Configuration

VectorWave provides robust and type-safe configuration management based on pydantic-settings. All settings can be controlled via a .env file or system environment variables.

1. Connection & Schema

Defines the Weaviate database connection and the collection (Table) names where data will be stored.

Variable Name	Description	Default Value
WEAVIATE_HOST	Weaviate instance host address	localhost
WEAVIATE_PORT	HTTP port (REST API)	8080
WEAVIATE_GRPC_PORT	gRPC port (for bulk data transfer)	50051
COLLECTION_NAME	Collection name for storing function metadata (static info)	VectorWaveFunctions
EXECUTION_COLLECTION_NAME	Collection name for storing execution logs (dynamic info)	VectorWaveExecutions
GOLDEN_COLLECTION_NAME	Collection name for storing golden datasets for testing	VectorWaveGoldenDataset
IS_VECTORIZE_COLLECTION_NAME	Whether to store function definitions (static data) (If False, only logs are stored)	True

2. AI & Vectorizer Strategy

Sets the engine and model for embedding (Vectorizing) data.

Strategy Selection (VECTORIZER)

Value	Description
huggingface	(Default) Uses local sentence-transformers models. Free & secure.
openai_client	Uses OpenAI API. High accuracy, costs apply.
weaviate_module	Delegates to Weaviate's internal modules (text2vec-*).
none	Disables vectorization (storage only).

Detailed Settings

Variable Name	Description	Default Value
HF_MODEL_NAME	HuggingFace model name (used in local mode)	sentence-transformers/all-MiniLM-L6-v2
OPENAI_API_KEY	OpenAI API key (sk-...)	None
WEAVIATE_VECTORIZER_MODULE	Module name to use in Weaviate module mode	text2vec-openai
WEAVIATE_GENERATIVE_MODULE	Weaviate generative module name (for RAG)	generative-openai

3. Performance & Batching

Balances real-time processing and bulk data processing. VectorWave uses Async Batching by default.

Variable Name	Description	Default Value
BATCH_THRESHOLD	Batch buffer size. Sends to DB when this count is reached.	20
FLUSH_INTERVAL_SECONDS	Forcibly sends if this time passes even if the buffer isn't full.	2.0 (seconds)

Tip: In high-traffic production environments, increase BATCH_THRESHOLD to 100 or more to maximize throughput.

4. Security & Data Masking

Automatically masks sensitive personal information or secret keys to prevent them from being stored in the Vector DB.

Variable Name	Description	Default Value
SENSITIVE_FIELD_NAMES	List of keywords to mask (comma separated). Arguments containing these keywords are replaced with [MASKED].	password,api_key,token,secret,auth_token

# Example: password argument is automatically masked and stored
@vectorize
def login(username, password): ... 

5. Monitoring & Alerting

Sends external notifications when an error occurs.

Variable Name	Description	Default Value
ALERTER_STRATEGY	Alert strategy (webhook, log, none)	none
ALERTER_WEBHOOK_URL	Slack/Discord webhook URL	None
ALERTER_MIN_LEVEL	Minimum log level to trigger alerts (INFO, WARNING, ERROR)	ERROR

6. Advanced Analytics

Settings for detecting data changes or for recommendation systems.

📉 Data Drift Detection

Detects when the distribution of input data deviates from training (or past) data.

Variable Name	Description	Default Value
DRIFT_DETECTION_ENABLED	Whether to enable drift detection	False
DRIFT_DISTANCE_THRESHOLD	Vector distance threshold to determine drift (0~2)	0.25
DRIFT_NEIGHBOR_AMOUNT	Number of recent neighbors to compare	5

🎯 Recommendation Margins

Adjusts the search range for similarity search.

Variable Name	Description	Default Value
RECOMMENDATION_STEADY_MARGIN	Stable (similar) recommendation range	0.05
RECOMMENDATION_DISCOVERY_MARGIN	New (diverse) recommendation discovery range	0.15

7. File Path Settings

Variable Name	Description	Default Value
CUSTOM_PROPERTIES_FILE_PATH	Path to custom metadata schema file	.weaviate_properties
FAILURE_MAPPING_FILE_PATH	Path to error code mapping file	.vectorwave_errors.json