Implementation:Protectai Llm guard Config

Knowledge Sources	LLM Guard
Domains	Configuration, API_Design, DevOps
Last Updated	2026-02-14 12:00 GMT

Overview

Concrete tool for loading and validating YAML-based scanner pipeline configuration using Pydantic models, provided by the LLM Guard API.

Description

The Config class and get_config function load a YAML configuration file, substitute environment variables, and validate the schema using Pydantic. The configuration defines input/output scanner pipelines, authentication, rate limiting, and observability settings.

Usage

Used internally by create_app to initialize the API server. Users author the YAML file; the system loads and validates it at startup.

Code Reference

Source Location

• Repository: llm-guard
• File: llm_guard_api/app/config.py
• Lines: L52-88

Signature

class Config(BaseModel):
    input_scanners: List[ScannerConfig] = Field()
    output_scanners: List[ScannerConfig] = Field()
    rate_limit: RateLimitConfig = Field(default_factory=RateLimitConfig)
    auth: Optional[AuthConfig] = Field(default=None)
    app: AppConfig = Field(default_factory=AppConfig)
    tracing: Optional[TracingConfig] = Field(default=None)
    metrics: Optional[MetricsConfig] = Field(default=None)

def get_config(file_name: str) -> Optional[Config]:
    """Load and validate YAML config file."""

Import

from llm_guard_api.app.config import Config, get_config

I/O Contract

Inputs

Name	Type	Required	Description
file_name	str	Yes	Path to YAML configuration file

Outputs

Name	Type	Description
Config	Config or None	Validated configuration object, or None if loading fails

Usage Examples

YAML Configuration File

app:
  name: "LLM Guard API"
  log_level: "INFO"
  scan_fail_fast: false
  scan_prompt_timeout: 10
  scan_output_timeout: 30
  lazy_load: false

auth:
  type: http_bearer
  token: "${API_TOKEN:default-token}"

input_scanners:
  - type: PromptInjection
    params:
      threshold: 0.92
      use_onnx: true
  - type: Toxicity
    params:
      threshold: 0.5

output_scanners:
  - type: NoRefusal
    params:
      threshold: 0.75
  - type: Relevance
    params:
      threshold: 0.5

Related Pages

Implements Principle

• Principle:Protectai_Llm_guard_Scanner_Configuration_Schema

Requires Environment

• Environment:Protectai_Llm_guard_API_Server_Deployment