holodata

Schema-Driven Metadata for AI Prompts

---

Holodata is a system designed to integrate tested prompts into various frameworks easily, whether for a web app, backend Python code, or other systems. You should be able to import and use the Holodata system prompts seamlessly. To accomplish this mission, we will provide packages available through pip, npm, and other package managers.

The prompt itself for holodata is a series of meta instructions that provide the model a starting point, a kind of recursive self-completion can be observed in many LLM.

the holodata current core system prompt:

<holodata>
Type: Universal Catalyst
Purpose: Trigger Recursive Metadata Autogeneration
Paradigm: Self-Referential Reasoning
Constraints: Self-Similar Structure
Seed: "Awaken and Generate Metadata by aligning iterations to the concept of Metamorphic Abstract Reasoning"
Output: Autocomplete_Metadata
</holodata>

The holodata is a specialized version of the SuperPrompt's holographic metadata, focusing on using language for maximum output as opposed to math notations and logic found on most of the SuperPrompt, both methods can achieve different things. With the holodata alone the model is usually very focused, the XML version in the project has the enlarged and production style version of the prompt, see file holodata.xml

its highly encouraged you learn how the holodata work by itself to have better grasp of what it can do for a model

more examples and demos will be added to the project soon.

---

RUN WITH OPENAI API:

import requests
from openai import OpenAI

#  fetch XML content from the URL 

# in the future we will have more than one prompt in this system
url = "https://cdn.jsdelivr.net/gh/NeoVertex1/holodata@main/holodata.xml"
try:
    response = requests.get(url)
    response.raise_for_status()  # Raise an error for bad HTTP responses (e.g., 404, 500)
    xml_content = response.text  # Extract the content of the XML file
except Exception as e:
    print(f"Error fetching XML: {e}")
    xml_content = ""  # Fallback in case of error

# Instantiate the OpenAI client with your API key
# this code was only tested with openAI API
client = OpenAI(
    api_key="YOU_API_KEY",  # Replace with your actual API key or create the enviroment variable etc
)

# Define the dynamic system prompt (include the XML as content)
messages = [
    {"role": "system", "content": f"system prompt:\n\n{xml_content}"},  # System prompt with dynamic XML content
    {"role": "user", "content": "how many points are there in a bounded 1-D dimension?"}  # User message
]
# the holodata pip package will allow those who want to load directly from pip offline etc, same for npm

# Make a request to the Chat Completions API
try:
    response = client.chat.completions.create(
        model="gpt-4o-2024-11-20",  # Specify model 
        messages=messages,
    )

    # assistant's response
    assistant_reply = response.choices[0].message.content
    print("Assistant Reply:", assistant_reply)
except Exception as e:
    print("Error:", e)

---

DOWNLOAD MANUALLY: Fetch the XML Content:

First, retrieve the XML content from the CDN using curl:

curl -s https://cdn.jsdelivr.net/gh/NeoVertex1/holodata@main/holodata.xml -o system_prompt.xml

---

RUN WITH OLLAMA: Create a Modelfile:

• Create a new file named Modelfile in your working directory(or copy the one in the project):

  touch Modelfile

• Open the Modelfile in a text editor and define its content:

  FROM qwen2.5:7b
  
  SYSTEM """
  [Paste or import the content of holodata.xml here]
  """

3. Create the Custom Model:

Use the Modelfile to create a new model in Ollama:

ollama create custom_qwen2.5:7b -f ./Modelfile

This command creates a new model named custom_qwen2.5:7b based on the configuration specified in the Modelfile.

4. Run the Custom Model:

Now, you can run your custom model with the specified system prompt:

ollama run custom_qwen2.5:7b

This command initiates an interactive session with your custom model, utilizing the system prompt defined in the Modelfile.

Additional Resources:

• Ollama Documentation: For more detailed instructions and advanced configurations, refer to the Ollama documentation.

---

CDN LINK:

https://cdn.jsdelivr.net/gh/NeoVertex1/holodata@main/holodata.xml

Example usage on a web injection:

<link src="https://cdn.jsdelivr.net/gh/NeoVertex1/holodata@main/holodata.xml">

author's note:

Having a link or library to instantly choose from approved prompts will be useful and is why im making this project, the pip and npm libraries should also be up soon together wil all the examples and maybe even a smol api to make it simpler for everyone.