Sanar.
← Build Log

Introduction: What is Sanar?

If you're reading this, first let me start by saying that I appreciate you investing the time to see what this project is about. You may have stumbled upon me through Twitter, Reddit, LinkedIn, Discord, or even Slack. Regardless of how you've ended up here, my name is Trent. I am the founder/developer of Sanar.

There is no perfect way to start this post. Some people may want to know my technical background; some people may just want to know what Sanar actually is.

First, the etymology:

sanar

Q. mind

sanar noun "mind" (literally "thinker" or "reflector", suggesting an underlying verb #sana- "to think, to reflect") (VT41:13)

Taken from Quenya, the language developed by J.R.R. Tolkien.

Lovely word, very literal. Possibly too on the nose. It also happened to be readily available for domains, so there was that.

Sanar began its life as a tool for construction management workflows, originally intended for mostly document parsing, data analytics, and cost management tools. I quickly figured out that the juice wasn't worth the squeeze, there was a reason why the existing tools sucked. Mostly, construction companies just weren't all that interested in spending on high-end software when their workforce hated technology.

The second take was a scrappy attempt at an AI orchestration tool. I had lots of ideas, very little familiarity with the space, and I worked very hard to find that I had...

Built n8n with extra steps.

Awesome, not quite what I was going for. Lots of lessons learned though, and I now had a better understanding of the space I wanted to work in, as well as a crystalized vision for what I wanted to build.

This led to the new vision for Sanar:

AI infrastructure.

What exactly is it? AI infrastructure can mean a lot of things, and the goal with Sanar is to develop products in both the digital and physical space. Digitally, this looks like:

  • Intelligent codebase indexing that reduces query time and provides semantic diff capabilities for code reviewers that utilize AI in their production teams. This is currently the primary focus for the MVP debut scheduled for the end of Q1 2026. (WIP, future posts will discuss research as well as development)
  • Security as a core feature, not as an afterthought. We are fully integrated with AWS KMS, have a 4-layer model for multi-tenant isolation, magic link authentication, and tamper-evident audit logging. All of this ships on day ONE. These aren't WIP, these are fully functional.
  • Polyglot architecture, Sanar utilizes TypeScript/Node.js for API orchestration and CLI, Rust for performance-critical operations, and Python sidecar for isolated automation runtimes, with process-level sandboxing.
  • The plugin architecture is designed, and core commands are registered. It includes trust levels, SBOM, and security scan tracking along with checksum version management.
  • There are pluggable adapters for ITSM, issue tracking, multiple auth methods, and normalized state mapping across systems.
  • An observability stack with Pino structured logging, tamper proof audit logs with hash chains, OpenTelemetry tracing, multi-export telemetry.
  • Metering & quota management, including full support for event-based metering (tokens, compute hours, API calls), real-time quota aggregation, and budget alerts.

You might notice that only one item is marked as WIP. The bottom line: This isn't a product, it's a platform. Every piece is designed to be extended, replaced, or scaled independently. This is for two reasons:

  1. I build data centers for a living; I build digitally the same way I do physically. Everything has a lifespan, and you need to be able to expand as needed. Scalability is critical.
  2. By having focused on core functionality and designing modules that naturally carry through the rest of the codebase, I can rapidly iterate multiple ideas and prototype features without having to rebuild debugging and logging, all I have to do is build unit/integration testing for them.

In the physical realm, two ASUS Ascent GX10 units have been procured. These units will be used to move cloud training pipelines to a local level, allowing for higher parameter training at a much smaller cost than cloud compute. Additionally, it opens the door to cost effective from-scratch training for custom models to operate in the Sanar ecosystem. These will also serve as mobile demo platforms.

In summary, Sanar is a full-scope AI platform, with the goal of providing an end-to-end developer environment, including compute resources at reasonable terms as a benefit of being on our platform. If you're interested in working on this project, you won't be restricted to a regimented vision of what cloud infrastructure needs to be, and you'll have access to computational resources that will allow you to develop the way YOU want to.

I am actively seeking beta users/teams to partner with in 2026.

To connect with me, you can reach me at trent@sanar.co or on Twitter @trntvalentine