Displaying 1 to 21 of 21 repositories
LLM Guard injection detection model (Python). ONNX-based dual-pass classifier detecting prompt injec
1m
1.3K
Presidio PII detection microservice. Provides named entity recognition and pattern-based PII detecti
1m
955
Semantic similarity detection using embeddings and vector store. Detects prompt injection via semant
1m
1.2K
Language detection service using fastText. Detects input language and applies language-specific dete
1m
905
Qdrant vector store persistence container. Maintains embedding vectors for semantic similarity detec
1m
1.5K
Content safety worker for Polish. Detects harmful content using specialized Polish language model.
1m
870
Content safety worker for English. Detects toxic and harmful content using multilingual content mode
1m
841
Scope drift detection worker (Python). Uses semantic embeddings to classify conversation drift from
1m
996
Qdrant vector store initialization. Pre-seeds vector embeddings for semantic detection.
1m
1.4K
Scope drift detection orchestrator. Coordinates semantic analysis to classify prompts as on-topic, n
1m
10K+
Decision arbiter combining signals from multiple detection workers. Applies weighted scoring and bus
1m
828
Backend API for configuration interface. Manages settings, rule sets, policies, and monitoring dashb
1m
915
Public REST API (Express.js) serving detection requests. Handles authentication, rate limiting, mess
1m
979
PII detection worker. Identifies sensitive data (names, emails, credit cards, PESEL, NIP) in prompts
1m
1.1K
Web-based configuration and investigation interface for managing Vigil Guard deployment and reviewin
1m
782
Heuristics-based detection worker. Analyzes prompts using pattern matching and statistical analysis.
1m
1.2K
Asynchronous logging worker. Ingests detection events into analytics database with configurable rete
1m
862
Content moderation orchestrator. Coordinates content safety analysis across multiple language-specif
1m
858
Helper scripts and utilities. Includes stack initialization, health checks, and management tools.
1m
1.7K
NATS integration layer for LLM Guard detection model. Provides request-reply interface to injection
1m
1.2K
SIEM log forwarder. Sends normalized detection events to external security systems via Vector integr
1m
770