Hithe

Introduction

Hithe is an LLM gateway and proxy. Every application that needs to call a model routes through Hithe instead of hitting the provider directly, and in exchange, Hithe adds metering, governance, auditability, and compliance controls on top of that traffic. It isn't a workflow builder. It's infrastructure that sits underneath wherever workflows eventually make their model calls, positioned to eventually replace Horizon, Dais's node-based agentic workflow product.

The Goal

Design a gateway interface that a compliance officer, not just an engineer, can trust at a glance, one where the parts of the product carrying legal and safety weight are visually and structurally distinct from routine configuration.

The Challenge

A metering and governance product succeeds or fails on a specific kind of trust: the customer isn't just asking "does this work," they're asking "can I prove to my own auditors that this worked, every time, for every request." That's a harder design problem than most SaaS products solve, because the interface itself is part of the compliance story, not just a way to configure the product.

Discovery Phase

Building the structure

David reconstructed the product screen by screen from the live product, following an initial discovery conversation with Dais's VP of Engineering about Hithe's direction. The audit surfaced structural issues invisible from any single screen: duplicate settings, conflicting billing surfaces, and an approval flow with no second reviewer to use it.

The Research

Competitive Audit

The LLM gateway market splits cleanly along one axis: self-hosted, developer-first tools built for routing flexibility, versus managed, compliance-first platforms built for regulated enterprises. Hithe sits, or needs to sit, firmly in the second category.

Takeaway

Hithe's positioning: the products winning in regulated-industry deals aren't winning on routing features. LiteLLM already does that for free. They're winning on making compliance and audit posture visible and provable inside the product itself. That's the gap Hithe's interface needs to close, and none of these four fully solve it at the interface level either.

The Pain Points

User Personas

Hi-Fidelity Designs

Current State

These are the current, in-progress high-fidelity mockups for the proposed direction. They reflect design decisions made from the audit findings alone; they have not been through usability testing or stakeholder feedback sessions, since that phase hadn't started before David's involvement with the product ended. Treat these as a working direction, not a validated one.

Going Forward

Impact

The audit surfaced concrete, specific risks in a product built for a compliance-sensitive audience, most notably that its two safety-critical flows were visually indistinguishable from routine settings pages. That's a finding with real product implications regardless of who acts on it next.

What David Learned

The next phase would have been validating the proposed direction with the people who actually carry the compliance and legal risk, likely legal or compliance stakeholders alongside engineering, before any of it touched a real interface. That validation never happened. Ownership of Hithe was still in transition when a round of layoffs, including David's own, ended his involvement with the product.

Next Steps

The next phase would have been bringing the proposed direction into validation with compliance and legal stakeholders, not just engineering, before any of it shipped. Knowing precisely what that phase should have been, and why it matters for this specific product category, is itself part of the work.

Want to know more about David Rivas? You can also check out Design Skills & Guardrails or Battery Trader.