SochDB

SochDB is an embedded database that combines SQL, vector search, and context memory into a single engine for AI applications. It solves the problem of managing multiple databases and glue code for LLM apps, agents, and RAG systems. It targets AI/LLM developers, agent builders, RAG system builders, and platform engineers. It is positioned as a local-first, embedded alternative to stitching together separate SQL, vector, and cache databases.

Key features

• Context Query Builder with token budgets
• Hybrid Search: HNSW vectors + BM25 keywords
• Graph Overlay for agent memory
• Embedded-First: ~700KB binary, no dependencies
• Full ACID: MVCC + WAL + Serializable Snapshot Isolation
• Columnar Storage for efficient reads
• SQL support: SELECT, INSERT, UPDATE, DELETE, DDL
• Joins: INNER, LEFT, RIGHT, FULL, CROSS
• Aggregates: COUNT, SUM, AVG, MIN, MAX, MEDIAN, STDDEV
• Policy hooks with pre-built templates
• Tool routing for multi-agent coordination
• Multi-vector documents with chunk-level aggregation