Model stacking

Model stacking is Achiral's pattern for composing memory, assistants, tools, and model providers into one working system.

Use this section when you need to understand how Achiral shapes model behavior, routes requests, controls tool use, and manages training or fine-tuning workflows.

Pages in this section

• Smart routing: how Achiral routes work across memory, assistants, tools, and model providers.
• AI behavior: communication style, escalation behavior, and learning controls.
• Assistant capabilities: the capability model, registry semantics, executable handlers, and scope.
• EA capabilities: the user-facing controls for enabling and configuring personal EA capabilities.
• Training and fine-tuning: model training jobs, tuning workflows, feedback loops, and versioning.