repowiseniko1221/strata

niko1221 / strata

Qwen3.8-Flash-Next on any consumer hardware: one-click install for Windows / Linux. Strata inference engine, OpenAI/Anthropic API on localhost, optional image input.

C++★ 32,091
Use in your agentAsk this codebase

github.com/niko1221/strataIndexed at 7dd5cf1Up to date with upstream

What is strata?

Strata is an inference engine for running the Qwen3.8-Flash-Next model, a 125-billion-parameter model, on a consumer PC instead of a server. It splits the model's work across graphics memory, RAM and an SSD so the model fits in the memory the computer already has. The repository is written mainly in C++, with Python, TypeScript, JavaScript and Rust files for its tools and apps.

Written by a model from the README and the source

3,950 files and 33,309 symbols in 19 modules, led by Python (784 files) and C++ (760).

Start reading

Code health

5.1out of 10Needs work

Hotspots average 2.9, the place fixes pay off most.

Full health report

Use niko1221/strata in your agent

Point Claude Code, Cursor, Codex or VS Code at this repository. Your agent gets the index this page is built from: cited answers, callers, history and health for any file.

  1. 1.In Claude, open Customize → Connectors, click +, then Add custom connector. On a Team or Enterprise plan an owner adds it first under Organization settings → Connectors.
  2. 2.Paste the server URL below as the Remote MCP server URL, then click Add.
  3. 3.Click Connect. A Repowise sign-in window opens. Approve access.
  4. 4.Repowise's tools are now available in your Claude chats.

Connect with one click. Sign in with Repowise (OAuth), no API key to copy. Works in Claude on the web, desktop, and mobile.

Docs
bash
https://api.repowise.dev/mcp/niko1221/strata
Editors need an API key: sign in to create one.Sign in to connect

This repo is public, so anyone with a Repowise account and an API key can connect.

Try asking

  • get_overview()
  • get_answer("how is strata put together?")
  • get_why("why is strata structured this way?")

Explore strata

Key features of strata

  • Runs Qwen3.8-Flash-Next on a gaming PC with 32 GB of RAM and graphics memory together.
  • Connects coding agents such as Claude Code, Codex CLI and OpenCode through the OpenAI and Anthropic APIs.
  • Reads pictures, with picture input enabled by default.
  • Runs with no account, no cloud and no telemetry, and is released under the MIT license.

How niko1221/strata works

serve/server.py is an entry point in the serve directory, which exposes the OpenAI and Anthropic APIs on localhost for chat apps and coding agents. The generation work is in src/program/generate.cpp, a hotspot file that drives the production of each answer. src/core/expert_source.cpp is a hotspot file that handles where the model's experts are read from, as the README describes, with most of them kept on the SSD and read a few rows at a time. src/core/verify.cpp is a hotspot file in the same core directory that checks the guesses of the small helper model against the big model. The desktop app starts from apps/desktop/src-tauri/src/main.rs, a Rust entry point, and the web front end starts from apps/host/src/main.tsx, a TypeScript entry point.

The codebase has 3,950 files and 33,309 symbols in 19 modules, led by Python, C++ and JavaScript. Code health is 5.1 out of 10, rated needs work.

Getting started with strata

  1. Start Strata, which asks a few questions, downloads what it needs and starts serving.
    strata launch

Commands copied from the README

niko1221/strata FAQ

What is Strata?

Strata is an inference engine that runs the Qwen3.8-Flash-Next model on a consumer PC, a model that normally needs a server. It serves OpenAI and Anthropic APIs on localhost so chat apps and coding agents can connect to it.

How does Strata work?

Strata spreads the model across your graphics card, RAM and SSD, keeping the most often used experts on the graphics card and reading the rest from RAM or the SSD as needed. A small helper guesses the next few words, and the big model checks them in one pass and keeps the correct ones.

How do I install Strata?

Strata is installed on Linux and Mac with a one-line curl command, and on Windows with a one-line PowerShell command. Then run strata launch to choose a model size, download what it needs and start serving.

What language is Strata written in?

Strata is written mainly in C++. The repository also includes Python, TypeScript, JavaScript, shell and Rust files.

How is the Strata code organized?

The Strata repository has top-level directories such as apps, tools, sycl, bench, src, tests, include and serve. The core engine code is in src, and the server code is in serve.

Who is Strata for?

Strata is for people who want to run Qwen3.8-Flash-Next on their own gaming PC, Mac or unified-memory PC. It needs at least 32 GB of RAM and graphics memory together.

What license does Strata use?

Strata is released under the MIT license, whose SPDX id is MIT. The license lets you use the project without an account or cloud service.

The summary, features and answers on this page are written by a model from the README and the source, and refreshed with each index.

About this page

Get a page like this for your own codebase

Repowise built this map of niko1221/strata from its source and git history. It parses every symbol, draws the dependency graph, scores each file's health from complexity, duplication, coverage and churn, finds hotspots and owners in the history, and writes documentation for every module. Your coding agent reads the same index over MCP, so its answers cite real files.

Index your repo freeSee how Repowise worksFree for 2 public repos, with a page like this one.

Prefer local? pip install repowise && repowise init indexes your own checkout, no account needed.