Today LangSmart becomes APERION.
When we started this company, we called it LangSmart because we were building something smart with language models. That was accurate. It was also small. What we have actually built, and what enterprises are telling us they desperately need, is something much bigger: the governance control plane that sits between an organization and every AI model, agent, and workflow it deploys.
We call that category AI Sovereignty. The thesis is simple: cloud-based AI governance will never work for regulated industries. You cannot outsource control of your AI to the same vendors selling you the AI. Enterprises in financial services, healthcare, and defense need to own their AI infrastructure end-to-end. On their hardware. Under their policies. With their data never leaving.
Why Now
One week ago, the most widely used open-source LLM proxy in the Python ecosystem was supply-chain attacked. LiteLLM versions 1.82.7 and 1.82.8 were compromised by the threat actor group TeamPCP through a cascading attack that originated in a compromised security scanner. The malicious packages contained credential-stealing malware. The entire LiteLLM package remains quarantined on PyPI. According to Wiz, LiteLLM was present in 36 percent of all cloud environments.
Smartflow, our flagship product, was unaffected. Not because we were fortunate. Because our architecture makes this category of attack irrelevant. Smartflow is deployed as an on-premises software appliance, behind the enterprise firewall, with no dependency on public package registries and no external CI/CD pipelines in the customer deployment path. The PyPI supply chain is not in Smartflow’s threat model.
Since the LiteLLM incident, web traffic to our site has increased over 200 percent, driven by search terms related to LiteLLM security and AI gateway alternatives. The enterprises reaching out are not startups. They are the institutions that define what production-grade means.
What We Are Launching Today
The company name changes from LangSmart to APERION. The product name does not change. Smartflow remains Smartflow.
Alongside the rebrand, we are releasing the Smartflow SDK, a Python library that provides enterprise developers with an immediate path from evaluation to production-grade AI governance. The SDK detects whether a Smartflow appliance is available. If yes, it enables the full feature set: log retrieval, per-user and per-group limits, semantic caching at p95, and inline policy enforcement. If no appliance is detected, it installs a standalone software gateway with feature parity to alternatives including LiteLLM and OpenRouter, giving developers a governed AI environment without hardware commitment.
We have also published a public migration whitepaper documenting the step-by-step process for transitioning from OpenRouter and other OpenAI-compatible proxies to a Smartflow instance.
Where We Stand
Smartflow is in production with paying enterprise customers, including DDA, and under evaluation at financial-services institutions. Our patent-pending control plane covers enterprise AI governance, sovereign model deployment, and autonomous AI control plane architecture. MetaCache, our four-phase BERT semantic cache, was benchmarked at NVIDIA GTC 2026.
Smartflow is built in Rust. Kubernetes-native. Authenticates against enterprise identity providers over OIDC, including Entra ID and Okta, and composes with SAML and Kerberos single sign-on. No-code compliance policy engine mapping to EU AI Act, NIST AI RMF, FINRA, and HIPAA requirements.
The market for AI governance infrastructure, which we estimate at $40 to $50 billion based on a 5 to 15 percent governance capture of overall AI infrastructure spend, is just beginning to form. The LiteLLM incident accelerated the timeline. Regulated enterprises that were evaluating AI gateways on features and latency are now evaluating them on deployment model, supply chain provenance, and architectural security posture.
APERION was built for that evaluation.
The product name does not change. Smartflow remains Smartflow. The company name now matches the mission.
Put this in the path of your own agents.
Policy enforced inline between your agents and every model and tool they reach, with a record bound to the human who owns it.
Request a Demo Read the docs