Dual-form memory
SQLite provides indexed access while Markdown remains a human-readable source representation.
A runnable reference implementation where memory, routing, policy, tools, workflows and orchestration remain explicit engineering boundaries.
The problem
Personal AI systems become difficult to trust when memory, provider routing, tools and workflows are implicit. This runtime makes those boundaries explicit and testable.
Verified capabilities
These capability descriptions are intentionally limited to the evidence-backed implementation surface.
SQLite provides indexed access while Markdown remains a human-readable source representation.
Chat, reasoning, embedding and vision workloads route through provider health checks.
A local-only route is rejected before a cloud provider adapter is invoked.
Registered capabilities are checked against execution context before invocation.
Lifecycle state is persisted so workflows can resume across process boundaries.
Idempotency keys, bounded concurrency and retry classification make jobs explicit.
Architecture
The diagram below represents the current verified architecture boundary; planned or environment-dependent components are described in the engineering notes rather than silently promoted.
Deterministic IDs and hashes connect database memory to its human-readable representation.
Engineering deep dive
Technology names matter less than the engineering contracts they enforce.
Deterministic IDs and hashes connect database memory to its human-readable representation.
Policy is evaluated before provider invocation, not after the request has already crossed the boundary.
Every job has an idempotency key and retries are limited to retryable failures.
The provider port allows local, hosted or BYOK adapters without changing application policy.
Project status
Runnable reference implementation with memory, routing, local-only policy, tools, workflows and orchestration covered by tests.
Public implementation — source and test suite are available.