AI Prompt Builder

This Streamlit application helps users refine their system prompts for AI assistants using either a local Ollama LLM or Azure OpenAI.

Features

• Prompt Refinement: Submit an initial system prompt and receive a more effective and well-structured version generated by a Large Language Model (LLM).
• Provider Selection: Choose between Ollama (local LLM) and Azure OpenAI (cloud-based LLM) as your AI provider.
• Model Selection: Select from a list of available models for the chosen provider directly within the application.
• Configurable System Prompt: The core system prompt for the prompt builder itself is externalized in config.yaml, allowing for easy modification and versioning.
• LLM Call Logging: Every interaction with the LLM is logged in a structured JSON format (llm_calls.log), providing details such as timestamp, prompt ID, model used, tokens, latency, and outcome.
• Secure Credential Management: Azure OpenAI API keys and endpoints are loaded securely from a .env file, keeping sensitive information out of the codebase.

Setup

Follow these steps to get the AI Prompt Builder running on your local machine.

Prerequisites

• Python 3.8+: Ensure you have Python installed.
• Ollama (Optional, for local LLM): If you plan to use Ollama, you need to have it installed and running. Download it from ollama.ai.
• Ollama Models (Optional): If using Ollama, download the required LLM models (e.g., llama2, mistral, codellama) using the Ollama CLI. For example:

  ollama pull llama2
  ollama pull mistral
  ollama pull codellama

• Azure OpenAI Account (Optional, for cloud LLM): If you plan to use Azure OpenAI, you will need an active subscription, an Azure OpenAI resource, and a deployed model.

Installation

1. Clone the repository (if applicable) or navigate to the project directory:

   cd /Users/m/projects/work/od/promptER

2. Create a Python virtual environment:

   python3 -m venv .venv

3. Activate the virtual environment:

   • On macOS/Linux:

     source .venv/bin/activate

   • On Windows:

     .venv\Scripts\activate

4. Install dependencies:

   pip install -r requirements.txt

Configuration

config.yaml

The application's behavior can be configured via the config.yaml file:

{
    "provider": "ollama",
    "ollama": {
        "model": "llama2",
        "models": [
            "llama2",
            "mistral",
            "codellama"
        ]
    },
    "azure_openai": {
        "models": [
            "gpt-4",
            "gpt-35-turbo"
        ]
    },
    "system_prompt": {
        "id": "sp-4f8c2e",
        "version": "0.1.0",
        "purpose": "To assist users in refining system prompts for AI assistants.",
        "owner": "Gemini",
        "date_created": "2025-10-23",
        "intended_model": "llama2",
        "tags": [
            "prompt-engineering",
            "ai-assistant",
            "refinement"
        ],
        "notes": "This is the initial version of the system prompt.",
        "content": "You are a prompt engineering assistant. Your task is to take a user's system prompt and rewrite it to be a more effective and well-structured prompt for an AI assistant. Provide only the refined prompt in your response."
    }
}

• provider: The default LLM provider to use (ollama or azure_openai).
• ollama.model: The default Ollama model to be selected.
• ollama.models: A list of available Ollama models that the user can select from. Ensure these models are pulled in your Ollama instance.
• azure_openai.models: A list of available Azure OpenAI models (deployment names) that the user can select from.
• system_prompt: An object containing metadata and the content of the system prompt used by the prompt builder itself.
  • id: Unique identifier for the system prompt.
  • version: Semantic version of the system prompt.
  • purpose: Description of the system prompt's role.
  • owner: Creator of the system prompt.
  • date_created: Date the system prompt was created.
  • intended_model: The model this system prompt was designed for.
  • tags: Keywords for categorization.
  • notes: Any additional notes.
  • content: The actual system prompt text that guides the LLM's behavior in refining user prompts.

.env file

If you are using Azure OpenAI, create a file named .env in the root directory of the project (the same directory as app.py and config.yaml). This file will store your sensitive API credentials.

.env.template:

AZURE_OPENAI_API_KEY=
AZURE_OPENAI_ENDPOINT=
AZURE_OPENAI_DEPLOYMENT_NAME=

Example .env file (replace with your actual credentials):

AZURE_OPENAI_API_KEY="your_azure_openai_api_key_here"
AZURE_OPENAI_ENDPOINT="https://your-resource-name.openai.azure.com/"
AZURE_OPENAI_DEPLOYMENT_NAME="your-deployment-name"

Important: Do not commit your .env file to version control. Add it to your .gitignore file.

Running the Application

1. Ensure your virtual environment is activated.

2. Ensure your Ollama server is running (if using Ollama).

3. Start the Streamlit application:

   streamlit run app.py

   (Note: If you ran it in the background previously, you might need to kill the old process first: kill $(ps aux | grep streamlit | grep -v grep | awk '{print $2}'))

4. Open your web browser and navigate to the URL provided by Streamlit (usually http://localhost:8501).

Logging

All LLM calls are logged to llm_calls.log in JSON format. This file can be used for monitoring, debugging, and analysis of LLM interactions.

{"message": "LLM call successful", "prompt_id": "sp-4f8c2e", "prompt_version": "0.1.0", "model_used": "llama2", "tokens_used": 123, "latency": 0.543, "outcome": "success", "time": "2025-10-23T10:30:00.123456"}