CLI & REPL Reference

AnamDB includes a unified command-line tool anam (compiled from crates/anam-cli) that provides both an interactive REPL shell and a gRPC server.

---

Server Mode

To start the database server, use the serve command. This launches the highly concurrent gRPC server ready to receive remote queries and synchronization changesets:

anam serve --port 8080

Server Command Options

• --port <PORT>: The port to listen on (defaults to 8080).
• --host <IP>: Host address to bind the server to (defaults to 127.0.0.1).
• --gpu: Force enable GPU acceleration if compatible hardware (CUDA, Metal) is found.

---

Interactive REPL

Run anam without arguments to enter the interactive SQL & shell runtime:

$ anam
anam>

Inside the REPL, you can execute standard SQL queries directly or invoke metadata control operations using dot-commands.

---

Command Reference

.ingest

Ingests a raw CSV file and writes it as a versioned, Arrow-backed Lance dataset.

• Syntax: .ingest <csv_path> [lance_path]
• Example:

  anam> .ingest demo/data/transactions.csv demo/data/transactions.lance
  ✓ Ingested 100,000 rows (4.7 MB CSV → Lance columnar)

.load

Loads a Lance table into the current query session catalog.

• Syntax: .load <lance_path> [table_name]
• Example:

  anam> .load demo/data/transactions.lance txns
  Registered table 'txns'

.model load

Registers an ONNX model as an in-process Function-as-Operator (FAO).

• Syntax: .model load <onnx_path> <operator_name> <num_features> <latency_ms> <accuracy>
• Example:

  anam> .model load demo/models/fraud_detector.onnx fraud_detector 3 5.0 0.95
  ✓ Loaded ONNX model 'fraud_detector'

.logic

Compiles and registers a symbolic Datalog rule/constraint.

• Syntax: .logic <rule_name> "<datalog_expression>"
• Example:

  anam> .logic high_risk "fraud_prob > 0.90 AND amount > 10000"
  ✓ Registered rule 'high_risk'

.nl

Translates a natural language prompt into a deterministic Datalog rule via an LLM.

• Syntax: .nl <rule_name> <table_name> <english_description>
• Prerequisites: Requires an OPENAI_API_KEY configured in the .env file.
• Example:

  anam> .nl night_risk txns Flag any transaction between midnight and 5am over $5000
  Compiling NL → Datalog via LLM...
  ✓ Generated and registered rule 'night_risk':
    Datalog: night_risk(X) :- txns(X), X.time >= '00:00', X.time < '05:00', X.amount > 5000.

.models / .operators

Lists all loaded neural model operators registered in the AI-Tables catalog.

• Example:

  anam> .models
  • fraud_detector v1.0.0 [onnx] — 5.0ms, 0.95 accuracy
  • fraud_fast v1.0.0 [onnx]     — 0.5ms, 0.75 accuracy

.rules

Lists all compiled and active symbolic Datalog constraints.

• Example:

  anam> .rules
  • high_risk       ← fraud_prob > 0.90 AND amount > 10000
  • night_risk      ← (LLM-generated Datalog)

.devices

Displays the active device pool and current workload allocations.

• Example:

  anam> .devices
  [ 0–7] CPU-0..CPU-7 (1x) — idle
  [   8] Metal: Apple M2 (10x) — idle

.explain

Prints the complete multi-dimensional reasoning trace for the most recently executed query. This trace shows provenance configuration, compiled logic rules, model statistics, the Pareto frontier, hardware utilization, and any semantic anomalies.

• Example: See Explainability & HITL for details.

.hub

Manages downloading and searching for packages in the Community Hub registry.

• Syntax:
  • .hub search <query>
  • .hub install <package_id>
• Examples:

  anam> .hub search fraud
  anam> .hub install anamdb/financial-compliance@1.0.0

.quit

Closes the interactive shell session.

• Syntax: .quit or Ctrl+D.

---

Executing SQL

You can run standard SQL commands directly. If your queries target tables with Datalog logic or neural operators defined on them, the engine transparently executes model inference and logic checks in-memory.

anam> SELECT region, COUNT(1) AS count FROM txns WHERE fraud_prob > 0.90 GROUP BY region;
+--------+-------+
| region | count |
+--------+-------+
| APAC   | 5321  |
| EU     | 36033 |
| US     | 48018 |
+--------+-------+