The Knowledge Image is the mountable artifact. Evidence is lexical, spanned, and receipted. V5.1.0 extends the same file into a transactional runtime.
Knolo contract
V5.1.0 ships. V4 stays compatible.
We build AI systems on Knolo, a verifiable knowledge runtime for agents. V5.1.0 ships on npm as @knolo/core: a verifiable Knowledge Image and transactional runtime — one ledger for knowledge, policy, memory, authority, and agent runs. V4 Knowledge Images remain the compatible prior line: portable .knolo files with deterministic lexical evidence, receipts, and local execution.
Compatible prior line
V4 Knowledge Images
Portable .knolo files with deterministic lexical evidence, receipts, and local execution. V4 pack and retrieval APIs remain unchanged. Open-source contract lives in knolo-core.
Published on npm as @knolo/core 5.1.0. The same .knolo file becomes a transactional runtime: one ledger for knowledge, policy, memory, authority, and agent runs.
V5.1.0 · @KNOLO/CORE · SAME .KNOLO FILE
Builder-facing layers
Image, evidence, runtime. Models are replaceable.
01
Knowledge Image (.knolo)
The mountable artifact. A portable file you can hash, ship, revoke, and run locally. V4 ships this today. It is not a pile of embeddings and not a chat memory.
02
Evidence path
Lexical grounding, spans, plan hash, read receipt. The same question against the same image under the same policy is replayable. Approximate search, if used, sits behind that contract — it is not the trust root.
03
V5 runtime path
State roots, write receipts, run receipts, default-deny authority, resumable runs. V5.1.0 ships on npm as @knolo/core. Same file, transactional runtime.
The artifact
A file you can mount.
Vector search is a useful index. It is not a knowledge runtime. Cosine distance is not an audit trail. When counsel asks why a passage appeared, the answer is a span, a plan hash, and a read receipt — not a nearest-neighbor vibe.
Hive Forensics AI implements that runtime inside the systems we ship — on-prem or in a VPC — rather than bolting a chat UI onto someone else’s index.