Inventory and connectors
Connects to Chroma, pgvector, Pinecone, Qdrant, Weaviate, Milvus, Redis, OpenSearch, MongoDB, OpenAI vector stores, Azure AI Search, Amazon Bedrock Knowledge Bases and Vertex AI. The Schema Explorer shows each collection's metadata fields, types and how often each is filled.
Sensitive-data scans with stated coverage
Detects personal data, health information, cardholder data and credentials in chunk text. Collections that were sampled or not scanned are listed as such instead of being reported as clean.
Posture and access
Checks network exposure, TLS, encryption at rest where the store exposes it, and dependency CVEs. Compares granted access with intended access, flags stale or over-privileged API keys, and probes for cross-tenant retrieval.
RAG integrity and query monitoring
Finds prompt-injection text, chunks without provenance and anomalous embeddings. Query monitoring, from replayed logs or a wrapper around your connector, alerts on bulk reads, enumeration and collection sweeps.
DSAR, erasure and DPIA
Finds a data subject's records across every registered store, runs erasure with a verification pass afterwards, and prepares the inputs for a GDPR Article 35 DPIA from what the scans found.
Fixes with approvals
Label, tag, redact or quarantine chunks, assign a missing tenant, or split tenants into their own collections. Each change is staged, approved by a second person, executed and read back. Fixes no connector can make come as store-specific steps, checked by a re-scan.
Masked review and audit log
Findings keep a masked preview, not the text. Raw values need the compliance role and a stated reason. Reveals, approvals and executions go into a hash-chained audit log.
Evidence and operations
Exports evidence mapped to GDPR, HIPAA, PCI-DSS, SOC 2, CCPA/CPRA and the EU AI Act. Deploys with Docker, Helm or Terraform, including air-gapped. API roles, JWT auth, health probes and Prometheus metrics are built in.