Configuration

CheckAI can be configured through CLI flags, environment variables, and request-level parameters.

Server Configuration

All server settings are passed as CLI flags to checkai serve:

Setting	CLI Flag	Default	Description
Port	--port	8080	HTTP server port
Host	--host	0.0.0.0	Bind address
Data directory	--data-dir	data	Storage for active/archived games
Opening book	--book-path	—	Polyglot .bin file
Tablebase	--tablebase-path	—	Syzygy tablebase directory
Analysis depth	--analysis-depth	30	Minimum plies for analysis engine
TT size	--tt-size-mb	64	Transposition table memory in MB
Max retained jobs	--analysis-max-jobs	—	Maximum number of completed analysis jobs kept
Max concurrent jobs	--analysis-max-concurrent-jobs	—	Maximum number of analysis jobs run in parallel
Completed-job TTL	--analysis-completed-ttl-secs	—	TTL for completed analysis jobs in seconds (e.g. 86400=24h)

Environment Variables

Variable	Description
CHECKAI_LANG	Override locale (e.g. de, fr, es)
RUST_LOG	Log level (error, warn, info, debug, trace)
CHECKAI_PORT	Port when using Docker Compose

Language / Locale

The locale is resolved in priority order:

1. CLI flag: --lang de
2. Environment variable: CHECKAI_LANG=de
3. System locale: auto-detected via sys-locale
4. Fallback: en

Per-Request Locale (API)

API responses can be localized per request:

# Via query parameter
curl http://localhost:8080/api/games?lang=de

# Via Accept-Language header
curl -H "Accept-Language: de" http://localhost:8080/api/games

Data Directory Structure

data/
├── active/     # Currently running games (binary format)
└── archive/    # Completed games (zstd-compressed)

Games are automatically moved from active/ to archive/ when they end (checkmate, draw, resignation). Archived games are compressed with zstd for efficient storage.

Analysis Configuration

The analysis engine settings control the depth and memory used for game analysis:

Parameter	Min	Recommended	Description
analysis-depth	30	30–40	Deeper = slower but more accurate
tt-size-mb	1	64–256	Larger = fewer transposition collisions

Performance Note

Analysis depth above 35 can take significantly longer per move. A depth of 30 is sufficient for most use cases and provides move classifications (Best through Blunder) with centipawn accuracy.