For BrainRun AI, Webzenz delivered a complete AI platform with multi-model chat, document search (RAG), AI agents, credit billing and AWS infrastructure.
Overview
BrainRun AI is an AI platform that brings several language models, a user's own documents, AI agents and usage-based billing together in one application. Users can work with different models side by side, use their documents as a knowledge source and automate tasks with agents.
Webzenz was responsible for the full technical delivery: architecture, frontend, backend, servers and deployment on AWS. The platform covers AI chat, document processing, knowledge management, AI agents, user accounts, subscriptions, credits and an admin portal.
Starting point
The platform had to combine several language models and data sources without losing control of cost, security or response times. It needed a modular architecture in which individual areas can evolve and scale independently.
Webzenz built specialised backend services behind a central API gateway for this. Each service has a clearly defined task, from chat and document processing to billing.
Features and delivery
Multi-model chat
Chat with several language models, streamed responses and an evaluation pipeline for quality and cost.
Document search with RAG
Text extraction, OCR, chunking, embeddings and semantic search turn a user's own documents into a knowledge source.
Knowledge and context management
A combination of vector search, relational storage and caching for fast, relevant answers.
AI agents
Agents with versioning, scheduling, result management and an estimate of the credits required.
Web search as an extra source
Integrated web search adds current sources to what the models know.
Multilingual interface
Interface in German, English, French and Italian plus a translation service with a multi-level cache.
Subscriptions and credits
Stripe subscriptions combined with an internal credit system that reflects token usage and model costs.
Admin portal and security
Admin portal with roles, sign-in by email, Google and Apple, JWT, role-based access, rate limiting and audit logs.
Monitoring and scheduled jobs
End-to-end monitoring of metrics, logs and traces plus scheduled background jobs.
Webzenz scope of work
- System architecture with microservices and an API gateway
- Frontend development with Next.js and React
- Backend development with Python and FastAPI
- AI integration, RAG pipeline and model connectivity
- Database design with PostgreSQL, Redis and vector databases
- Servers, deployment and operations on AWS
- Quality assurance, testing and monitoring
Technologies used
- Frontend
- Backend and API
- Database and data layer
- AI and models
- Knowledge search and documents
- Payments and billing
- Authentication and security
- Servers, cloud and hosting
- Monitoring and operations
- Quality assurance
Result
BrainRun AI is an AI platform that unites language models, a user's documents, agents and billing in one application. Its modular architecture makes it possible to add new models, features and markets without major rework.
The project shows that Webzenz can deliver complex AI platforms end to end, from the user interface and backend services to the infrastructure on AWS.
BrickMatch.ch: Infomaniak Server Setup and Docker Deployment in Switzerland
Webzenz set up the Infomaniak server environment for BrickMatch.ch and took the Next.js platform live with Docker, staging, GitHub Actions and Swiss hosting.
View project: BrickMatch.ch
APANI Audiophile: Multilingual High-End Audio Website with Next.js and React
Webzenz built a high-end audio website for APANI in 10 languages: Next.js, animated product pages, international SEO, JSON-LD and one codebase for four domains.
View project: APANI Audiophile
Audio Insider Zurich: Custom Registration System for Events and Listening Sessions
Webzenz built a bilingual registration system for events and listening sessions for Audio Insider Zurich: React, PHP API, MySQL, admin dashboard and Hostpoint hosting.
View project: Audio Insider Zürich