Skip to content

SQL_MODEL.md

Latest commit

 

History

History
85 lines (64 loc) · 2.2 KB

SQL_MODEL.md

File metadata and controls

85 lines (64 loc) · 2.2 KB

Relational SQL & Vector Search (⛩️)

ToriDB provides a high-performance relational engine that combines classic SQL integrity with modern vector storage capabilities.

1. Schema Enforcement

ToriDB is schema-first for relational tables. You must define a table before inserting data.

1.1 Supported Data Types

Type Description Example
int 64-bit signed integer 42
string UTF-8 UTF-8 string "Hello"
float 64-bit floating point 3.14
bool Boolean (true/false) true
datetime ISO8601 or Timestamp "2024-01-01"
blob Binary data (Base64) "SGVsbG8="
vector Float array embedding [0.1, 0.2, ...]
json Native JSON Document '{"key": "val"}'

1.2 Table Definition (SQL)

CREATE TABLE users (
    id:int:pk,
    email:string,
    profile:json,
    features:vector
)
  • :pk marks a column as Primary Key.
  • :fk(table.col) marks a Foreign Key relationship.

2. Querying Data

2.1 Standard Select

SELECT name, email FROM users WHERE age >= 18 AND status = "active"

2.2 Aggregates & Grouping

ToriDB supports real-time aggregation over in-memory sets, including joined tables.

  • Selectors: COUNT(*), SUM(col), AVG(col), MAX(col), MIN(col).
  • Grouping: GROUP BY col1, col2.
  • Filtering: HAVING count(*) > 5.

2.3 Table Joins

Efficient in-memory joins using Hash-Join implementation.

SELECT orders.id, users.email 
FROM orders 
JOIN users ON orders.user_id = users.id

3. Vector Similarity Search

Perform K-Nearest Neighbor (KNN) searches across high-dimensional vectors.

3.1 Syntax

SEARCH <table> <column> <pivot_vector> <limit>

SEARCH products embedding [0.12, 0.45, 0.22] 5

3.2 Details

  • Similarity Metric: Cosine Similarity.
  • Normalization: Vectors are auto-normalized for consistent similarity scoring.
  • Performance: Calculated parallelly across the worker pool.

4. Hybrid JSON Pathing

Query inside json columns using the Arrow Operator (->).

SELECT email FROM users WHERE profile->settings->theme = "dark"

Back to Document Index