Configuration¶

NeuroWeave uses a three-tier configuration system: field defaults → YAML file → environment variables (highest priority).

Configuration Methods¶

Programmatic (recommended for library use)¶

from neuroweave import NeuroWeave

nw = NeuroWeave(
    llm_provider="anthropic",
    llm_api_key="sk-ant-...",
    llm_model="claude-haiku-4-5-20251001",
    enable_visualization=True,
    server_port=8787,
    log_level="INFO",
    log_format="console",
)

YAML File¶

nw = NeuroWeave.from_config("config/default.yaml")

config/default.yaml

llm_provider: "anthropic"
llm_model: "claude-haiku-4-5-20251001"
extraction_enabled: true
extraction_confidence_threshold: 0.3
graph_backend: "memory"
server_host: "127.0.0.1"
server_port: 8787
log_level: "INFO"
log_format: "console"

Environment Variables¶

All fields can be overridden with NEUROWEAVE_ prefixed environment variables:

export NEUROWEAVE_LLM_PROVIDER=mock
export NEUROWEAVE_LLM_API_KEY=sk-ant-...
export NEUROWEAVE_LOG_FORMAT=json
export NEUROWEAVE_SERVER_PORT=9000

The API key is also read from ANTHROPIC_API_KEY for convenience.

All Settings¶

Setting	Type	Default	Description
`llm_provider`	`"anthropic"` \| `"mock"`	`"anthropic"`	LLM backend for extraction
`llm_model`	`str`	`"claude-haiku-4-5-20251001"`	Model identifier
`llm_api_key`	`str`	`""`	API key (required for anthropic provider)
`extraction_enabled`	`bool`	`true`	Enable/disable extraction pipeline
`extraction_confidence_threshold`	`float`	`0.3`	Minimum confidence to store a relation
`graph_backend`	`"memory"`	`"memory"`	Graph storage backend
`server_host`	`str`	`"127.0.0.1"`	Visualization server bind address
`server_port`	`int`	`8787`	Visualization server port
`log_level`	`str`	`"INFO"`	Logging level (DEBUG, INFO, WARNING, ERROR)
`log_format`	`"console"` \| `"json"`	`"console"`	Log output format

Logging¶

NeuroWeave uses structlog for structured logging.

Console mode (development) — colored, human-readable output:

2026-02-17 10:30:15 [info] extraction.complete  entity_count=3 relation_count=2 duration_ms=47.2

JSON mode (production) — machine-parseable output:

{"event": "extraction.complete", "entity_count": 3, "relation_count": 2, "duration_ms": 47.2, "timestamp": "2026-02-17T10:30:15Z"}

Set via log_format="json" or NEUROWEAVE_LOG_FORMAT=json.