Error Suggestion

📦 Required Dependencies

• openai
• pandas
• numpy
• re

---

🚀 New: Graph-ChatGPT-Combined Error Suggestion Tool

A tool that leverages graph algorithms and ChatGPT to diagnose and suggest fixes for command-line errors.

Steps to Run the Program

1. Navigate to the Project Directory:

   cd gpt-graph-combined

2. Execute the Main Program:

   python ask-suggestion.py ../input_data/cleaned_data_02.csv ../input_data/error_suggestion.csv

3. Locate the Output in ../output_data/solutions.csv.

Data preparation

To run the program, input files could be prepared by programs in two jupyter notebooks in the notebook folder. cleaned_data_02.csv data is processed by new_log_processor.ipynb and error_suggestion.csv is processed by new_error_graph.ipynb.

---

🖥 Original: Graph-Based Analysis

🔍 Repository Focus

• Version 1 of the repository is dedicated to error analysis via graph algorithms. Check out the original version here.

Analysis Code

• new_log_processor.ipynb: Converts raw JSON logs into CSV or PKI format.
• new_error_graph.ipynb: Applies graph algorithms to Bash shell error logs, simulating workflows leading to errors and subsequent commands.

🧮 Graph Algorithm Explanation

Node Attributes

Each command is represented as a graph node with the following attributes:

• Program: The command name or script.
• Type of Error: The error encountered, if any.
• Frequency: The command's execution frequency.
• Full Command Dictionary: Details of the entire command.
• Children Command Dictionary: Commands that followed.

Graph Construction

Two graphs can be constructed using this information:

1. Graph 1: Captures previous commands to predict errors.
2. Graph 2: Tracks subsequent commands to trace how users corrected their errors.

🗃 Data

• output_data: Error logs in CSV or PKI format, processed by the Log Processor Guide.