On-premise AI assistant for OpenText Content Management
tara runs entirely inside your infrastructure. tara is an enterprise AI assistant for OpenText Content Management (OTCM) that deploys on your own Kubernetes. The retrieval engine, the vector & metadata store and your content all stay inside your network, governed by your existing OpenText permissions. The only pluggable piece is the language model — which you can self-host for a fully air-gapped deployment.
Why tara works on-premise
One boundary tells the whole story: the platform and OpenText Content Manager run inside your infrastructure. Only the language model can be external — and it is pluggable.
tara UIExperience layer · same tara, multiple ways to work
Standalone tara App
Full-page chat workspace
tara Widget
Embedded panel inside the OTCM UI
tara Mobile coming next
tara on the go — in the works
tara Platform on-premise · kubernetes
REST API
Integrates tara into any system across your enterprise landscape
rag engineproprietary
Cascade Retrieval
Hybrid content searchMetadata-awareACL filterGrounded context
Vector & Metadata Store
Full-text index & metadata
external · 3rd party
Chat-LLM
Pluggable model provider — bring your own
- Azure AI Foundry
- Amazon Bedrock
- Google Vertex AI
- Self-hosted / custom model
syncReal-time sync keeps the index up to date
Change Observer
Delta Queue
Delta changes from OTCM
Stateless Workers
Queue processors · scale horizontally
integrations3 supported ways to integrate with OTCM
Database Connector
REST API Connector
WebReports Power Pack
ot
source of record
OpenText Content Manager (OTCM)
How your data stays on-premise
Runs on your Kubernetes
tara ships as containers you deploy on your own hardware or private cloud. No tara-hosted SaaS, no outbound dependency — the platform lives entirely inside your infrastructure.
Your content never leaves your network
The RAG engine and the vector & metadata store sit inside the on-premise boundary. Retrieval, indexing and grounding all happen locally — documents stay where they already are.
Governed by your OpenText ACLs
Every answer is filtered by existing OpenText access rights. Users only see content they are already permitted to open — no separate permission model to maintain.
Pluggable, self-hostable LLM
Point the language model at your own Azure, Bedrock or Vertex tenant, or run a self-hosted open model inside the boundary for a fully air-gapped deployment.
Want to see how tara works live?
We offer a focused live demo of tara. We show how it works in real time, explain key capabilities, walk through the vision and answer your questions.
Hi, I'm tara.
Book your live demo