The intelligence layer
Shared infrastructure for intelligence across HLNA Labs — not a chatbot.
01
HLNA AI is the shared intelligence and orchestration layer across the HLNA ecosystem.
Rather than embedding an isolated chatbot into every product, HLNA AI is intended to provide common capabilities for reasoning, retrieval, agents, persistent context and controlled tool use.
Each project still owns its data, its rules and its interface. What it does not have to do is build and maintain its own model integrations, agent framework, retrieval pipeline and safety controls. Those are solved once, carefully, and reused.
Other HLNA products can use the same layer where useful, without each project reinventing its own AI stack.
HLNA AI architecture
02
Four responsibilities, deliberately kept separate. Knowledge is not intelligence, intelligence is not permission to act, and acting is not the same as having done the job correctly.
Knowledge
Structured, sourced knowledge and history — the context intelligence works with.
Intelligence
HLNA AI
Reasoning, retrieval, agents and orchestration over that context.
Controlled execution
Policy-bounded, audited actions when intelligence needs to do something.
Verification
Evidence-based checks that the work was done correctly and within the rules.
Brainbase = knowledge · HLNA AI = intelligence · Relay = controlled execution · Proof = verification
03
What the shared layer is being designed to provide.
Model abstraction
Projects ask for a capability, not a specific vendor's model. Models can change without rewriting products.
Reasoning
Multi-step problem solving over the information a project provides.
Agents
Task-focused agents that can plan and carry out work within defined limits.
Tool calling
Structured access to project functions and external systems.
Retrieval
Finding the right knowledge — from Brainbase or project data — at the right moment.
Persistent context
Context that survives beyond a single conversation or session.
Project-aware intelligence
Each project brings its own data model and rules; the shared layer respects them.
Workflow orchestration
Coordinating multi-step work across models, tools and agents.
Controlled execution
Actions run through Relay-style controls: bounded, logged and verifiable.
Local and cloud models
Designed so local models can sit alongside cloud models where that is practical.
04
HLNA AI is in development as internal infrastructure for HLNA Labs projects. It is not offered as a standalone product.
05
Personal Knowledge
Your personal knowledge and operating system.
Brainbase brings together information scattered across conversations, documents, projects, tasks and digital history, turning it into structured, searchable knowledge while retaining its original source and context.
AI Reliability
Know whether your AI actually followed the rules.
Proof evaluates AI behaviour against explicit requirements and evidence. Relay provides the controlled execution layer around AI actions, helping make autonomous work observable, constrained and verifiable.