Stable identity
Typed, path-independent object identities and canonical paths separate durable identity from a mutable namespace.
research://synapsefs
SymbioRAP SynapseFS
A deterministic persistence layer for long-lived agents: files, evidence, versions and state remain inspectable even when models, tools and runtime components change.
RESEARCHRESEARCH · ARCHITECTURE IN DEVELOPMENT
research://thesis
Persistence is part of agent intelligence, not an implementation detail.
Long-lived personal AI needs a source of truth that survives process restarts, model swaps and changing retrieval strategies. SynapseFS investigates a storage-neutral core where identity, evidence and temporal state are explicit rather than hidden in transient model context.
focus://synapsefs
Research focus
Immutable evidence & versions
Evidence can remain addressable while newer interpretations or versions supersede earlier state without erasing provenance.
Deterministic recovery
WAL/event-log, checkpoint, snapshot and recovery semantics are researched as first-class correctness mechanisms.
Storage-neutral core
RAM and NVMe/file backends sit behind explicit StorageBackend capabilities so persistence logic is not tied to one vendor or legacy container.
Hardware-aware Fast-Path selection
An Architecture/Execution Policy Layer detects CPU and platform capabilities at startup and selects only among Fast Paths that have already passed benchmark, correctness and crash validation, for example for Intel, AMD or ARM/Neoverse. Unvalidated profiles fall back to the portable baseline path.
architecture://working-model
Current implementation line: small deterministic core, explicit backends, stable contracts.
The current direction uses a C++23 core with typed 128-bit identities, traversal-safe canonical paths, scopes and backend capability contracts. Hot-path full copies, accidental global locks and cache/source-of-truth confusion are treated as architecture risks. WAL, checkpoints, crash recovery and replication are staged work rather than assumed capabilities. At runtime, capability detection maps the observed hardware profile to a validated Execution Policy. Selection is reproducible and evidence-driven: only Fast Paths confirmed for the concrete architecture are enabled; otherwise the portable baseline remains active.
Namespace & IDs
Stable object identity, canonical paths, scopes and deterministic traversal semantics.
Evidence & versions
Immutable content/evidence, explicit supersession and auditable version relationships.
Transaction & recovery path
Event/WAL semantics, checkpoints, snapshots and deterministic restart behavior.
StorageBackend
Capability-based adapters for memory, local storage and later high-performance persistent backends.
Architecture / Execution Policy
CPU and platform capabilities are detected at startup and mapped to a previously validated execution path. Intel, AMD, ARM/Neoverse and future paths remain independently benchmarkable and correctness-testable.
research://questions
Open research questions
Which identity and version semantics remain stable across renames, moves, deduplication and compaction?
How can copy-on-write, chunking, hashing and indexing reduce write amplification without making recovery opaque?
Which concurrency model minimizes locking while preserving deterministic state transitions?
How should persistence expose temporal and provenance primitives to the memory layer without leaking backend details?
How should hardware profiles, benchmark evidence, correctness/crash tests and fallback policies be versioned so automatic Fast-Path selection remains reproducible and safe?
product://symbiofs
From research to a general-purpose storage core
SymbioFS productizes the persistence and storage concepts as a high-performance, versioned, content-addressed and crash-consistent core for AI state, object storage, event streams, edge/local-first and other systems.
status://research
SymbioRAP Research
The points shown describe research and implementation directions, not guarantees of finished product functionality.
connect://research
Contribute to the research architecture
For research, funding and technology partners, we share architecture questions, benchmark design and technical work states in qualified collaboration.
Next useful step
Tell us your idea – a few bullet points are enough.
You do not need to define the technical solution. Describe what you want to build or improve and we will map feasibility, a sensible first scope and next steps.
Get a free idea assessment
Four fields are enough. You will receive a concrete response on how to start small and extend the project later.