One box. One process. Structured logs in SQLite — query over HTTP, tail over SSE.

LogLite is a small logging service for a single machine: ingest with POST /logs or built-in harvesters (files, sockets, ZeroMQ), store in your SQLite schema, search and tail without standing up Elasticsearch or a DB server. Ships as pip install loglite (C++ core inside the wheel) or a standalone binary (~5 MB musl build).

• One device, many services — replace a pile of log files with one indexed & queryable store.
• Edge & appliances — gateway, robot, or box that collects logs from nearby services or devices over HTTP.
• Low ops, low RAM — no JVM, single process, low memory footprint; backup is copying a .sqlite file.

Multi-node aggregation, sharding, or enterprise SIEM. If you need Loki, Elastic, Splunk, or ClickHouse-scale search across a fleet, use those tools. LogLite does not federate peers or isolate tenants.

Core is C++20 (Asio/Beast); Python is a thin CLI and harvester layer. Same config.yaml for the wheel or the standalone binary.

pip install loglite          # Python 3.10+
pip install "loglite[zmq]"   # To enable the ZeroMQ harvester

Or grab the standalone C++ binary — see cpp/README.md

config.yaml (minimal):

host: 0.0.0.0
port: 7788
log_table_name: Log
sqlite_dir: ./db
auto_rollout: true

sqlite_params:
   # Any valid SQLite PRAGMA key/value pairs
   auto_vacuum: INCREMENTAL
   journal_mode: WAL  # Highly recommended
   synchronous: NORMAL

migrations:
  - version: 1
    rollout:
      - |
        CREATE TABLE Log (
            id INTEGER PRIMARY KEY,
            timestamp DATETIME NOT NULL,
            message TEXT NOT NULL,
            level TEXT NOT NULL CHECK (level IN ('DEBUG', 'INFO', 'WARNING', 'ERROR', 'CRITICAL')),
            service TEXT NOT NULL,
            extra JSON
        );
      - CREATE INDEX IF NOT EXISTS idx_timestamp ON Log(timestamp);
      - CREATE INDEX IF NOT EXISTS idx_level ON Log(level) WHERE level IN ('WARNING', 'ERROR', 'CRITICAL');
    rollback:
      - DROP INDEX IF EXISTS idx_level;
      - DROP INDEX IF EXISTS idx_timestamp;
      - DROP TABLE IF EXISTS Log;

Run the server:

loglite server run -c config.yaml

Send a log:

curl -X POST http://localhost:7788/logs \
  -H "Content-Type: application/json" \
  -d '{"timestamp":"2026-05-05T12:00:00.123Z","message":"hello","level":"INFO","service":"demo"}'

Tail in real time:

curl -N --output - -H "Accept: text/event-stream" "http://localhost:7788/logs/sse?fields=message,timestamp,level"

Since v1.2.0, LogLite includes an optional web dashboard (see frontend/README.md): live tail, log search, stats, and settings. Use it for observability and debugging —— insights into service performance, availability, and storage. The docker image is published per release.

Full configuration reference, HTTP API, harvester plugin guide, and recipes: loglite.lu-d.com

• Bulk insert with backlog
• Column-based compression for enum-like fields
• Harvester plugin system (file / socket / ZMQ / custom)
• Native C++ core
• /stats endpoint for DB and background-task metrics
• Built-in web UI for browsing logs
• Time-based partitioning (one SQLite file per day or month)

MIT