What is Augini about?
augini enables synthetic data generation in an easy to use interface.
Some datasets are hard to get access to. Synthetic data generation allows one to run quick experiments on datasets that are very similar to ground truth.
The tool currently exposes three easy to use interfaces/methods to get to work quickly.
User-Facing Methods:
augment_columns:
Purpose: This method is used to augment (add new synthetic data to) specific columns in a DataFrame.
Usage: Users can specify which columns/column names they want to augment and optionally provide a custom prompt.
Example:
augmented_df = augini.augment_columns(df, columns=["column1", "column2"], custom_prompt="Custom prompt here", use_sync=False)
Columns can be of arbitrary size. That is, they can be more than the two presented above.
augment_single:
Purpose: This method is used to augment a single column in a DataFrame.
Usage: Similar to augment_columns, but specifically for a single column.
Example:
augmented_df = augini.augment_single(df, column_name="column1", custom_prompt="Custom prompt here", use_sync=False)
print(augmented_df.head())
chat:
Purpose: This method allows users to ask natural language questions about a DataFrame and receive AI-generated responses.
Usage: Users provide a query and the DataFrame, and the method returns an AI-generated answer based on the DataFrame's context.
Example:
You can ask follow up questions by simply querying augini again.
response = augini.chat(query="What is the average presented in this case? The median or the mean?", df=df)
print(response)
Chat oriented methods:
get_conversation_history:
Purpose: Retrieves the conversation history in either 'full' or 'summary' mode.
Usage: Users can get an overview of the current conversation context. It can be helpful to know what data and conversation details augini will be interacting with.
Example:
full_history = augini.get_conversation_history(mode='full')
print(full_history)
summary_history = augini.get_conversation_history(mode='summary')
print(summary_history)
The full mode presents an overview of the entire conversation. Summaries make use of embeddings to aim to reduce the cost of api queries ensuring that only contextually relevant information is used in any query.
clear_conversation_history:
Purpose:Clears the conversation history based on the specified mode.
Usage: By clearing the conversation history you start from a new slate making it less likely that a previously error filled context affects new instructions thereby increasing accuracy.
Example: