Introduction to RunAI
More and more often, I find myself meeting with companies of all types and sectors to discuss improving business efficiency.
The Data Security Challenge
Many of these companies want to develop solutions based on artificial intelligence, but most operate in environments where data security is a critical and fundamental factor to maintain.
This often creates a dilemma, as the most advanced AI technologies require access to large amounts of data to function optimally and frequently reuse the same user data to retrain and improve their models.
From this perspective, companies that handle sensitive and private data, with contracts containing information such as customer names and financial details, may face serious difficulties in sharing their data with external AI platforms.
The Solution: An Internal Inference Data Center
Why not invest in creating a small internal inference data center?
An internal inference data center would allow companies to use AI models without having to share their sensitive data with third parties, ensuring maximum security and control over their data.
RunAI Architecture
This is how RunAI was born. RunAI offers a secure platform with a simple yet robust architectural scheme.
Data Security and Privacy
The client sends a request to RunAI.it, protected by SSL TLS 1.3 connection, which acts as a proxy. A load balancer assesses the least busy server and redirects the call to that node. Proxy and Nodes are connected via an end-to-end encrypted VPN tunnel.
The request arrives at the node and is processed by the model. Once the response is generated, it is deleted one millisecond after being sent to the client. This approach ensures that sensitive data is never stored on the server, guaranteeing the highest level of security and privacy.
The logs saved by the platform are strictly technical logs, containing no information about the data sent by users or the responses provided.
The infrastructure is entirely Italian and complies with GDPR regulations.
RunAI Models
RunAI utilizes state-of-the-art language models, trained on vast and diverse text datasets to ensure accurate, creative, and relevant responses to a wide range of prompts.
RunAI Core
This is the core model, to be used as a starting point; it is a balanced and reliable model suitable for most requests. Costs are calculated per million tokens in input and output. RunAI Core is therefore the most cost-effective model for applications requiring a good balance between performance and cost. It is available on both Chat and API platforms.
RunAI Insight
This model specializes in clear, in-depth, and well-structured explanations. It is ideal for those who need to understand the mechanisms behind the responses generated by AI or to obtain detailed analyses on a specific topic. RunAI Insight is also available on both Chat and API platforms. Cost calculated per million tokens in input and output.
RunAI Vision
This is our multimodal model, which includes both text and images. It can generate detailed descriptions of images, answer questions about visual content, and interpret documents with OCR and layout recognition capabilities. It is available on both Chat and API platforms. Cost calculated per million tokens in input and output.
RunAI Swift
This is a very fast, lightweight, and fluid model: perfect for immediate responses in Chatbots, real-time applications, or interactive interfaces that require high-speed responses. It is available on both Chat and API platforms. Cost calculated per million tokens in input and output.
RunAI Codex
This model specializes in code generation and understanding. It can generate code in various languages, translate code from one language to another, and even explain how a specific scope of your code works. It is available on both Chat and API platforms. Cost calculated per million tokens in input and output.
RunAI Nano
This is the lightest model. Optimized for simple operations at a very low cost. Ideal for applications that require efficient resource usage, such as text-based chatbots or automated support systems. It is available only on the API platform. The cost is calculated per million tokens in input and output.
RunAI Vector Engine
This is the semantic engine that calculates representation vectors for text, images, and other types of data (which are always converted to text, e.g., base64). It can be used for applications such as semantic search, sentiment analysis, and creating recommendation systems. It is available on the API platform. The cost in this case is calculated per seconds of processing.
Other Features of RunAI
The RunAI portal offers other interesting features. For example, TTS and STT (Text-to-Speech & Speech-to-Text). The important thing to highlight is that to provide these two services, RunAI does not make calls to external cloud services but uses the Web Speech APIs of Google Chrome locally. This also ensures high speed and data security.
All chat history is only stored on the user's device; all conversations and chats can be deleted locally. The user can even delete individual messages as they would in any WhatsApp chat and can delete all conversations in one go.
Payment Methods and Credit Management
RunAI operates on a pay-as-you-go credit system. We did not want to introduce subscription plans where users are forced to pay a monthly fee even if they do not use the service. In RunAI, users make a top-up, just like a phone recharge, and have free access to all models both in chat and on the API platform. Each call is counted with a cost logic per million tokens in input and output and inference time for the Vector Engine.
This system allows for total control of one's budget. Users can choose how much to spend and are always aware of the costs incurred. The remaining credit is automatically updated after each individual call. There are no unexpected surprises; once the credit is exhausted, RunAI stops automatically, and there is no risk of unforeseen charges. The system sends emails to notify low credit when reaching 10 euros remaining and a second email when reaching 5 euros remaining in budget.
Users can recharge their credit independently from their reserved area via PayPal or Stripe.
Invitation to Sign Up and Gift
Visit the website runai.it, sign up, and you will receive a gift of 5 euros in credit to start using the platform.
