapper function for OpenAI's ChatGPT completion.
 
@@ -45,7 +45,8 @@ def openai_chatgpt(prompt, model, temperature, max_tokens, top_p, n, fp):
         client = openai.OpenAI(api_key=os.getenv('OPENAI_API_KEY'))
         response = client.chat.completions.create(
             model=model,
-            messages=[{"role": "user", "content": prompt}],
+            messages=[{"role": "system", "content": system_prompt}, 
+                      {"role": "user", "content": prompt}],
             max_tokens=max_tokens,
             n=n,
             top_p=top_p,