Admin User Guide

Admin Features

This section is a guide for Rapidflare Admins. Rapidflare Admins help manage your Rapidflare account, train and deploy AI Agents and overall help get the best out of Rapidflare. Rapidflare Admins get access to the Rapidflare Dashboard, where they can do administrative tasks, configuration and tuning of AI Agents, or get access to conversations, analytics and more!

Conversation History

You can access the history of all conversations your enterprise had with the AI Agent via the Home > History navigation link. You can view the history for a given time range. Various time range filters are available to help you zoom in or out to a time range.

You can page through the list of conversations, filter by feedback, user ID, and other attributes. You can sort and filter by various columns such as timestamp, user ID, feedback and number of messages.

You can click on the Title column to bring up a past conversation from history.

Reviewing Feedback

In the history view, you can easily review feedback provided by users. Since feedback is at an individual message level, we roll up individual message feedback up to each conversation, and visualize the summary in the table. Here are some scenarios

• If a conversation has 1 or more messages with thumbs-up feedback, we summarize the conversation's feedback as "Good".
• If a conversation has 1 or more messages with thumbs-down feedback, we summarize the conversation's feedback as "Bad".
• If a conversation has 1 or more messages with thumbs-down feedback and 1 or more messages with thumbs-up feedback, we summarize the conversation's feedback as "Mixed".
• If a conversation has no messages with thumbs-up or thumbs-down feedback, we summarize the conversation's feedback as "None".

You can also quickly review comments provided with the feedback ratings by hovering over the icon presented representing one of the scenarios above. You will see a bulleted list of comments provided, if any.

Exporting history

You can export history by clicking on the Export button and choose the format preferred. We currently support exporting conversation history in Excel, JSON, Markdown and HTML formats. If you would like other history export in other formats, please get in touch with us at support@rapidflare.ai.

Demo

Here's a video demo illustrating conversation history and history export.

Sharing a conversation

You can share a conversation by clicking on the Share icon. This creates a link that can be shared. There are 4 modes of sharing a conversation:

• Share via URL: You can always share the conversation with others within your enterprise, by sharing this link (or just grab it off of the browser bar when you are in the Converse view). Only authenticated users into the dashboard will be able to view and interact with the AI Agent using this link.

• Share via Email: This option lets you share the conversation with specific email addresses, sending them a notification that includes the conversation link and your custom message. Only authenticated users into the dashboard will be able to access and interact with the conversation through this method as well. Note: The conversation link shared via this method will expire 30 days after being generated.

• Share via Hubspot: This option lets you share the conversation by plugging in the URL in your CRM like Hubspot. Using this you can add more qeury paramters like email, firstname and lastname to share personalized links. All users that the link is shared with will be able to view the conversation. Authenticated is not required. Therefore this should be used with caution. Note: The conversation link shared via this method will expire 30 days after being generated.

• Share with anyone: This option lets you share the conversation with anyone who has access to the link. Authenticated is not required. Therefore this should be used with caution. Note: The conversation link shared via this method will expire 30 days after being generated.

Configuration

You can configure how your AI Agent experience looks like by setting name, description, placeholder text, and logo image of your choice. You can also change colors such as background, accent color and text color to match your brand's theme, using the Configure navigation link.

Here's a video demo illustrating AI Agent configuration:

Starter prompts

You can choose to show custom starter prompts to your users. You can group starter prompt questions by their category names and also provide a simple icon and a help description per category which will be shown to users on hover.

Here's a video demo illustrating starter prompts configuration:

You can also auto-generate these prompts using the AI Agent, here's a video demo of auto generating starter prompts:

API keys

Manage your API keys from Hubs » Configure » Security. One table lists every key with its type, bot protection, allowed domains and expiry. The chips above the table filter by type and show how many keys of each type you have.

Rapidflare has two kinds of API key. You choose the type when you create a key and it cannot change afterwards.

The prefix shows a key's type at a glance. Keys created before the prefixes were introduced start with rf_ alone and work exactly as before. Check the key's type in the table if you are unsure.

Currently, Rapidflare supports secret keys for instantiating and using embedded widgets in protected automation environments. This lets you test your Rapidflare widgets without running into reCAPTCHA checks, failures and human verification gates. A secret key is intended only for developers running automation tests, and it does not give open-ended access to your hub. It must never be logged, stored on disk, or accessible to your end users.

Create a key

Click Create key, choose the type, name the key (up to 50 characters) and choose when it expires: 90 days, 1 year, 2 years (the default) or a custom date and time. A publishable key also needs at least one allowed domain and a bot protection setting.

A secret key needs only a name and an expiry.

The dashboard shows the new key once. Copy it somewhere safe before you close the window; if you lose it, create a new key. Everywhere else the dashboard shows only ***** and the last five characters.

Update, disable or delete a key

Click Update on a key's row to change its name, turn it on or off, change its expiry and, for a publishable key, its allowed domains and bot protection. The type and the key value cannot be changed. The status switch in the table enables or disables a key without opening the dialog.

A disabled or expired key stops working immediately. Delete is permanent: any embedded agent or API call that uses the key stops working at once.

If a key leaks

A publishable key is meant to be visible in your page code, so seeing it there is not a leak. A secret key is different: it skips the domain and bot checks, so it must stay private. Never log it, store it on disk, or make it accessible to your end users. If a secret key ends up in a web page, an app, a repository, a chat, a ticket or a log, delete it at once and create a new one. Creating a separate key for each site or environment lets you delete one without breaking the others.

Allowed domains

A publishable key works only on the domains you list, and requests from any other domain are blocked. Every publishable key needs at least one allowed domain.

You can add up to 250 domains. Type a domain and press Enter or comma, or paste a list. Do not include a protocol such as http:// or https://, slashes, port numbers or query parameters. If your website is https://www.example.com, add example.com. All sub domains are included automatically, such as www.example.com or child.example.com.

Domains must be real, public hostnames. IP addresses, such as 10.0.0.1, are not accepted: a browser always sends the domain of the site the widget is on, so the domain is what identifies it. Local names (localhost, *.localhost and [::1]) are not accepted either. An internationalised domain name in its xn-- form is accepted. To test the widget before launch, use a hostname you control, such as staging.yoursite.com, which yoursite.com already covers. For automated testing in a protected environment, use a secret key.

Keys created before domains were required keep working. The next time you edit one, the dashboard asks you to add a domain.

Bot protection

Every request from a publishable key is checked by Google reCAPTCHA Enterprise. Google's terms of service and privacy policy apply.

• Score check (always on). Each visitor session is scored invisibly. Nothing is shown to the visitor, and it cannot be turned off.
• Show a challenge. When a visitor's score is too low and this switch is on (the default for new keys), the visitor sees a one-click checkbox from Google and carries on after solving it. If the switch is off, the request is refused with no way through.

The keys you create here initialize your AI Agent widget as described here. Here's a sequence diagram of how Google's reCAPTCHA system works under the hood.

Here's a video walking through creating, updating and deleting an API key.

Web search controls

Choose when your agent can search public websites, restrict allowed domains, and set retrieval limits. See Control web search for your hub for the three modes, conversation toggle behavior, and advanced configuration.