Model routing
In pilotRoute each step to a model or to plain code based on task type, cost and data sensitivity. Running in our own engineering operations.
Lamiak Core
The system underneath every LAMIAK product.
Lamiak Core is the layer under every agent: it assembles context, routes each step, holds actions for approval, and measures what happened. Parts of it are built and tested today; every part below carries its current status.
The loop
Understand the work.Act with control. Learn from outcomes.
01 Context
What the system knows: records, documents, permissions, history.
02 Action
What an agent or workflow does, inside the rules you set.
03 Outcome
What actually happened, recorded with its evidence.
04 Learning
What changes next time, promoted only after review.
The loop, today and next
Five stages are built today. Four come next.
Context, approval, routing, evaluation and feedback are built in part and exercised in testing; each shows its status. Dataset snapshots, a unified model registry, promotion gates and training from outcomes are planned and come next. No stage is offered to customers yet, and each product page shows its own status.
Today
Context
In developmentGather the records and permissions a task needs into one packet before an agent acts.
Approval
Human approvalIn developmentHold selected actions for a person, and record who decided and when.
Routing
In pilotSend each step to the right model for the task and the sensitivity of the data.
Evaluation
In developmentScore agents against suites and graders before trusting a change.
Feedback
In developmentCapture edits, rejections and outcomes from the people doing the work.
Next
Dataset snapshots
PlannedFreeze feedback and outcome data into versions a decision can cite.
Model registry
PlannedOne identity for every model, adapter and prompt, with its evaluation history.
Promotion gates
PlannedA candidate proves better on evaluation before it is given authority.
Training from outcomes
PlannedAdapt models on outcome data a customer has agreed to share.
Next is where the loop closes: outcomes become data, data becomes a candidate, and a candidate earns its place on evaluation before it is given authority.
The five parts
Each part, and how far along it is.
One page per part. Each separates what is built today from what is designed, and carries the status of the registry entries it rests on.
Context
What the system knows before it acts.
Assembling the records, documents and permissions a task needs into one packet, before an agent is allowed to do anything with them.
TodayIn developmentAgents
Agents that propose, inside limits you set.
Bounded agents that draft and propose rather than act, with each step sent to a model or to plain code depending on what the step is.
TodayIn developmentWorkflows and approvals
Steps that run the same way, and the person who decides.
Ordinary repeatable steps, with a person approving where the policy for that workflow says so, and each decision recorded with who made it and when.
TodayIn developmentEvaluation and learning
Measure first, then decide what changes.
Evaluation suites and graders run by hand today. Scorecards, promotion gates, dataset snapshots and training from outcomes are designed and not built.
TodayIn developmentDeployment and governance
Where your records sit, and what governs a change.
How records are kept apart per organisation, which outside services receive data and why, and what would have to be true before something changes.
TodayIn developmentArchitecture
Products on top. One foundation beneath.
Each product is built for its own customers. The parts that decide what an agent may see, do and learn are meant to be shared, so they can be built and checked once; today some still exist in more than one version.
Products
Workspaces and journeys built for a specific kind of work.
- Andiamo SMB
- Andiamo Enterprise
- Hordago Labs
- BioContext7
Workflows and approvals
Deterministic steps, with people approving where policy requires it.
- Approval queues
- Execution records
Agents and routing
Agents do bounded tasks; each step goes to a model or to plain code.
- Agents
- Model routing
Context
The records, documents and permissions a task needs, assembled for review.
- Context assembly
Evaluation and learning
Scorecards, feedback and outcomes that decide what changes next.
- Evaluation
- Feedback
- Promotion gates
Status
Every part, with its status.
The whole Lamiak Core registry in one place, sorted by how far along each entry is, so the list can be read without opening five pages. Some parts are built and tested inside our own systems, and some inside Andiamo. None is yet offered to customers; each product page shows its own status.
Approvals and decision records
In developmentSelected actions wait for an owner or admin to approve or reject them, and each decision is recorded with who decided and when; not yet switched on in a live product.
Human approval
Context assembly
In developmentGather the records and permissions an agent needs for a task into one packet before it acts; built for selected workflows, and packets are not yet saved for review.
Evaluation
In developmentEvaluation suites and graders exist for some agents and are run by hand; automatic scorecards and release gates are still being built.
Feedback capture
In developmentRecord how people respond to AI suggestions, including edits and rejections, and whether they found them useful; captured in a few places today, not yet as one feedback stream.
Separation between businesses
In developmentEach business's records are tagged to its organization, and database access rules limit staff to their own organization's data; known gaps in this separation are still being closed.
Dataset snapshots
PlannedVersioned snapshots of the feedback and outcome data used for training and evaluation.
Model registry
PlannedA registry of models, adapters and prompts with their evaluation history.
Model promotion gates
PlannedA candidate must prove better on evaluations before it receives authority in production.
Training from outcomes
PlannedAdapt models on outcome data a customer has agreed to share.
Start with one workflow that matters.
We map the work, the systems it touches and the approvals it needs before anything is automated.