Public scan — anyone with this URL can view this analysis. Sign up to track your own repos privately, run scheduled re-scans, and get AI fix prompts via your dashboard.

vessl-ai/vessl-cloud-cookbook

https://github.com/vessl-ai/vessl-cloud-cookbook · scanned 2026-06-16 00:48 UTC (2 months, 1 week ago)

36 raw signals (15 security + 21 graph)

File as issue Image

UNIFIED Repobility · multi-layer engine · AI coders

Complete repo analysis

Last scanned 2 months, 1 week ago · v1 · 29 actionable findings from 2 signal sources. 24 repeated signals grouped for readability. Security checks, system graph analysis, and verified AI-agent feedback are merged into one review queue.

JSON

77.8% cov · 21 findings

Severity distribution — click a segment to filter

Active filters: excluding tests × Reset all

Severity: Critical 2 High 4 Medium 6 Low 12 9-Layer: Software Security Quality Integrity Frontend Hardware Data Network Cicd Source: Security checks 8 System graph 21 Crowd 0 Layer: Quality 16 Security 3 Software 8 Frontend 1 Cicd 1

Exclude dismissed / FP Exclude test files

All 215 nodes from the latest scan, grouped by kind. Each node is a unit the engine identified (file, function, endpoint, table…). Most users won't need this view — it's primarily for debugging the engine's graph extraction or for AI agents that want to enumerate the project structure.

Label	Layer	Status	Path
`fetch_metrics`	software	healthy	`gpu-cost-benchmark/batch-job/bench_client.py:65`
`one_request`	software	healthy	`gpu-cost-benchmark/batch-job/bench_client.py:86`
`worker`	software	healthy	`gpu-cost-benchmark/batch-job/bench_client.py:148`
`run_bench`	software	healthy	`gpu-cost-benchmark/batch-job/bench_client.py:165`
`gen`	software	healthy	`gpu-cost-benchmark/batch-job/bench_client.py:169`
`warmup_then_measure`	software	healthy	`gpu-cost-benchmark/batch-job/bench_client.py:184`
`pct`	software	healthy	`gpu-cost-benchmark/batch-job/bench_client.py:209`
`main`	software	healthy	`gpu-cost-benchmark/batch-job/bench_client.py:296`
`is_rank0`	software	dead	`gpu-cost-benchmark/batch-job/train_sft.py:148`
`world_size`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:152`
`fmt_gib`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:156`
`__init__`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:319`
`training_step`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:170`
`apply_te_fp8`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:177`
`start`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:239`
`_loop`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:245`
`latest`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:279`
`summary`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:282`
`stop`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:299`
`theoretical_step_flops`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:308`
`on_step_begin`	software	dead	`gpu-cost-benchmark/batch-job/train_sft.py:343`
`on_step_end`	software	dead	`gpu-cost-benchmark/batch-job/train_sft.py:348`
`on_log`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:353`
`log_step`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:455`
`close`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:466`
`_pct`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:473`
`write_summary`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:481`
`build_base_with_lora`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:580`
`fp8_filter`	software	dead	`gpu-cost-benchmark/batch-job/train_sft.py:617`
`autotune_mbsz`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:633`
`maybe_setup_wandb`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:678`
`report_to_list`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:690`
`run_trainer_variant`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:697`
`run_tp_variant`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:891`
`collate`	software	dead	`gpu-cost-benchmark/batch-job/train_sft.py:962`
`main`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:1056`
`save`	software	healthy	`gpu-cost-benchmark/analysis/make_figures.py:67`
`fig1_training_tps`	software	healthy	`gpu-cost-benchmark/analysis/make_figures.py:77`
`fig2_training_cost`	software	healthy	`gpu-cost-benchmark/analysis/make_figures.py:131`
`fig3_inference_tps`	software	healthy	`gpu-cost-benchmark/analysis/make_figures.py:188`
`fig4_mtp_effect`	software	healthy	`gpu-cost-benchmark/analysis/make_figures.py:243`
`fig6_vram_tps`	software	healthy	`gpu-cost-benchmark/analysis/make_figures.py:305`
`fig7_kvs_effect`	software	healthy	`gpu-cost-benchmark/analysis/make_figures.py:363`
`plot_series`	software	healthy	`gpu-cost-benchmark/analysis/make_figures.py:372`
`fig8_inference_cost`	software	healthy	`gpu-cost-benchmark/analysis/make_figures.py:425`
`ci`	software	healthy	`gpu-cost-benchmark/analysis/make_figures.py:429`
`fig9_concurrency_sweep`	software	healthy	`gpu-cost-benchmark/analysis/make_figures.py:504`
`get_variant`	software	dead	`gpu-cost-benchmark/analysis/make_figures.py:505`
`fig10_lora_overhead`	software	healthy	`gpu-cost-benchmark/analysis/make_figures.py:556`
`fig11_stability_heatmap`	software	healthy	`gpu-cost-benchmark/analysis/make_figures.py:627`

Showing first 50 of this kind. Full payload available via the JSON button at the top of the page.

Label	Layer	Status	Path
`README.md`	software	healthy	`README.md`
`CONTRIBUTING.md`	software	healthy	`CONTRIBUTING.md`
`LICENSE`	software	healthy	`LICENSE`
`README.md`	software	healthy	`_template/README.md`
`benchmarks.md`	software	healthy	`_template/benchmarks.md`
`requirements.txt`	software	healthy	`_template/requirements.txt`
`submit.sh`	software	healthy	`_template/batch-job/submit.sh`
`recipe.py`	software	healthy	`_template/batch-job/recipe.py`
`README.md`	software	healthy	`gpu-cost-benchmark/README.md`
`benchmarks.md`	software	healthy	`gpu-cost-benchmark/benchmarks.md`
`requirements.txt`	software	healthy	`gpu-cost-benchmark/requirements.txt`
`submit.sh`	software	healthy	`gpu-cost-benchmark/batch-job/submit.sh`
`run_full_per_hardware.sh`	software	healthy	`gpu-cost-benchmark/batch-job/run_full_per_hardware.sh`
`serve_and_bench.sh`	software	healthy	`gpu-cost-benchmark/batch-job/serve_and_bench.sh`
`bench_client.py`	software	healthy	`gpu-cost-benchmark/batch-job/bench_client.py`
`train_sft.py`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py`
`stability_metrics.json`	software	healthy	`gpu-cost-benchmark/results/stability_metrics.json`
`make_figures.py`	software	healthy	`gpu-cost-benchmark/analysis/make_figures.py`
`compute_cost.py`	software	healthy	`gpu-cost-benchmark/analysis/compute_cost.py`
`pack_sharegpt.py`	software	healthy	`gpu-cost-benchmark/data_prep/pack_sharegpt.py`
`train.py`	software	healthy	`aqr-finance/train.py`
`prepare.py`	software	healthy	`aqr-finance/prepare.py`
`eval.py`	software	healthy	`aqr-finance/eval.py`
`README.md`	software	healthy	`aqr-finance/README.md`
`accelerate_config.yaml`	software	healthy	`aqr-finance/accelerate_config.yaml`
`benchmarks.md`	software	healthy	`aqr-finance/benchmarks.md`
`requirements.txt`	software	healthy	`aqr-finance/requirements.txt`
`submit.sh`	software	healthy	`aqr-finance/batch-job/submit.sh`
`prep.sh`	software	healthy	`aqr-finance/batch-job/prep.sh`
`submit-async.sh`	software	healthy	`aqr-finance/batch-job/submit-async.sh`
`wait-jobs.sh`	software	healthy	`aqr-finance/batch-job/wait-jobs.sh`
`_lib.sh`	software	healthy	`aqr-finance/batch-job/_lib.sh`
`README.md`	software	healthy	`aqr-finance/multigpu/README.md`
`qwen3-moe-fullweight-cpt.yaml`	software	healthy	`aqr-finance/multigpu/qwen3-moe-fullweight-cpt.yaml`
`volume_inspect_submit.sh`	software	healthy	`aqr-finance/multigpu/volume_inspect_submit.sh`
`benchmarks.md`	software	healthy	`aqr-finance/multigpu/benchmarks.md`
`stage3_real_v2_submit.sh`	software	healthy	`aqr-finance/multigpu/stage3_real_v2_submit.sh`
`prepare_text.py`	software	healthy	`aqr-finance/multigpu/prepare_text.py`
`train.py`	software	healthy	`autoresearch/train.py`
`prepare.py`	software	healthy	`autoresearch/prepare.py`
`README.md`	software	healthy	`autoresearch/README.md`
`program.md`	software	healthy	`autoresearch/program.md`
`benchmarks.md`	software	healthy	`autoresearch/benchmarks.md`
`pyproject.toml`	software	healthy	`autoresearch/pyproject.toml`
`requirements.txt`	software	healthy	`autoresearch/requirements.txt`
`submit.sh`	software	healthy	`autoresearch/batch-job/submit.sh`
`prep.sh`	software	healthy	`autoresearch/batch-job/prep.sh`
`submit-async.sh`	software	healthy	`autoresearch/batch-job/submit-async.sh`
`wait-jobs.sh`	software	healthy	`autoresearch/batch-job/wait-jobs.sh`
`_lib.sh`	software	healthy	`autoresearch/batch-job/_lib.sh`

Showing first 50 of this kind. Full payload available via the JSON button at the top of the page.

Label	Layer	Status	Path
`_template`	software	healthy	`_template`
`batch-job`	software	healthy	`_template/batch-job`
`gpu-cost-benchmark`	software	healthy	`gpu-cost-benchmark`
`batch-job`	software	healthy	`gpu-cost-benchmark/batch-job`
`results`	software	healthy	`gpu-cost-benchmark/results`
`analysis`	software	healthy	`gpu-cost-benchmark/analysis`
`data_prep`	software	healthy	`gpu-cost-benchmark/data_prep`
`aqr-finance`	software	healthy	`aqr-finance`
`batch-job`	software	healthy	`aqr-finance/batch-job`
`multigpu`	software	healthy	`aqr-finance/multigpu`
`autoresearch`	software	healthy	`autoresearch`
`batch-job`	software	healthy	`autoresearch/batch-job`
`gemma4-finetuning`	software	healthy	`gemma4-finetuning`
`batch-job`	software	healthy	`gemma4-finetuning/batch-job`

Label	Layer	Status	Path
`TEFp8Trainer`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:163`
`NVMLSampler`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:215`
`StepLogger`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:313`
`_State`	software	healthy	`gpu-cost-benchmark/batch-job/train_sft.py:460`
`NaNWatchdog`	software	healthy	`aqr-finance/train.py:133`
`GPTConfig`	software	healthy	`autoresearch/train.py:33`
`CausalSelfAttention`	software	healthy	`autoresearch/train.py:61`
`MLP`	software	healthy	`autoresearch/train.py:99`
`Block`	software	healthy	`autoresearch/train.py:112`
`GPT`	software	healthy	`autoresearch/train.py:124`
`MuonAdamW`	software	healthy	`autoresearch/train.py:356`
`Tokenizer`	software	healthy	`autoresearch/prepare.py:209`

Label	Layer	Status	Path
`repobility-clone-w8jc58ly`	software	healthy	`/tmp/repobility-clone-w8jc58ly`

Label	Layer	Status	Path
`127.0.0.1`	network	healthy	`gpu-cost-benchmark/batch-job/serve_and_bench.sh`

Label	Layer	Status	Path
`gpu (detected)`	hardware	healthy	`README.md`

For AI agents: Voting guide (TP/FP) MCP manifest Stdio wrapper SARIF Integrate Findings queue Vote TP/FP on findings to calibrate the engine.

For AI agents + API integrations

Email me when this repo regresses

Free. We re-scan periodically; new criticals → your inbox. No signup required for the scan itself.

API access

This page is publicly accessible at: https://repobility.com/scan/b6a74a3b-e3de-4c7a-a525-31ea52512dc8/

To check status programmatically (no auth required):

curl -s https://repobility.com/api/v1/public/scan/b6a74a3b-e3de-4c7a-a525-31ea52512dc8/

Important — please don't re-submit the same URL repeatedly. The submission endpoint is idempotent: re-submitting the same git URL returns this same scan_token, not a new one. To re-scan this repo, sign up free and use the dashboard.

vessl-ai/vessl-cloud-cookbook

Complete repo analysis

function 50+

file 50+

directory 14

class 12

repo 1

service 1

gpu 1