An on-premise, low-code platform to build and operate AI agents and the processes around them — inside your perimeter.
PAIrasat («Система ИИ») lets a bank, pension fund, utility or state body build customer-facing and internal AI assistants without a development team. Administrators upload documents into named knowledge bases, design agents as block diagrams and wire them into backend processes in a visual flow editor. Every answer is grounded in your documents with citations and a confidence score; everything is journaled. Delivered as Docker Compose or K3s and runs entirely on your infrastructure. It is the core on which fAInance, AI-Lawyer, the AI Call Center and the other ALDIS agents run.
Runs inside your network on ALDIS on-premise servers. No public LLM APIs, role-based access, full audit log, personal-data masking. See the Security page for the full control list.
An agent is a block diagram: instruction, knowledge base, tools, model, topic filter and risk passport. No code required to create or change one.
Named knowledge bases with document upload and retrieval; each answer cites the fragments it used and carries a confidence score.
Visual DAG editor for backend processes: REST, Kafka, RabbitMQ and e-mail intake, database and HTTP nodes, sandboxed Python and JavaScript steps.
When confidence is low or the topic requires a person, the dialogue goes to an operator queue with full context.
A provider layer routes requests to local models for data that must not leave the perimeter, or to cloud models where policy allows — with personal-data masking in between.
Conversations, retrieved fragments, tool calls and security events are logged; security events ship to your SIEM.
Docker Compose or K3s inside your network; SSO and roles connected to your directory.
Upload regulations, product documents and FAQs into knowledge bases; test retrieval quality.
Compose agents and backend processes in the visual editors; set escalation rules and the risk passport.
Monitor quality, review operator queues, update knowledge; add agents on the same installation.
A platform. Chatbots, document assistants and back-office automations are built on it; the ALDIS product agents are examples of what it runs.
No. With local models it runs in a closed network. Cloud models can be enabled per agent if your policy allows.
Risk passport per agent, AI disclosure to users, operator hand-off, immutable journaling, role-based access and MFA, plus a conformity matrix delivered with the installation.
Yes — that is the point. After a two-day training your administrators create and modify agents and flows without ALDIS involvement.
Real-time reporting, forecasts and risk warnings for the CFO — from your own 1C and bank data.
Learn moreLiveContract review, drafting and legal research — on Kazakhstan's legislation, inside your perimeter.
Learn moreDeployedAn omnichannel assistant that answers in Kazakh and Russian — on WhatsApp, Instagram, web and your ERP.
Learn moreDeployedOffline Russian ↔ Kazakh translation of official documents, routing and drafting to state standards.
Learn moreAvailableResume screening, ranking, skills assessment and a candidate assistant — without sending personal data outside.
Learn moreNext agents on the same secure core, sold as add-on modules to existing installations.
A 30-minute demo on your use case: finance, legal, HR or customer service.