TellusR

ACH for developers

Headless AI platform for search, RAG and assistants

TellusR is a toolkit for building, customising and scaling advanced AI search, generative services and assistant solutions – with a flexible architecture and documented APIs.

Book a demo Contact us
TellusR for developers – API and platform modules

How to get started with TellusR

TellusR Docs API Docs

Run where you want – with models you trust

TellusR can run on the infrastructure you prefer – on a local server or in your chosen cloud. When you adopt TellusR you get access to four main modules:

Import module (Update assistant)

Upload and parse documents, metadata and other content.

Search/retriever module (Flow)

Finds and ranks relevant passages across all content, delivering precise context for search and AI features.

Inference service (NLP service)

Text vectorisation and text generation.

Assistant module (Dialogue)

Generative assistants with a full framework.

The search/retriever module uses compact language models downloaded and run locally with TellusR. This means indexing, vectorisation and search can run without external API calls. The assistant module connects to an external LLM (or several) for text generation, and model choice is fully controlled via configuration.

TellusR is platform-agnostic and designed to avoid vendor lock-in: you are free to choose which language models to use, and can switch providers without changing the solution at its core.

TellusR platform architecture

Quick wins – up and running fast

  • Fast install on your own server. A one-liner to download.
  • A Docker Compose file that brings the solution up.
  • Supports Word, PDF, PowerPoint, HTML and Markdown.
  • Proven default config and prompts, with flexible tuning.
  • End-to-end traceability and observability across the chat pipeline.
  • AI search and AI chat integrated with a built-in AI test framework.

Key features

Platform and architecture

  • Headless and API-first architecture – all functionality integrates directly into your own applications.
  • Highly configurable with full control via documented API endpoints and configuration interfaces.
  • GPU support and scalable execution across multiple servers.
  • Flexible deployment: run locally, on your own server or in any cloud – without vendor lock-in.

RAG, search and context delivery

  • Advanced hybrid search (traditional/semantic).
  • Full control over context, chunking and source prioritisation.
  • Local execution of retriever models.

Assistant framework

  • Support for building your own assistants and static assistant pipelines.
  • Support for MCP (Model Context Protocol).
  • Built-in prompt editor with tooling for testing and QA of prompt changes.
  • Free LLM choice.

Observability and quality assurance

  • Built-in test framework with static tests and LLM-as-a-judge scenarios – easy to add, maintain and run individual tests or full suites.
  • Automated quality monitoring: compare test results across versions, catch regressions early and document quality before production.
  • Full traceability and deep debugging across the pipeline – insight into context, prompts, intermediate steps, helper assistants and generated answers, with drill-down at every step for troubleshooting and testing.

Extra value for consultancies and software teams

  • Standardised APIs mean low vendor lock-in – build your own services on top, extend or swap infrastructure without losing value.
  • Flexible model choice lets you use the best from open source, hyperscalers or your own training, with low switching cost and no change to your codebase.
  • Enterprise-grade security – all data processing and AI execution can happen behind the firewall, on local hardware or in national cloud.
  • An actively used test framework helps you deliver production-ready solutions faster and document quality to your customers.

Secure AI on your terms. Trusted by:

Direktoratet for byggkvalitetAcosRedpill LinproKongsberg kommuneRælingen kommuneNord PoolStiftelsen AstaFritzøe EngrosAllstadEpinovaTeknaHastingsElektroimportørenArtifexAI

Read also

Webinar: AI in your own software without losing control

Morten Krogh-Moe · 18 February 2026

Webinar recording: AI in your own software without losing control

Norwegian software companies are pushed to build in AI without losing control of data and margins. TellusR ran a webinar on how to get to market fast without sacrificing security or profitability.

Read the article →
Digital alternatives exist

Endre Dingsør and Morten Krogh-Moe · 21 September 2025

Digital alternatives exist – if we dare to use them

Norway’s dependence on US technology is undeniable, but real alternatives exist if business and government dare to choose them. We must own the most critical parts of the infrastructure and make the decisive choices now.

Read the article →
The paradox of the AI billion

Endre Dingsør and Morten Krogh-Moe · 22 May 2025

The paradox of the AI billion

Norway’s “AI billion” risks deepening dependence on American technology platforms. The attempt to strengthen national AI competence simultaneously locks us closer to systems we neither own nor control.

Read the article →
New infrastructure for the AI era

Morten Krogh-Moe · 18 November 2025

New infrastructure for the AI era

Norwegian organisations struggle to connect AI safely to yesterday’s infrastructure. TellusR proposes an AI Communications Hub (ACH) as a new control layer — as decisive for AI as CMS was for the internet.

Read the article →

Start your TellusR journey – build smarter, faster and keep full control!

See what is possible in our static API documentation, and read our step-by-step documentation.

Book a demo Docs API Docs