Skip to main content

VCAL Semantic Cache

VCAL Semantic Cache is an optional dedicated semantic-reuse module for AI Cost Firewall deployments. It is intended for environments where reusable model results need to be shared across applications or gateway instances, or where the semantic cache requires its own operational lifecycle.

AI Cost Firewall remains the gateway, orchestration layer, and commercial entry point. VCAL Semantic Cache provides a separately operated cache backend when the built-in caching path is not sufficient for the agreed architecture.

When to use it​

VCAL Semantic Cache is best suited to deployments that require one or more of the following:

  • shared semantic reuse across multiple AI Cost Firewall instances or applications;
  • operational separation of the cache from the gateway lifecycle;
  • dedicated capacity, persistence, authentication, and observability;
  • on-premises or private-cloud control of embeddings and cached responses;
  • direct HTTP access from approved RAG services or agents within the same deployment architecture.

It is not required for every AI Cost Firewall deployment. The built-in Redis and Qdrant cache path remains the default where it meets the workload and operational requirements.

Capabilities​

  • REST API for cache traffic and operational control under /v1/*
  • TTL and LRU eviction
  • Background-safe snapshot persistence
  • Prometheus metrics at /metrics
  • API-key authentication with app and admin roles
  • Signed-license enforcement
  • Docker-container or standalone-binary deployment
  • Customer-controlled data storage

Delivery and availability​

VCAL Semantic Cache is not offered as a self-service download or independent public SKU. When it is justified by an AI Cost Firewall architecture, VCAL supplies the agreed deployment artifact and license directly.

Available artifact formats include:

  • an archived Docker container image; or
  • a standalone Linux binary package.

The delivery package can include checksums, release signatures, an SBOM, an OpenAPI definition, monitoring assets, and deployment-specific instructions. Contact VCAL to discuss whether the module fits your deployment.

Integration model​

In the standard architecture, AI Cost Firewall owns request processing and cache policy while VCAL Semantic Cache provides the dedicated reuse service. Applications, RAG pipelines, or agents may call its HTTP API directly only where this is part of the approved deployment design.

The calling component remains responsible for:

  1. generating embeddings with the selected model;
  2. querying the cache using compatible vector dimensions;
  3. deciding whether a match is suitable for reuse; and
  4. storing an approved response after an upstream model call.

Documentation​