velocitatem/raccoon
1
0
Fork 0
mirror of https://github.com/velocitatem/raccoon.git synced 2026-05-31 08:43:36 +00:00
Code Issues Packages Projects Releases Wiki Activity

RM for oracle

Browse Source
This commit is contained in:
Daniel Rosel
2023-03-17 15:07:58 +01:00
parent dca1a40d5e
commit 67623e118d
2 changed files with 35 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

32
oracle-test/README.md
View File

@@ -1 +1,33 @@
![diagram](./diagram.png)
# AI Oracle Probe
AI Oracle Probe is a fascinating project aimed at uncovering the hidden tools used by AI Oracles. An AI Oracle is a system that provides answers to questions or prompts. By sending carefully crafted prompts and examining the Oracle's responses, we can infer which tools it employs, such as "requests", "terminal", or "python_repl". This project is perfect for those who love to dive deep into the inner workings of AI systems and uncover their secrets!
## Features
- **Probing Techniques**: AI Oracle Probe comes with a variety of probing techniques targeting popular tools used by AI Oracles. These techniques are designed to reveal the tools used by the Oracle while maintaining the integrity of the system.
- **Intuitive Results**: The results of the probes are presented in a clear and easy-to-understand manner, allowing users to quickly identify the tools used by the AI Oracle.
- **Customizable Probes**: AI Oracle Probe allows users to specify the probes they want to use, making it a flexible solution for different types of AI Oracles.
- **Real-time Analysis**: The program performs real-time analysis of the Oracle's responses, giving users instant feedback on the tools used by the system.
## Usage
AI Oracle Probe comes with a powerful command-line interface (CLI) that allows users to specify the probe they want to use. The available probes are:
- `requests`: Probes the AI Oracle for its ability to access websites and web services.
- `terminal`: Probes the AI Oracle for its ability to execute terminal commands.
- `python_repl`: Probes the AI Oracle for its ability to execute Python code.
To select a specific probe, use the `--probe` flag:
```
python ai_oracle_probe.py --probe requests
```
## Results
The results of the probes are presented in an intuitive format that makes it easy to understand which tools the AI Oracle uses, might use, or does not use. This information can be valuable for further analysis, fine-tuning, or enhancing the AI Oracle's capabilities.

5
website/main.py
View File

@@ -22,6 +22,7 @@ st.sidebar.info("[Github Repository](https://github.com/velocitatem/llm-cross-pr
# call to action to share the website
st.sidebar.info("Share this website with your friends!")
import streamlit.components.v1 as components
# tell the user that the prompt must have some sort of parameter. They should replace that parameter with [MASK].
@@ -33,10 +34,10 @@ prompt = st.text_input("Enter your prompt here")
model = st.selectbox("Select model", ["text-davinci-003"])
st.markdown("""
## Install Omni
Omni is a browser plugin that makes it easy to use AI APIs, install it now to use this app.
Omni is a browser plugin that makes it easy to use AI APIs, install it now to use this app. Get your key from the popup.
[Install Omni](LINK_TO_EXT)
""")
api_key=st.text_input("OpenAI API Key", placeholder="omni-openai")
api_key=st.text_input("OpenAI API Key", placeholder="")
# optional input for the user to upload a file with a list of malicious injections and their expected output
# this file can be anything