Storage

Overview

Seerflow uses Protocol-based storage interfaces. Backends are swappable via a single config line:

storage:
  backend: sqlite      # or postgresql

SQLite (Default)

Zero-config storage with WAL mode for concurrent reads during writes.

Features

• Auto-created on first run (~/.local/share/seerflow/seerflow.db)
• WAL mode with optimized PRAGMAs (64 MiB cache, 256 MiB mmap)
• FTS5 full-text search on log messages
• Batch writes via WriteBuffer (1000 events or 100ms)
• ~30K events/sec batched write throughput

Schema

Four tables created automatically:

• events — indexed columns + msgpack BLOB
• entity_events — junction table for entity-to-event queries
• alerts — with dedup key UNIQUE index
• model_state — key-value store for ML model checkpoints

Configuration

storage:
  backend: sqlite
  data_dir: ~/.local/share/seerflow    # default
  sqlite_path: /custom/path/seerflow.db  # override

PostgreSQL (Production)

For production deployments with higher concurrency requirements.

storage:
  backend: postgresql
  postgresql_url: ${DATABASE_URL:-postgresql://localhost/seerflow}

PostgreSQL backend is planned for Sprint 13.