FYRAN

What is FYRAN?

FYRAN is a platform for creating hosted AI chatbots; users can upload their own content and use it to build bots that are capable of answering questions based on that content. It is suitable for converting documents, text, and audio into tools for engaging with customers, employees, or learners through dialogue.

The product offers a web-based interface, embeddable components, and REST APIs, in addition to voice responses provided by digital avatars. It is not an open model, nor is it an open-source project whose source code can be freely modified.

Main functions

Robot for generating a knowledge base from documents

Users can upload PDF, TXT, DOCX, or MP3 files; the system extracts the content from these files and creates a dedicated robot. The user can then pose questions in natural language, and the robot attempts to find relevant passages in the available information to generate answers.

Audio content Q&A

MP3 support allows the use of podcasts, interviews, training recordings, and similar content in Q&A sessions, thereby reducing the need to manually transcribe the content before organizing it into Q&A pairs. Audio quality, overlapping speech, accents, and specialized terminology can affect the accuracy of the extraction results; it is therefore advisable to test with sample materials first.

Digital human replies

The Freya digital avatar enables virtual avatars to read out the robot’s responses in natural language, and it allows the generated content to be downloaded for use in videos or presentations. This feature requires more credits than ordinary text-based responses; the exact amount needed can be checked in the account dashboard.

Multilingual Q&A

Users can ask the robot to respond in their preferred language, which is suitable for cross-lingual FAQs and learning scenarios. The translations and responses may still deviate from the original text; key policies and professional content should retain their original wording along with human review.

Boost Accuracy

The Boost option employs additional processing to improve the accuracy and quality of responses. It also increases the amount of points consumed, with the specific quantity depending on the length of the response and the number of calls made.

Web and app integration

Once created, access can be provided through hosted pages, website components, JavaScript, or HTML embedding. Teams that need more control can use APIs to create, query, and manage bots within their own applications.

Delete and rebuild the robot

Users can delete existing robots and create new ones using different settings. The product description indicates that there may be restrictions on the number of times deletion is allowed in the future; therefore, the current approach to deletion cannot be considered an unlimited option forever.

Input, Output, and Status

Tutorial for web version

1. Create an account, and on the account page check the current plan, points, and file size limits.
2. Choose a data set with low sensitivity, and first check whether you have the rights to upload it and use it.
3. Upload the file in a supported format, and wait for the system to extract the content and generate the robot.
4. Prepare factual questions, boundary questions, and questions that are not present in the materials for the robot.
5. Test the direct questioning method, the rephrased questioning method, the multilingual questioning method, and the long-question method separately.
6. Check whether the answers can be supported by the original material, and observe the behavior when there are no answers.
7. Enable Boost or the digital avatar when needed, and keep track of the additional points consumed.
8. After confirming the effectiveness, an embedding method is generated and deployed to a test page first.
9. Limit the robot’s visibility range, and include privacy notices as well as a link to human customer service.
10. After going live, continuously check for failure issues, expired data, quota limits, and abnormal calls.

API integration process

FYRAN provides public REST API documentation, but no official SDK is available. Developers must log in using their account email and API key; once they have obtained a JWT, they can access the protected interfaces.

1. Create or obtain an API key on the account page; do not put the key in the browser code.
2. The backend calls the login interface and stores the returned JWT temporarily.
3. After creating the robot, save the namespace and associate it with your own tenant and permissions.
4. On the server side, the length of the input, the user’s permissions, and the frequency of calls are verified.
5. Call the message interface, and record the response, relevant paragraphs, as well as the amount of points used.
6. Return only the necessary fields to the frontend, in order to prevent the exposure of tokens, internal errors, and other identifiers related to robots.
7. Add administrator permissions, confirmation, and audit logs for deletion operations.
8. Before going live, test failure states such as 400, 401, 407, timeouts, and insufficient quota.

Price packages

The price range displays seven monthly subscription plans; the free plan comes with clear limits on the number of robots, messages, and characters that can be used. In other sections of the page, it is stated that FYRAN is available for free at the moment, and the terms allow for payment either on a monthly or annual basis, so the information provided isn’t entirely consistent.

Typical queries cost one basic point or message; digital avatars and Boost options incur additional costs. Before making a purchase, it is necessary to check on the account’s settlement page to see whether the corresponding plan is still available, as well as to understand the annual pricing, taxes, rules for handling excess usage, and how points can be reset.

Subscription, cancellation, and refunds

The terms state that subscriptions can be paid in advance on a monthly or annual basis, with automatic renewal at the end of each billing cycle; users can cancel this renewal either through their account or by contacting support. Any price changes will take effect after the current cycle ends, and users will be notified in advance.

Unless required by law, paid subscription fees are non-refundable. The terms also state that if a robot cannot provide an answer or the outcome is unsatisfactory, this does not constitute a reason for issuing points or a refund.

Suitable for users and scenarios

• Customer Service Team: Convert the product instructions and frequently asked questions into Q&A sections on the website.
• Educators: Provide review robots based on course documents or recordings.
• Content creators: Enable readers to search for articles, podcasts, and collections of materials.
• Training team: Convert systems and operation manuals into performance support tools.
• Developer: Create and invoke custom bots in applications via APIs.
• Marketing team: Use digital avatars to create presentations or video content for some of the answers.

Advantages

• It supports both regular documents and MP3 files, making it suitable for reusing existing audio content.
• There is no need to manually write fixed answers for each question one by one.
• Digital humans expand the answers from knowledge bases into audio and presentation formats.
• It offers embedded components and APIs, providing both low-code solutions and the possibility for custom development.
• The free version allows you to test the quality of content extraction and responses before making a purchase.

Restrictions and risks

• AI responses may be incorrect, incomplete, or indicate uncertainty; they cannot replace professional advice.
• If the response is not satisfactory, a refund is generally not provided; the costs associated with testing must be taken into account when making a selection.
• The exact amount of digital human and Boost integration required is not listed explicitly on the public page.
• There is a conflict in the status of MP4 across the FAQ and API documentation; it should be tested in a real account.
• The package section is displayed alongside the “Currently Free” indication; the actual charging amount shall be based on the settlement page.
• When a company provides components to customers, it remains responsible for the content and privacy aspects.
• The public terms do not guarantee continuous availability, error-free operation, or uninterrupted service.

Data and Privacy

The product description states that only the content extracted from the uploaded files is stored, not the original files themselves, and that Azure and AWS servers are used for this purpose. Even if the original files are not kept for an extended period, the extracted content remains processed and stored; sensitive information should not be uploaded based solely on this description.

The personal data listed in the privacy policy include name, email address, cookies, IP address, device information, browser details, and usage records. This data can be used for providing services, offering support, conducting analyses, billing purposes, and marketing efforts; it may also be processed by service providers or transmitted to the United States.

• Stripe handles payment card information; the platform states that it does not store the details of the cards directly.
• Analysis services may include Google Analytics and Mixpanel.
• The retention period is determined based on the purpose, legal obligations, and the needs related to dispute resolution; there is no fixed number of days.
• Some users may request access, correction, deletion, restriction of processing, or withdrawal of consent.
• The service is not available for children under 16 years old.
• Companies should separately define the procedures for deleting end-user chat records, knowledge content, and backups.

Copyright, Commercial Use, and Open Source

Users retain the right to upload content, but they must ensure that they have the necessary authorization to use such materials and be responsible for the content they provide to their clients. The FYRAN platform, its functions, and its brand remain the property of the operator or licensor.

A paid subscription grants a limited, revocable, non-exclusive, and non-transferable license to install components on websites or software in order to serve customers. The terms prohibit reverse engineering, unauthorized use of the API, and unapproved automated access.

The source code of the FYRAN product, its official GitHub repository, or an open license have not been made available, so it should not be classified as open source. The availability of public APIs does not mean that the product is open source.

Frequently Asked Questions

Which files does FYRAN support?

PDF, TXT, DOCX, and MP3 are confirmed to be usable. MP4 is listed in the API documentation as an option for uploading, but the FAQ states that it will be available soon and requires actual testing.

How many bots can be created in the free version?

The price list specifies 2 robots, 50 messages, and 400,000 characters per robot. The details in the account dashboard may vary due to current promotions or policy changes.

Will digital humans incur an additional fee?

It is stated publicly that digital avatars consume more points than regular responses, but the amount used per time is not specified; the actual deduction can be checked in the backend.

Can it be embedded on a company’s official website?

It can be integrated using web components, JavaScript, HTML, or APIs. Before going live, it is necessary to set up privacy notices, permission settings, and a mechanism for manual upgrades.

Does FYRAN have an API?

Yes, the document includes interfaces for logging in to, listing, creating, sending messages to, and deleting robots. The number of times that a robot interface can be listed is limited to 5 times per minute.

Is an official SDK provided?

No standalone SDK has been confirmed; the documentation provides examples of requests in Python and PHP. Developers can integrate the system directly using REST interfaces.

Is FYRAN open source?

It is not a verifiable open-source product; the API and embedding features are part of hosted services subject to specific terms.

What gets saved after a file is uploaded?

The product description states that the extracted content is what is saved, rather than the original file. The exact details regarding the robot’s messages, chat records, and the retention period for backups should be confirmed with the service provider.

Can I get a refund if the answer isn’t satisfactory?

Generally, this is not possible; the terms explicitly exclude answers indicating ignorance or dissatisfaction as reasons for a refund, with legal requirements stipulating otherwise.

Can robot responses be used for professional advice?

It should not be used directly. The terms remind us that AI is not an expert and may make mistakes; medical, legal, financial, and other related matters must be reviewed by qualified professionals.

Summary

FYRAN’s strength lies in its ability to convert documents and MP3 content into knowledge base robots, offering various ways of interaction through digital avatars, components, and APIs. It is suitable for using the free tier to assess the quality of the content before deciding whether to increase the number of robots and the volume of messages.

Before formal deployment, it is essential to verify the actual price, the status of the MP4 files, the points associated with additional features, and data retention policies. Companies also need to establish their own procedures for informing end users, handling content licensing, reviewing responses, and controlling access.