Expand description
§AnamDB
The AI-Native, Differentiable Logic Kernel for Autonomous Agents.
AnamDB is a neurosymbolic database engine that natively integrates probabilistic neural inference with deterministic symbolic reasoning. Models are first-class citizens, logic is verifiable, and every query returns a provenance-backed reasoning trace.
§Modules
core— Session API, Arrow schemas, and semiring provenance.model— AI-Tables, Function-as-Operator (FAO) registry, and inference adapters.logic— Differentiable Datalog engine (Scallop) and NL-to-Logic compiler.execution— Extended DataFusion operators, Pareto optimizer, and heterogeneous dispatch.storage— Lance-backed table provider with snapshot versioning.hitl— Human-in-the-Loop semantic monitoring, self-repair, and interactive triage.sdk— Logic Pack SDK for building distributable domain-specific rulesets.
Re-exports§
pub use crate::core::error::AnamError;pub use crate::core::error::Result;pub use crate::core::session::Session;
Modules§
- client
- Rust client SDK for AnamDB — connects to a running AnamDB server.
- core
- Core module: error types, Arrow schemas, semiring provenance, BCNF catalog, and the Session API.
- execution
- Execution engine: custom DataFusion operators, Pareto optimizer, heterogeneous hardware dispatcher, distributed task routing, and 5-stage symbolic integration pipeline.
- hitl
- Human-in-the-Loop (HITL) debugging: semantic monitoring, interactive triage, syntactic self-repair, query result explanation, and global distributed lineage.
- logic
- Logic layer: differentiable Datalog engine and NL-to-Logic compiler.
- model
- Model management: AI-Tables, Function-as-Operator (FAO), inference adapters, and distillation.
- sdk
- Logic Pack SDK — modular, domain-specific neurosymbolic rulesets, and the AI-Tables Community Hub package manager.
- server
- AnamDB server — production wire protocol.
- storage
- Storage engine: Lance-backed table provider with snapshot versioning, streaming scans, write path, persistent catalog, and spatial/audio types.