Expand description
plugmem engine core.
The no_std + alloc, single-threaded memory engine on top of
plugmem_arena: the data model (facts, entities, edges,
bitemporality), the indexes (BM25, temporal, graph, vectors), hybrid
recall, and the snapshot/journal machinery. The core owns no I/O, no
clock and no threads — timestamps arrive as parameters, bytes leave
through the Storage trait.
Design: ..05. Implementation lands in stages 2-4; modules
appear here in dependency order as those stages complete.
Re-exports§
pub use config::Config;pub use error::Error;pub use id::EntityId;pub use id::FactId;pub use id::NONE_U32;pub use memory::FactFault;pub use memory::FactView;pub use memory::LinkInput;pub use memory::MaintainReport;pub use memory::Memory;pub use memory::OpenReport;pub use memory::RecallQuery;pub use memory::RecallResult;pub use memory::RecallScratch;pub use memory::RecalledEdge;pub use memory::RecalledFact;pub use memory::RememberInput;pub use memory::RememberOutcome;pub use memory::Similar;pub use memory::SimilarReason;pub use memory::Stats;pub use model::EdgeSlot;pub use model::EntityByName;pub use model::EntityRecord;pub use model::FactAux;pub use model::FactRecord;pub use model::TemporalSlot;pub use model::VALID_TO_OPEN;pub use model::fact_flags;pub use storage::MemScratch;pub use storage::MemStorage;pub use storage::Scratch;pub use storage::Storage;
Modules§
- config
- Engine configuration.
- error
- Engine error type.
- id
- Typed identifiers.
- index
- The recall indexes: shared posting storage, the BM25 lexical index, and plain id-list indexes (tag → facts, entity → facts) with sorted-list intersection.
- journal
- Journal record framing.
- memory
- The engine: data model state, verbs, journaling and replay (/03, /05).
- model
- Record layouts of the data model.
- snapshot
- Snapshot container: the file format’s header, section table and checksums.
- storage
- The
Storagetrait and the in-memory reference implementation - tokenizer
- The core tokenizer, v2.