Conversations
Ava conversations offer a simple, chat-based interface where you can enter freeform questions about your data and leverage generative AI to produce insights.
Open a New Conversation
To start a conversation with Ava, simply click the New Conversation button in the sidebar. This opens an "Unnamed" conversation in the conversation history.
Each individual conversation is distinct from your other conversations. Ava remembers the questions asked and responses given in each conversation, but these questions and responses can only be referenced in the same conversation they were created in.
Load Data
When opening a new conversation, the a Files form appears. Ava supports the following file types:
- Comma-separated values (.csv)
- Tab-separated values (.tsv)
- Excel (.xlsx)
- JavaScript Object Notation (.json)
To load a file:
Under Upload Files, either drag and drop your file, or browse your local machine.
Optionally, you can instead select a previously uploaded file from My Files.
Click Load into conversation.
If the window is closed without any loaded data, you can click the Load Data button to reopen the window.
Once loaded, Ava provides a sample of the dataset.
Name Your Conversation
Conversations by default are "Unnamed". We recommend renaming conversations to avoid multiple "Unnamed" conversations in your conversation history. You can rename a conversation by clicking More options > Rename and entering a new name.
Collapse the Sidebar
Optionally, you can collapse the sidebar to put your conversation in focus:
Ask Questions
Once data is loaded into your conversation, you can begin asking questions. Enter your question in natural language, then click the wand icon or press Enter to let Ava find the answer. For example, consider a dataset on AQI levels. If you wanted to know how many records are in the current dataset, you could simply ask: “How many records are in this dataset?”
Responses
After entering a question, Ava automatically finds and runs the steps needed to get to your answer. You can watch the solution progress in Ava's response.
In this example, Ava provides us with an answer to our original question, showing us that fall has the highest average riders.
Responses from Ava can vary, depending on the questions asked. For example, more in-depth questions that require more analysis can produce a response including multiple tables and charts. Using the previous example, if you then ask "Which factors most impact AQI_level?", Ava's response may look more like this:
Click either the right or left arrows to navigate between charts and tables.
Answer Explanation
When a response is provided, you can view the summary of steps taken to get that answer in the response header.
To view a more technical explanation, you can click the Steps button to see the GEL steps used. In DataChat's complementary product, DataGrid, GEL steps can be edited for more granular control.
Feedback Buttons
When a response is provided, thumbs-up and thumbs-down feedback buttons appear in the top-right of the response. You can click either button to give feedback about Ava's response and whether or not it was beneficial to your data investigation. Providing feedback helps Ava to improve your experience.
You can also submit more detailed feedback to the DataChat team by clicking the purple Feedback button on the right.
Next Steps
From here, you can continue your data investigation by asking Ava more questions in the same conversation. If you'd like to start a new topic using the same data, you can sweep the current conversation, or if you'd like to explore a different dataset, you can create a new conversation.
Sweep Conversation
If you'd like to begin a new topic with Ava using the same datasets, you can sweep the current conversation instead of opening a new conversation. Sweeping prevents your previous questions and Ava's previous responses from further being used when asking new questions. To sweep the conversation, simply click the Sweep button.
When the conversation is successfully swept, the Sweep button becomes disabled until more questions are asked.
Share Outputs
Entire Conversation
To share your conversation history with others:
Click the Export icon. This opens a form displaying a preview of the conversation.
Click Create Link.
Select to share either by copying the URL or embedding the corresponding code into an external site or dashboard.
Visualizations
If you'd like to only share the visualizations Ava generated, you can create a public link to the response that can then be shared with others:
Click the Share icon next to the output you'd like to share. This opens a form where you can select which chart to share.
Click Create Link.
Select to share either by copying the URL or embedding the corresponding code into an external site or dashboard.
Delete a Conversation
To delete an existing conversation, in the Conversations sidebar, click More options > Delete next to the conversation you'd like to remove.
