Struct Session 

pub struct Session {
    pub config: SessionConfig,
    /* private fields */
}

Primary entry-point for all AnamDB operations.

Fields

config: SessionConfig

Session-level config.

Implementations

impl Session

pub async fn new() -> Result<Self>

Create a new session with default settings (CPU-only).

pub async fn new_with_npu() -> Result<Self>

Create a new session with NPU / GPU acceleration enabled.

pub async fn with_config(config: SessionConfig) -> Result<Self>

Create a session from an explicit SessionConfig.

pub async fn register_table(&self, name: &str, path: &str) -> Result<()>

Register an existing Lance dataset as a queryable table.

Uses the streaming Lance provider for push-down projection and filter support (no eager memory loading).

pub async fn insert( &self, _table_name: &str, lance_path: &str, batches: Vec<RecordBatch>, schema: Arc<Schema>, ) -> Result<WriteResult>

Insert rows into a registered Lance table.

Appends the given batches to the underlying Lance dataset.

pub async fn delete( &self, _table_name: &str, lance_path: &str, predicate: &str, ) -> Result<WriteResult>

Delete rows from a registered Lance table matching a predicate.

pub async fn register_logic_from_nl( &self, name: &str, table: &str, nl_description: &str, ) -> Result<()>

Define a logical constraint from natural language.

pub fn register_logic(&self, name: &str, datalog: &str) -> Result<()>

Register a raw Datalog rule directly.

pub async fn sql(&self, query: &str) -> Result<QueryResult>

Execute a neurosymbolic SQL query.

pub async fn refine_query(&self, correction: &str) -> Result<QueryResult>

Refine a query after human feedback on an anomaly.

pub fn models(&self) -> &ModelRegistry

Access the model registry for AI-Table operations.

pub fn logic_engine(&self) -> &Arc<RwLock<LogicEngine>>

Access the logic engine.

pub fn device_pool(&self) -> &DevicePool

Access the heterogeneous device pool.

pub fn load_onnx_model( &self, name: &str, model_path: &str, function_id: &str, num_input_features: usize, ) -> Result<String>

Load an ONNX model from disk and register it as both an AI-Table entry and an FAO operator.

pub fn load_onnx_model_with_metrics( &self, name: &str, version: &str, model_path: &str, function_id: &str, num_input_features: usize, avg_latency_ms: f64, accuracy: f64, ) -> Result<String>

Load an ONNX model with custom performance metrics.

Use this to register multiple model variants with different latency/accuracy trade-offs for Pareto optimization.

pub fn load_logic_pack(&self, pack: &LogicPack) -> Result<String>

Load a Logic Pack into this session.

Registers all bundled rules and models from the pack.

pub fn explain_query( &self, batches: &[RecordBatch], level: ExplainLevel, ) -> Result<QueryExplanation>

Generate an explanation of the last query results.

• Coarse: High-level summary (rules, models, stats, hardware).
• Fine: Per-row provenance trace with source record lineage.

pub fn self_repair( &self, error_msg: &str, operator_name: &str, context: &str, ) -> Result<RepairReport>

Trigger the syntactic self-repair agent for a failed operation.