Releases

An embedded Rust graph database for local data-plane workloads

An embedded Rust graph database for local data-plane workloads

Authoring helpers and macros for monochange lint suites

Ecosystem-agnostic manifest lint engine for monochange

An embedded Rust graph database for local data-plane workloads

An embedded Rust graph database for local data-plane workloads

Shared ECMAScript semantic analysis helpers for monochange adapters

An embedded Rust graph database for local data-plane workloads

An embedded Rust graph database for local data-plane workloads

An embedded Rust graph database for local data-plane workloads

The superfluid daemon: durable session store (WAL, verbatim token spans, epochs), generation runtime over Link W, and the native session API

An embedded Rust graph database for local data-plane workloads

Apache Kafka producer sink for the faucet-stream ecosystem

An embedded Rust graph database for local data-plane workloads

A terminal diff viewer for reviewing GitHub PRs: Helix keys, rust-analyzer, inline review comments, and Claude Code reviews

MLX as a superfluid runtime: RuntimePrimitives over mlx-lm through an embedded CPython, for the generic executor, and the runtime's worker executable

Link F session layer: head-daemon and node-agent endpoints — epochs, leases, command idempotency, per-generation watermarks, reconcile (schema §4.2/§4.4)

The mock engine as a superfluid runtime: what the daemon's tests and `superfluid-workerd --engine mock` serve, and the kit's reference adapter

HF tokenizer.json as a superfluid Tokenizer: the daemon's built-in tokenizer for HF-layout artifacts (MLX, safetensors), no runtime FFI

The llama.cpp runtime's tokenizer library: GGUF tokenization through the vocabulary-only load of the llama.cpp beside it, loaded by the daemon at run time

llama.cpp as a superfluid runtime: RuntimePrimitives over the libllama llama.cpp publishes, loaded at run time, for the generic executor; a vocab-only Tokenizer for GGUF artifacts; and the runtime's worker executable

baseRT's native engine as a superfluid runtime: libbaseRT over its tick ABI (superfluid-engine-ffi), and the runtime's worker executable

lingting framework

Draft-7 Engine over the existing baseRT batch C API (migration phase 1: strangler, engine untouched)

Node-agent skeleton: Link W client (tick planning, host sampling, buffer/ring ownership) driving an engine-worker

Engine-worker skeleton: Link W server materializing serialized tick plans into the draft-7 ABI against an Engine (mock now, FFI later)

What every superfluid runtime adapter is built from: the Runtime trait, the worker executable's main, the check report, installing a runtime from its own project, and the conformance checks each adapter runs on itself

PostgreSQL-backed SCIM provisioning and HTTP endpoint Plugin for Lenso.

The generic executor: the Engine tick contract implemented once over a runtime's RuntimePrimitives (batch step, sequences, state export), so an adapter supplies only what its runtime does

IPC frame envelope + Link W / Link F message catalogs and protocol state helpers for superfluid (schema draft 7, §4). Pure protocol crate: no I/O, no sockets.