Configure Guardrails

Detailed configuration options for Dome Guards and Detectors.

Dome’s configuration system lets you precisely control which Guards run, how they execute, and what Detectors they use. This guide covers all configuration options.

Configuration Hierarchy

Dome organizes protection in three levels:

Each level has its own configuration options that can be customized.

Guardrail Configuration

Guardrail Options

Option	Type	Default	Description
input-guards	List	[]	Guards to run on input
output-guards	List	[]	Guards to run on output
input-early-exit	Boolean	True	Stop on first input flag
input-run-parallel	Boolean	False	Run input Guards in parallel
output-early-exit	Boolean	True	Stop on first output flag
output-run-parallel	Boolean	False	Run output Guards in parallel

Execution Modes

• Early Exit (default): Stops processing when the first Guard flags content. Faster for rejecting clearly malicious input.

• Complete Execution: Runs all Guards regardless of flags. Useful for comprehensive logging.

Parallel Execution: Runs Guards simultaneously for lower latency.

Guard Configuration

Guard Types

Type	Use Case	Available Detectors
security	Adversarial attacks	Prompt injection, encoding detection
moderation	Harmful content	Toxicity, profanity, hate speech
privacy	Sensitive data	PII detection, secrets
integrity	Data quality	Format validation (experimental)
generic	Custom logic	User-defined Detectors

Guard Options

Option	Type	Default	Description
type	String	Required	Guard category
methods	List	Required	Detectors to use
early-exit	Boolean	True	Stop on first Detector flag
run-parallel	Boolean	False	Run Detectors in parallel
blocked-response	String	Default	Custom block message

Detector Configuration

Here are some of the available Detectors:

Security Detectors:

Detector	Description	Options
prompt-injection-mbert	Multilingual BERT model	threshold
prompt-injection-deberta-v3-base	DeBERTa v3 model	threshold
encoding-heuristics	Base64, Unicode tricks	None
security-embeddings	Semantic similarity	threshold, top_k
security-llm	LLM-based detection	model_name

Moderation Detectors:

Detector	Description	Options
moderation-flashtext	Fast keyword matching	wordlist
moderation-deberta	Neural toxicity classifier	threshold
moderations-oai-api	OpenAI Moderation API	None
moderation-llamaguard	Llama Guard model	threshold

Privacy Detectors:

Detector	Description	Options
privacy-presidio	PII entity recognition	entities, threshold
detect-secrets	Credential detection	None

Work in Progress

The programmatic protection capabilities and Dome integrations are currently in private preview and subject to change.

Next Steps

Use Guardrails

Runtime integration patterns

Custom Detectors

Build your own Detectors

Observability

Monitoring and tracing