Chatbots

Creating a chatbot lets you collect and query a set of documents, such as business policies, support procedures, or product documentation. You can create a chatbot then share it with other users or members of your organization, so they can query the same document set. You can see chatbots you created and those shared with you in the Hub.

Chatbots rely on a knowledge base of files you upload during an origin conversation. These files are visible in the chatbot interface, letting users verify the chatbot’s answers against the knowledge base and understand which resources are referenced.

Creating chatbots

You can create a chatbot from conversations in your personal workspace.

1. Open an existing conversation or create a new one.

2. Add the documents that you want to include in the chatbot and enable any digitization settings, if needed. These documents become the knowledge base your chatbot uses to develop responses.

   See the Converse documentation for guidance on advanced settings available when uploading files, such as table or checkbox extraction.

3. (Optional) Query your documents to test chatbot performance, ensuring the message scope is set to all documents. During chatbot creation, you have the option to turn these queries into sample prompts that are visible to your chatbot’s users.

4. Click Create chatbot.

5. Enter a name and description for the chatbot. You can’t change the name after creating the chatbot, but you can update the description.

6. (Optional) Add a disclaimer to include in all chatbot responses. Disclaimers are useful for communicating legal information, technical limitations, or other important details.

7. Commercial & Enterprise Select sharing and access preferences, so other members of your organization can use the chatbot.

   For community accounts, all chatbots are shared but unlisted by default. Any chatbot you create is publicly available, meaning any AI Hub user with a link can access and query your chatbot, including being able to view any files in the chatbot’s knowledge base. Although available by link, your chatbots aren’t discoverable in the Hub—only after using your chatbot can another user see it in their Hub. When other users query your chatbot, you are not charged and you can’t see their queries or results.

   The ability to limit sharing to yourself or members of your organization is available at the commercial or enterprise tier.

8. (Optional) Add sample prompts to display in the chatbot. Queries from your origin conversation are imported, but you can edit the sample prompts list.

9. Click Create. Depending on the number of documents added to the chatbot, this can take up to several minutes. Do not refresh or navigate away from the page during chatbot creation—if you do, creation might fail.

After your chatbot is created, you can access it from your personal workspace or in the Hub to use, share, or edit it.

Updating chatbots

You can edit documents, prompts, and most settings for any chatbot you’ve created. You can’t change a chatbot’s name or author. You also can’t edit or delete another user’s chatbot.

During chatbot editing, you can add or remove documents in the chatbot’s knowledge base, add new sample prompts, and change settings and details. When you update your chatbot, a new chatbot version is published and all chat history for the previous version is deleted. If your chatbot is shared, other users also lose their chat history.

With each new version, the chatbot ID is also updated. For organization accounts, the shareable URL for your chatbot includes the chatbot ID. For convenience, when someone accesses a chatbot with a URL that has a previous version’s ID, they’re redirected to the chatbot’s latest version. For community accounts, your chatbot’s shareable URL uses the chatbot’s name instead of the ID, and users always see the latest version.

1. Open your chatbot and click Edit chatbot. The edit function returns you to the origin conversation.

   • To add files to the chatbot, click Add files in the file panel.

   • To remove files, click the Delete (trash can) icon next to the document in the file panel.

   • To edit advanced settings like digitization, click the Settings (gear) icon.

2. Click Update chatbot.

3. On the Update chatbot screen, make any changes you want to settings such as description, sharing settings, or sample prompts. You cannot change a chatbot’s name or author.

4. Click Update chatbot. Depending on the number of documents added to the chatbot, this can take up to several minutes. Do not refresh or navigate away from the page during chatbot creation—if you do, creation might fail.

Deleting chatbots

When you delete a chatbot, all documents and chat history for that chatbot are deleted.

1. Open the chatbot.

2. In the side panel, click Delete chatbot.

3. Confirm the deletion by clicking Delete.

Using chatbots

Because chatbots are built from origin conversations in Converse, they share many of the same features and limitations when making queries. Much of the Converse documentation’s guidance on conversing with your documents applies, as does the advice on getting the most out of Converse, such as interface tips and tricks and advice on writing good questions.

Some key differences between conversations and chatbots include:

• You can’t edit message scope in a chatbot. A chatbot query is always made to the entire chatbot knowledge base.

• Because message scope can’t be defined, the model used to generate a chatbot response is based on the number of documents in the knowledge base. Unless research mode is enabled, queries to a single-document knowledge base use the advanced model and queries to a multiple-document knowledge base use the multistep model. When using research mode, a more powerful but slower variant of the multistep model is used.

• Response feedback submitted in a chatbot is also visible to the chatbot’s creator, to help them improve the chatbot’s performance.

• Chatbots don’t support querying information retrieved using object detection, such as tables and checkboxes. If a chatbot’s origin conversation has object detection enabled, the same query can yield different responses in the chatbot.

• When using research mode in Converse, it’s noted in the response window. Chatbots don’t yet support highlighting which responses were generated using research mode.