HELIX V6

A runtime for AI workloads with local inference: orchestration, memory, tool routing and evaluation

Internal research and development, built solo: 121,000 lines of code across 633 source files, with more than 3,800 automated tests.

The problem

An agent is easy to demonstrate and hard to keep running. The difference is not the model. It is everything around it: the data it is allowed to see, the evaluation that tells you when it degrades, the fallback when it fails and the person who owns the outcome.

That surrounding layer is where AI workloads fail in production. HELIX V6 is NASCIA's internal research and development on that layer: a runtime for AI workloads with local inference.

What was built

A runtime for AI workloads with local inference, built solo by the founder. It covers orchestration, memory, tool routing and evaluation.

The codebase holds 121,000 lines of code across 633 source files, with more than 3,800 automated tests (founder's CV).

It is internal research and development. There is no public link and no product page.

How it works

HELIX V6, schematic: a request enters the runtime; orchestration works with memory, tool routing and a local model; evaluation is the fourth area

The diagram is schematic. It shows the four areas the runtime covers and a local model, and it claims no flow, protocol or storage detail beyond that; tools are drawn outside the runtime for legibility, not because their placement is published. Orchestration drives the workload. Memory holds state between steps. Tool routing directs each call to the tool that should handle it. Inference can run on a local model (founder's CV). Evaluation is one of the four areas the runtime covers. Separately, the codebase carries more than 3,800 automated tests (founder's CV).

Evidence

Three figures are published, all from one source: 121,000 lines of code (founder's CV), 633 source files (founder's CV) and more than 3,800 automated tests (founder's CV).

No usage, performance or benchmark figure is published, because none has been verified.

Stack

Inference can run on a local model (founder's CV). The runtime covers orchestration, memory, tool routing and evaluation (founder's CV). Languages, frameworks and model names are not published.

What it demonstrates

Define: the runtime has a named scope, orchestration, memory, tool routing and evaluation.

Instrument: more than 3,800 automated tests sit beside 121,000 lines of code across 633 source files (founder's CV).

Integrate: local inference and tool routing are the parts that put a model inside a workflow, which is what NASCIA ships for clients.

Hand over: a runtime with a test suite is one that another engineer can take over. The same discipline goes into every client system.

Evidence

lines of codeSource: founder's CV (helix-04), 24 September 2026: 121,000
source filesSource: founder's CV (helix-05), 24 September 2026: 633
automated testsSource: founder's CV (helix-06), 24 September 2026: More than 3,800