Crate synaptic_postgres 

PostgreSQL integration for the Synaptic framework.

This crate provides PostgreSQL-backed implementations of Synaptic traits:

• PgVectorStore — VectorStore using the pgvector extension for cosine-distance similarity search.
• PgStore — Store for key-value storage with JSONB values and optional full-text search via tsvector.
• PgCache — LlmCache for caching LLM responses with optional TTL expiration.
• [PgCheckpointer] — Graph checkpoint persistence (requires checkpointer feature).

Quick start

use sqlx::postgres::PgPoolOptions;
use synaptic_postgres::{PgVectorConfig, PgVectorStore};

let pool = PgPoolOptions::new()
    .max_connections(5)
    .connect("postgres://user:pass@localhost/mydb")
    .await?;

let config = PgVectorConfig::new("documents", 1536);
let store = PgVectorStore::new(pool, config);
store.initialize().await?;

Structs

ChatResponse

A response from a chat model containing the AI message and optional token usage statistics.

Document

A document with content and metadata, used throughout the retrieval pipeline.

Item

A stored item in the key-value store.

PgCache

PostgreSQL-backed implementation of the LlmCache trait.

PgCacheConfig

Configuration for PgCache.

PgStore

PostgreSQL-backed implementation of the Store trait.

PgStoreConfig

Configuration for PgStore.

PgVectorConfig

Configuration for a PgVectorStore table.

PgVectorStore

A VectorStore backed by PostgreSQL with the pgvector extension.

Traits

Embeddings

Trait for embedding text into vectors.

LlmCache

Trait for caching LLM responses.

Store

Persistent key-value store trait for cross-invocation state.

VectorStore

Trait for vector storage backends.