# darkdatter/gb10-repo architecture

> Architecture diagram of the GitHub repository darkdatter/gb10-repo, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-01). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/darkdatter/gb10-repo
- Repository: https://github.com/darkdatter/gb10-repo

## Overview

This repository is a measured deployment and benchmarking recipe for Qwen3.8-27B on DGX Spark, not the model server or SparkStation implementation itself. The documented workflow verifies GPU access, prepares the image and weights, serves the model standalone or through optional SparkStation, then benchmarks throughput and HumanEval quality. Tuning guidance and recorded results are also first-class product outputs. Relationships below are based on README statements; no source excerpts were available to verify implementation-level calls.

Main entry points: [`patches/sparkstation-models.yaml.md`](https://github.com/darkdatter/gb10-repo/blob/main/patches/sparkstation-models.yaml.md) (SparkStation gateway), [`scripts/01-build-and-fetch.sh`](https://github.com/darkdatter/gb10-repo/blob/main/scripts/01-build-and-fetch.sh) (Image and weights), [`scripts`](https://github.com/darkdatter/gb10-repo/tree/main/scripts) (Standalone serving), [`scripts/run-humaneval.sh`](https://github.com/darkdatter/gb10-repo/blob/main/scripts/run-humaneval.sh) (HumanEval runner)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_setup["Setup"]
  node_gpu_check["GPU verification<br/>[00-verify-gpu.sh]"]
  node_image_weights["Image and weights"]
end

subgraph group_serving["Serving"]
  node_standalone["Standalone serving"]
  node_sparkstation["SparkStation gateway"]
end

subgraph group_benchmarks["Benchmarks"]
  node_perf["Performance benchmark<br/>[perf.py]"]
  node_longctx["Long-context benchmark<br/>[longctx.py]"]
  node_humaneval_runner["HumanEval runner<br/>[run-humaneval.sh]"]
  node_humaneval_generate["HumanEval generation<br/>[generate.py]"]
  node_humaneval_execute["HumanEval execution<br/>[execute.py]"]
  node_humaneval_report["HumanEval report<br/>[report.py]"]
  node_benchmark_common["Benchmark helpers<br/>[common.py]"]
end

subgraph group_tuning["Tuning and results"]
  node_performance_results["Measured results<br/>[RESULTS.md]"]
  node_reproduction["Reproduction notes<br/>[REPRODUCTION.md]"]
  node_build_manifest["Build manifest<br/>[BUILD-MANIFEST.md]"]
  node_model_tuning["Model tuning notes"]
  node_gateway_health_patch["Health patch"]
  node_memory_patch["Memory override patch"]
  node_model_patch_notes["Model patch notes"]
end

node_user(("Operator"))
node_model_registry["Model repositories"]

node_user -->|"verifies GPU"| node_gpu_check
node_user -->|"prepares artifacts"| node_image_weights
node_image_weights -->|"fetches weights"| node_model_registry
node_user -->|"starts server"| node_standalone
node_user -->|"starts gateway"| node_sparkstation
node_standalone -->|"serves benchmark"| node_perf
node_standalone -->|"serves evaluations"| node_humaneval_runner
node_humaneval_runner -->|"executes problems"| node_humaneval_execute
node_user -.->|"runs benchmark"| node_longctx
node_perf -->|"informs measurements"| node_performance_results
node_humaneval_runner -->|"records quality"| node_performance_results
node_image_weights -->|"documents inputs"| node_build_manifest
node_sparkstation -->|"uses configuration"| node_model_tuning
node_sparkstation -->|"documents mitigation"| node_gateway_health_patch
node_sparkstation -->|"documents override"| node_memory_patch
node_model_tuning -.->|"describes settings"| node_model_patch_notes
node_performance_results -->|"compares runs"| node_reproduction

click node_gpu_check "https://github.com/darkdatter/gb10-repo/blob/main/scripts/00-verify-gpu.sh"
click node_image_weights "https://github.com/darkdatter/gb10-repo/blob/main/scripts/01-build-and-fetch.sh"
click node_standalone "https://github.com/darkdatter/gb10-repo/tree/main/scripts"
click node_sparkstation "https://github.com/darkdatter/gb10-repo/blob/main/patches/sparkstation-models.yaml.md"
click node_perf "https://github.com/darkdatter/gb10-repo/blob/main/bench/perf.py"
click node_longctx "https://github.com/darkdatter/gb10-repo/blob/main/bench/longctx.py"
click node_humaneval_runner "https://github.com/darkdatter/gb10-repo/blob/main/scripts/run-humaneval.sh"
click node_humaneval_generate "https://github.com/darkdatter/gb10-repo/blob/main/bench/humaneval/generate.py"
click node_humaneval_execute "https://github.com/darkdatter/gb10-repo/blob/main/bench/humaneval/execute.py"
click node_humaneval_report "https://github.com/darkdatter/gb10-repo/blob/main/bench/humaneval/report.py"
click node_benchmark_common "https://github.com/darkdatter/gb10-repo/blob/main/bench/common.py"
click node_performance_results "https://github.com/darkdatter/gb10-repo/blob/main/results/RESULTS.md"
click node_reproduction "https://github.com/darkdatter/gb10-repo/blob/main/results/REPRODUCTION.md"
click node_build_manifest "https://github.com/darkdatter/gb10-repo/blob/main/results/BUILD-MANIFEST.md"
click node_model_tuning "https://github.com/darkdatter/gb10-repo/blob/main/patches/sparkstation-models.yaml.md"
click node_gateway_health_patch "https://github.com/darkdatter/gb10-repo/blob/main/patches/sparkstation-gateway-health.patch"
click node_memory_patch "https://github.com/darkdatter/gb10-repo/blob/main/patches/sparkstation-mem-fraction-override.patch"
click node_model_patch_notes "https://github.com/darkdatter/gb10-repo/blob/main/patches/sparkstation-models.yaml.md"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_gpu_check,node_image_weights toneBlue
class node_standalone,node_sparkstation toneAmber
class node_perf,node_longctx,node_humaneval_runner,node_humaneval_generate,node_humaneval_execute,node_humaneval_report,node_benchmark_common toneMint
class node_performance_results,node_reproduction,node_build_manifest,node_model_tuning,node_gateway_health_patch,node_memory_patch,node_model_patch_notes toneRose
class node_user,node_model_registry toneIndigo
```

## Components

### Setup

- [GPU verification](https://github.com/darkdatter/gb10-repo/blob/main/scripts/00-verify-gpu.sh): `scripts/00-verify-gpu.sh`
- [Image and weights](https://github.com/darkdatter/gb10-repo/blob/main/scripts/01-build-and-fetch.sh): `scripts/01-build-and-fetch.sh`

### Serving

- [Standalone serving](https://github.com/darkdatter/gb10-repo/tree/main/scripts): `scripts`
- [SparkStation gateway](https://github.com/darkdatter/gb10-repo/blob/main/patches/sparkstation-models.yaml.md): `patches/sparkstation-models.yaml.md`

### Benchmarks

- [Performance benchmark](https://github.com/darkdatter/gb10-repo/blob/main/bench/perf.py): `bench/perf.py`
- [Long-context benchmark](https://github.com/darkdatter/gb10-repo/blob/main/bench/longctx.py): `bench/longctx.py`
- [HumanEval runner](https://github.com/darkdatter/gb10-repo/blob/main/scripts/run-humaneval.sh): `scripts/run-humaneval.sh`
- [HumanEval generation](https://github.com/darkdatter/gb10-repo/blob/main/bench/humaneval/generate.py): `bench/humaneval/generate.py`
- [HumanEval execution](https://github.com/darkdatter/gb10-repo/blob/main/bench/humaneval/execute.py): `bench/humaneval/execute.py`
- [HumanEval report](https://github.com/darkdatter/gb10-repo/blob/main/bench/humaneval/report.py): `bench/humaneval/report.py`
- [Benchmark helpers](https://github.com/darkdatter/gb10-repo/blob/main/bench/common.py): `bench/common.py`

### Tuning and results

- [Measured results](https://github.com/darkdatter/gb10-repo/blob/main/results/RESULTS.md): `results/RESULTS.md`
- [Reproduction notes](https://github.com/darkdatter/gb10-repo/blob/main/results/REPRODUCTION.md): `results/REPRODUCTION.md`
- [Build manifest](https://github.com/darkdatter/gb10-repo/blob/main/results/BUILD-MANIFEST.md): `results/BUILD-MANIFEST.md`
- [Model tuning notes](https://github.com/darkdatter/gb10-repo/blob/main/patches/sparkstation-models.yaml.md): `patches/sparkstation-models.yaml.md`
- [Health patch](https://github.com/darkdatter/gb10-repo/blob/main/patches/sparkstation-gateway-health.patch): `patches/sparkstation-gateway-health.patch`
- [Memory override patch](https://github.com/darkdatter/gb10-repo/blob/main/patches/sparkstation-mem-fraction-override.patch): `patches/sparkstation-mem-fraction-override.patch`
- [Model patch notes](https://github.com/darkdatter/gb10-repo/blob/main/patches/sparkstation-models.yaml.md): `patches/sparkstation-models.yaml.md`

### Other components

- **Operator**
- **Model repositories**

## Connections

- Operator → GPU verification: verifies GPU (evidence: `README.md`)
- Operator → Image and weights: prepares artifacts (evidence: `README.md`)
- Image and weights → Model repositories: fetches weights (evidence: `README.md`)
- Operator → Standalone serving: starts server (evidence: `README.md`)
- Operator → SparkStation gateway: starts gateway (evidence: `README.md`)
- Standalone serving → Performance benchmark: serves benchmark (evidence: `README.md`)
- Standalone serving → HumanEval runner: serves evaluations (evidence: `README.md`)
- HumanEval runner → HumanEval execution: executes problems (evidence: `README.md`)
- Operator → Long-context benchmark: runs benchmark
- Performance benchmark → Measured results: informs measurements (evidence: `README.md`)
- HumanEval runner → Measured results: records quality (evidence: `README.md`)
- Image and weights → Build manifest: documents inputs (evidence: `README.md`)
- SparkStation gateway → Model tuning notes: uses configuration (evidence: `README.md`)
- SparkStation gateway → Health patch: documents mitigation (evidence: `README.md`)
- SparkStation gateway → Memory override patch: documents override (evidence: `README.md`)
- Model tuning notes → Model patch notes: describes settings
- Measured results → Reproduction notes: compares runs (evidence: `README.md`)
