What is a semantic DAG?
A semantic DAG is a semantic graph database fused with a hash-linked, Ed25519-signed action history — a store where every link carries meaning and every change is a signed, time-travelable action, so it is neither a plain database nor a blockchain.
Meaning and history in one structure
A DAG — directed acyclic graph — is a set of nodes connected by directed links that never loop back on themselves. A semantic DAG adds two things to that shape. First, it is semantic: every link is a typed, meaningful relationship (a subject, a predicate, an object) rather than an anonymous pointer, so the graph itself records what things mean, not just that they are connected.
Second, it is a history, not only a state. Every change to the graph becomes an append-only action: hashed, cryptographically signed, and chained to the actions that came before it. Nothing is overwritten in place. The current state of the graph is the sum of its signed actions, which means you can replay it to reconstruct exactly how it looked at any point in the past.
This is deliberately not a blockchain. There are no blocks, no mining, no fees, and no global token — the actions are semantic assertions, not transfers of value. And it is more than a database: a database tells you what is true now; a semantic DAG tells you what is true, what it means, and how it came to be that way — all provable.
How a semantic DAG is built
Every fact is a triple
Data is stored as subject → predicate → object statements, so each relationship in the graph carries explicit meaning that both people and machines can query.
Every change is a signed action
Writing to the graph creates an action that is hashed (BLAKE3), signed (Ed25519), and linked to its parent actions — building a tamper-evident causal chain of who wrote what, when, and why.
Nothing is deleted, only superseded
The log is append-only. Removing a fact is itself a signed action, so the full audit trail — and the ability to travel back in time — is preserved by design.
Query by meaning
Because links are semantic, you can query the graph with SPARQL and traverse relationships directly, and reconstruct the graph's state at any past moment.
One structure that remembers and means
AI systems and connected devices increasingly act on data they cannot verify — and traditional stores force a bad trade-off: a database is fast but forgets how it got here, while a blockchain remembers everything but is heavy, expensive, and meaning-blind. A semantic DAG gives you a single structure that is queryable like a graph database, auditable like a ledger, and meaningful like a knowledge graph — without blocks, fees, or mining.
Verifiable by construction
Because every action is signed and chained, the history cannot be quietly rewritten — you can prove what happened.
Time-travel
Replay the signed actions to see the graph exactly as it was at any past moment, for audit, debugging, or provenance.
Meaning, not just data
Typed semantic links let both people and AI reason over relationships instead of guessing at opaque foreign keys.
How Apilium embodies it
AIngle is the first semantic DAG: a native semantic graph database fused with a signed, hash-linked action history, written in Rust. Every write becomes a DagAction — hashed with BLAKE3, signed with Ed25519, chained to its parents — and unsigned actions are rejected at the state-machine level. It exposes the same graph over REST, GraphQL, and SPARQL, and can replay to any point in time. AIngle powers Apilium's products: it is the engine beneath Akashi's grounded, signed knowledge and part of the fabric behind Mayros.
See it in AIngleCommon questions
Is a semantic DAG a blockchain?
No. A blockchain batches transactions into mined blocks secured by global consensus and usually a token. A semantic DAG has no blocks, no mining, and no fees — its entries are meaningful assertions (subject–predicate–object), each individually signed and chained to its parents. It borrows the tamper-evidence of a ledger without the cost and rigidity of a blockchain.
How is it different from a graph database?
A graph database stores relationships and lets you query them, but it typically keeps only the current state — update a value and the old one is gone. A semantic DAG keeps an append-only, signed history of every change, so you can prove how the graph reached its current state and replay it to any past moment.
What does 'directed acyclic' mean?
Directed means each link points one way, from one node to another. Acyclic means those links never form a loop back to where they started. Together they give a structure with a clear direction of history — actions build on earlier actions and never circle back — which is what makes a provable, replayable timeline possible.
What is Ed25519 doing here?
Ed25519 is the signature scheme that signs each action. Before any change is accepted into the graph, it is signed with the writing node's key; unsigned or invalid actions are rejected. That signature is what turns the history from 'trust me' into 'check it yourself'.
See the semantic DAG in AIngle
AIngle is the semantic DAG protocol — a signed, meaningful, time-travelable graph for AI and IoT.