- KV path-addressed instruction model (
[i,0]opcode,[i,-j]reads,[i,+j]writes) - typed values: int, float, bool, string, bytes, tensor
- builtin operators: arith, compare, logic, bitwise, cast, call, string, I/O
- single-layer IR — same syntax is VM instruction, high-level language, and compiler IR
- control flow: if/else, while (break/continue), for-in
- lower pass: if/while → block + br/goto
- TCO (tail-call optimization) for deep recursion
- multi-return values + discard (
_) - vthread coroutine model with concurrent workers
- soft links (path redirection)
- serve daemon with Redis persistence
- pipe / inline (
-c) / file execution modes -
kvspaceCLI: get, set, del, list, tree, clear, watch, notify -
vet(syntax check) +format(source formatter) - 87 tutorial examples: basics, functions, control flow, algorithms, LeetCode
- CI: build (Linux + macOS), tutorial tests (Redis), cross-compile releases
- tensor lifecycle ops: tensor.new / tensor.del / tensor.clone (heap-plat backed)
- tensor compute dispatch via kvspace op-plat routing
- kvspace-cpp client library for C++ backends (✅ created)
- GPU tensor ops (Triton / CUDA integration)
- multi-node vthread scheduling
- kvspace sharding
- RDMA backend for kvspace
- cluster mode (
kvlang serve --cluster)
- self-hosting compiler (kvlang → kvlang)
- WASM compile target
- foreign function interface (FFI)
- benchmarking suite
- playground / online demo (maybe never)
- improved error messages with source locations (maybe never)
-
go installsupport (maybe never) - package manager (maybe never)
- standard library: http, json, file I/O (maybe never)