The proprietary foundation
underneath DataInbox and APPSS.
Since 2014, Datastreams has developed declarative technology for data, events, state and applications: Dimml, event storage, CQRS, projections, rules, processing and restore in one runtime model.
Technology and IP layer — proof that the products are not wrappers around an LLM.
Declarative by design
Describe, execute and control the operation in one model.
Business complexity can grow without changing the shape of the configuration or adding a separate control plane for each requirement.
Describe
Define source, contract, context, rules, action and evidence as one versioned stream configuration.
Execute
Run validation, state, decisions and delivery inside one high-throughput runtime boundary.
Control
Validate, approve, inspect and recover changes through the same lifecycle for people and approved agents.
AI-supported authoring
From a business description to a tested runtime configuration.
AI works with the declarative model instead of generating an invisible production pipeline. Every proposal remains readable, comparable and subject to the same validation and authority gates.
- 01
Describe
A person or approved AI assistant describes the source, business meaning, context, rules and required outcome.
- 02
Model
The declarative model language turns that intent into an explicit, versioned configuration proposal.
- 03
Test
Contracts, references, permissions and representative events are validated before the live operation changes.
- 04
Approve
The responsible person or policy gate reviews the diff, processing purpose, legal basis and permitted actions.
- 05
Publish
Only an approved version is activated on the runtime, with its configuration and authority evidence retained.
AI is the model assistant, not the production authority.
An approved assistant can describe, generate, explain, validate and test configuration proposals. Identity, policy, review and approval still determine whether a version may become active.
One runtime, many operations
Hundreds of recipes can remain independently understandable.
A runtime can host many separately versioned stream configurations, subject to the agreed workload and capacity. Each recipe uses the same model language and control lifecycle without becoming a separate pipeline estate.
Purpose and legal basis
See which declared processing purpose and legal basis apply to a personal-data operation. The accountable organisation supplies and approves that basis; AI does not invent it.
Processing behaviour
Inspect accepted, rejected and recovered events, processing time, active state, retries and downstream delivery per configuration.
Decision evidence
Trace the active version, input context, policy result, actor or agent authority, action and resulting destination.
recipe → active version → processing purpose → legal basis → policy result → action → evidence
This makes the operation auditable while it runs, rather than requiring teams to reconstruct its meaning from source code, cloud consoles and logs afterwards.
Technology pages
Explore each capability without losing the complete picture.
Every page returns to the same technology foundation underneath DataInbox and APPSS.
The technology matters because live intelligence must remain explainable and under business control.
Start with the question, connected channels and required outcome. Use the technology detail to validate how processing, policy and evidence remain independent underneath it.
Explore Streaming Intelligence