shad0w

Compile the decisions your LLM already makes into a certified 2 MB table that answers in microseconds on a CPU.
It learns from your model's own answers, needs no labels, and tells you in writing how often it will disagree.

Python

Live demo · Docs · 2-minute video · Benchmarks

Your app sends small, typed decisions to a big model millions of times: which intent is this, which queue gets this ticket, is this spam, should this agent step run. Every call takes 20–1,500 ms and a round trip to someone else's GPU.

shad0w sits in front of that model like a shadow. It watches the answers, learns the decision, and compiles it into a tiny integer table. Before the table answers anything, it certifies how often it will disagree with your model. On public intent and topic benchmarks, it then served 45–71% of traffic itself, at under 5% certified disagreement and in microseconds. Everything else goes to your model, exactly as before.

import shad0w

sh = shad0w.Shadow("bundle/", teacher=ask_my_llm, log="teacher_log.jsonl")

sh.decide("my card was stolen yesterday")
# Decision(answer='lost_card', source='table', confidence=0.98, certified=True, latency_us=14)

~1 µs
_{per decision in C
7–11 µs JS · 15 µs Python}

45–71%
_{of LLM traffic served locally
at a certified 5% bound}

0 labels
_{learns from your model's
own logged answers}

2 MB
_{one file, no GPU, no network
256 KB still keeps 88.5%}

How it works

shadow → compile → certify → serve

Shadow. Your model keeps serving. Its answers are logged as (text, answer) pairs, either by the Shadow wrapper or by your own logging.
Compile. The log trains a hashed n‑gram table with int8 weights. There is no tokenizer and no GPU, and it takes seconds to minutes on a laptop.
Certify. On a held-out slice, a Learn‑then‑Test procedure picks the confidence threshold. Among answers the table serves, disagreement with your model is ≤ α (say 5%), with probability ≥ 1 − δ.
Serve. Certified answers come back in microseconds, in your process. Everything else goes to your model and is logged, so the share the table can take grows with traffic: 31% → 70% as AG News shadow traffic went from 2k to 100k.

Results

Measured on public datasets. Every number links to its raw result file on the benchmarks page.

	Result
LLM traffic answered locally (7 intent/topic tasks, 2 teachers), certified α = 5%	45–71%, realised disagreement 3.9–5.7% (mean of 3 seeds)
End-to-end accuracy (table + teacher) vs teacher alone	within 1 point on every task; +0.7 with Kev‑0.8B as teacher
Mean latency of the cascade, Kev‑0.8B → BANKING77	75 ms → 22 ms
Offload as shadow traffic grows (AG News, 2k → 100k decisions)	31% → 70% at α = 5%; 5% → 52% at 2%
Per decision	~1 µs C · 7–11 µs JavaScript · 15 µs Python · 0.19 ms over HTTP
Public benchmark (local‑jev‑bench, BANKING77, 20 options)	89.7% at 0.018 ms on 1 CPU thread; Kev‑0.8B 88% at 57.6 ms; Clef‑flash (9B) 97% at 1,274 ms
Out-of-scope queries (CLINC150) separated by confidence	AUROC 0.94
Reproduce on your laptop (`examples/shadow_demo.py`, MIT-licensed teacher, BANKING77)	69.4% answered locally at α = 5% (realised 4.6%); cascade 67.8% vs teacher 68.0%

Compared with what you would otherwise do

	Per decision	Runs on	Needs	Bound on what it serves
Keep calling the LLM / decision API	20–1,500 ms	GPU or vendor API	—	it is the reference
Semantic cache (GPTCache, gateway caches)	ms (embed + vector search)	embedding model + store	nothing	similarity threshold, no bound
Embedding router (e.g. semantic-router)	ms	embedding model	example utterances	threshold, no bound
Distil into a smaller LLM	10–100 ms	GPU	teacher outputs	none
Hand-built fastText / SetFit / Model2Vec	µs–ms	CPU	human labels	none
shad0w	1–15 µs	CPU, in-process, browser, edge	your model's logged answers	finite-sample, against your model

It is not new science. The certificate is Learn‑then‑Test, in the spirit of Trust or Escalate and BARGAIN. The table is fastText-shaped on purpose. What is new is the packaging: one workflow from shadow logs to a certified, KB-to-MB artifact, with the same decisions in C, Python and JavaScript.

Quick start

pip install "shad0wllm[compile]"        # the runtime alone is: pip install shad0wllm   (numpy only, C core included)

The package installs as shad0wllm and imports as shad0w. The command-line tool is shad0w.

1. Log your model's answers. Wrap the call you already make. With no bundle yet, every call goes to your model and is logged:

import shad0w

sh = shad0w.Shadow(None, teacher=ask_my_llm, question="intent", log="teacher_log.jsonl")
answer = sh.decide(text).answer          # = ask_my_llm(text), logged

2. Compile and certify once you have 1,000+ lines (10k–100k for high offload):

shad0w init                                        # starter schema.json: the options your model chooses from
shad0w shadow --schema schema.json --data teacher_log.jsonl --teacher my-llm-v3 --alpha 0.05 --out bundle/
shad0w report --bundle bundle/                      # the certificate, in plain words and numbers

3. Serve. Point the wrapper at the bundle. Certified answers come from the table and the rest from your model. 1% of certified answers are spot-checked against your model, so sh.stats() shows live disagreement:

sh = shad0w.Shadow("bundle/", teacher=ask_my_llm, log="teacher_log.jsonl")

4. Re-certify on fresh traffic every week, or when traffic changes:

shad0w certify --bundle bundle/ --data last_week.jsonl

Everywhere else


JavaScript (Node, browsers, Workers, Deno, Bun), zero dependencies	`npm i shad0wllm` → `const b = await Bundle.load("bundle/"); b.decide(text)`
HTTP sidecar, any language	`shad0w serve --bundle bundle/` → `POST /v1/decide`
C, any language with an FFI	`shad0w/_native/reflex.c`: `s0_load`, `s0_decide`
Decorator	`@shad0w.cascade("bundle/")` on your existing classify function
Human labels instead of a teacher	`shad0w compile --data labels.jsonl` + `shad0w calibrate` (~300 labels)
No teacher, no labels	`shad0w compile --unlabeled logs.txt` (option names + raw logs, `pip install "shad0wllm[logs]"`)

Integration recipes for FastAPI, LiteLLM, LangGraph, Cloudflare Workers and the browser are in the docs.

When not to use it

The limits below are measured. Knowing them is what makes the certificate useful.

Sentiment, tone, emotion. A hashed table tops out near 79% on SST‑2 even with perfect labels. shad0w certifies almost nothing there and defers everything to your model, which is the correct behaviour.
When you need the most accurate answer. A 9B decision model is 7 points better on the public BANKING77 file, at ~70,000× the latency. Put shad0w in front of such a model, not instead of it.
The bound is against your model, not the truth. A certified table in front of a wrong model is wrong in the same way. For a bound against the truth, calibrate with ~300 real labels.
The bound holds on traffic like the calibration slice. A big shift in the request mix can break it. The drift guard flags confidence shifts, the audits watch live disagreement, and shad0w certify re-certifies.
Reasoning or world knowledge. It learns surface decisions: intents, routes, topics, policies with distinctive wording.

Documentation

Getting started · Shadow mode guide · Integrations · The certificate, explained · API & CLI · Architecture & file format · Benchmarks · Limitations · FAQ

Contributing

Run it on one of your decisions and share your numbers, including failures. That is the most valuable contribution. Code contributions: see CONTRIBUTING.md.

Citation

@software{chauhan2026shad0w,
  author = {Chauhan, Gaurav},
  title  = {shad0w: certified microsecond decisions compiled from the model you already run},
  year   = {2026},
  url    = {https://github.com/2796gaurav/shad0w}
}

Built by Gaurav Chauhan · Apache-2.0

Metadata

Release files for shad0wllm 0.1.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for shad0wllm 0.1.0
File	Size	Uploaded
shad0wllm-0.1.0.tar.gz	52.8 kB	Oct 7, 2026	Details

Built distributions (wheels)

Table of built distributions (wheels) for shad0wllm 0.1.0
File	Interpreter	ABI	Platform	Reset
shad0wllm-0.1.0-cp313-cp313-win_amd64.whl 54.5 kB Oct 7, 2026	CPython 3.13	CPython 3.13	Windows x86-64	Details
shad0wllm-0.1.0-cp313-cp313-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl 67.2 kB Oct 7, 2026	CPython 3.13	CPython 3.13	Linux glibc 2.27+ x86-64, Linux glibc 2.28+ x86-64	Details
shad0wllm-0.1.0-cp313-cp313-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl 68.6 kB Oct 7, 2026	CPython 3.13	CPython 3.13	Linux glibc 2.27+ ARM64, Linux glibc 2.28+ ARM64	Details
shad0wllm-0.1.0-cp313-cp313-macosx_11_0_arm64.whl 52.0 kB Oct 7, 2026	CPython 3.13	CPython 3.13	macOS 11.0+ ARM64	Details
shad0wllm-0.1.0-cp313-cp313-macosx_10_13_x86_64.whl 51.7 kB Oct 7, 2026	CPython 3.13	CPython 3.13	macOS 10.13+ x86-64	Details
shad0wllm-0.1.0-cp312-cp312-win_amd64.whl 54.5 kB Oct 7, 2026	CPython 3.12	CPython 3.12	Windows x86-64	Details
shad0wllm-0.1.0-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl 67.1 kB Oct 7, 2026	CPython 3.12	CPython 3.12	Linux glibc 2.27+ x86-64, Linux glibc 2.28+ x86-64	Details
shad0wllm-0.1.0-cp312-cp312-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl 68.5 kB Oct 7, 2026	CPython 3.12	CPython 3.12	Linux glibc 2.28+ ARM64, Linux glibc 2.27+ ARM64	Details
shad0wllm-0.1.0-cp312-cp312-macosx_11_0_arm64.whl 52.0 kB Oct 7, 2026	CPython 3.12	CPython 3.12	macOS 11.0+ ARM64	Details
shad0wllm-0.1.0-cp312-cp312-macosx_10_13_x86_64.whl 51.7 kB Oct 7, 2026	CPython 3.12	CPython 3.12	macOS 10.13+ x86-64	Details
shad0wllm-0.1.0-cp311-cp311-win_amd64.whl 54.5 kB Oct 7, 2026	CPython 3.11	CPython 3.11	Windows x86-64	Details
shad0wllm-0.1.0-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl 67.0 kB Oct 7, 2026	CPython 3.11	CPython 3.11	Linux glibc 2.27+ x86-64, Linux glibc 2.28+ x86-64	Details
shad0wllm-0.1.0-cp311-cp311-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl 68.4 kB Oct 7, 2026	CPython 3.11	CPython 3.11	Linux glibc 2.28+ ARM64, Linux glibc 2.27+ ARM64	Details
shad0wllm-0.1.0-cp311-cp311-macosx_11_0_arm64.whl 52.0 kB Oct 7, 2026	CPython 3.11	CPython 3.11	macOS 11.0+ ARM64	Details
shad0wllm-0.1.0-cp311-cp311-macosx_10_9_x86_64.whl 51.8 kB Oct 7, 2026	CPython 3.11	CPython 3.11	macOS 10.9+ x86-64	Details
shad0wllm-0.1.0-cp310-cp310-win_amd64.whl 54.5 kB Oct 7, 2026	CPython 3.10	CPython 3.10	Windows x86-64	Details
shad0wllm-0.1.0-cp310-cp310-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl 66.9 kB Oct 7, 2026	CPython 3.10	CPython 3.10	Linux glibc 2.27+ x86-64, Linux glibc 2.28+ x86-64	Details
shad0wllm-0.1.0-cp310-cp310-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl 68.4 kB Oct 7, 2026	CPython 3.10	CPython 3.10	Linux glibc 2.28+ ARM64, Linux glibc 2.27+ ARM64	Details
shad0wllm-0.1.0-cp310-cp310-macosx_11_0_arm64.whl 52.0 kB Oct 7, 2026	CPython 3.10	CPython 3.10	macOS 11.0+ ARM64	Details
shad0wllm-0.1.0-cp310-cp310-macosx_10_9_x86_64.whl 51.8 kB Oct 7, 2026	CPython 3.10	CPython 3.10	macOS 10.9+ x86-64	Details

Total release size: 1.2 MB

Release files / shad0wllm-0.1.0.tar.gz

Download URL	shad0wllm-0.1.0.tar.gz
Size	52.8 kB
Tags	Source
SHA-256 checksum How to use checksums	`a1ef58a3b87afdbbbf3bbb1cec5c0c0ba4382b0a0792b550c7a3108329c1e24d`
BLAKE2b-256 checksum How to use checksums	`e1c2c52933ff13a91e7ae836f57ab8bcfcf434cf1f6d643a2f00f48233e933f4`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0.tar.gz
SHA-256 checksum: a1ef58a3b87afdbbbf3bbb1cec5c0c0ba4382b0a0792b550c7a3108329c1e24d Verifying attestations

Release files / shad0wllm-0.1.0-cp313-cp313-win_amd64.whl

Download URL	shad0wllm-0.1.0-cp313-cp313-win_amd64.whl
Size	54.5 kB
Tags	CPython 3.13 Windows x86-64
SHA-256 checksum How to use checksums	`343cda3fd080ed4fc9a16893ccd84f4c1290a2fa61f2f513e6e1589361294f09`
BLAKE2b-256 checksum How to use checksums	`857401d853c1915a233433de3c1743b4744e8e5669f47e1921bf645701bf032c`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp313-cp313-win_amd64.whl
SHA-256 checksum: 343cda3fd080ed4fc9a16893ccd84f4c1290a2fa61f2f513e6e1589361294f09 Verifying attestations

Release files / shad0wllm-0.1.0-cp313-cp313-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl

Download URL	shad0wllm-0.1.0-cp313-cp313-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
Size	67.2 kB
Tags	CPython 3.13 Linux glibc 2.27+ x86-64 Linux glibc 2.28+ x86-64
SHA-256 checksum How to use checksums	`26dc0fda14d602dcd2304e039730dda7c2727547a3a1d251c2290b4266ff2515`
BLAKE2b-256 checksum How to use checksums	`1761b08acec5573860f0f7cc9c99c1c10035213c9830f4a2fb2ddb9a48869abd`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp313-cp313-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
SHA-256 checksum: 26dc0fda14d602dcd2304e039730dda7c2727547a3a1d251c2290b4266ff2515 Verifying attestations

Release files / shad0wllm-0.1.0-cp313-cp313-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl

Download URL	shad0wllm-0.1.0-cp313-cp313-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl
Size	68.6 kB
Tags	CPython 3.13 Linux glibc 2.27+ ARM64 Linux glibc 2.28+ ARM64
SHA-256 checksum How to use checksums	`28c53535424fff7cb99baa792710d9ee4b400449344d57420b2db0465fb71ade`
BLAKE2b-256 checksum How to use checksums	`1f2192a649fa4d8de3460f4fae687a9870292620a5a25e0173f04b48bee941a7`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp313-cp313-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl
SHA-256 checksum: 28c53535424fff7cb99baa792710d9ee4b400449344d57420b2db0465fb71ade Verifying attestations

Release files / shad0wllm-0.1.0-cp313-cp313-macosx_11_0_arm64.whl

Download URL	shad0wllm-0.1.0-cp313-cp313-macosx_11_0_arm64.whl
Size	52.0 kB
Tags	CPython 3.13 macOS 11.0+ ARM64
SHA-256 checksum How to use checksums	`a4dd7f568c2be3ef7a7c779fdeee47c491405c4089854ddd13d056608d8e9a76`
BLAKE2b-256 checksum How to use checksums	`a3a2a821aab4e964b30c592a753cfce76aaa46a65d04709c92fc3e9568c43846`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp313-cp313-macosx_11_0_arm64.whl
SHA-256 checksum: a4dd7f568c2be3ef7a7c779fdeee47c491405c4089854ddd13d056608d8e9a76 Verifying attestations

Release files / shad0wllm-0.1.0-cp313-cp313-macosx_10_13_x86_64.whl

Download URL	shad0wllm-0.1.0-cp313-cp313-macosx_10_13_x86_64.whl
Size	51.7 kB
Tags	CPython 3.13 macOS 10.13+ x86-64
SHA-256 checksum How to use checksums	`2f6a67d758131f539d4a5bd4236b39bf56e14216991ca5c98b08c4179eef8fd9`
BLAKE2b-256 checksum How to use checksums	`8982dcc1afe87c35cbe83450c1aa05e6500f33b44e5e8b8a77df8fd6f590273f`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp313-cp313-macosx_10_13_x86_64.whl
SHA-256 checksum: 2f6a67d758131f539d4a5bd4236b39bf56e14216991ca5c98b08c4179eef8fd9 Verifying attestations

Release files / shad0wllm-0.1.0-cp312-cp312-win_amd64.whl

Download URL	shad0wllm-0.1.0-cp312-cp312-win_amd64.whl
Size	54.5 kB
Tags	CPython 3.12 Windows x86-64
SHA-256 checksum How to use checksums	`a185cfb821ec1abbf1f808e4e100b1de464cd496f608b930992e0d51fc45f6cb`
BLAKE2b-256 checksum How to use checksums	`8137a3b2c3e148af70086d26db83e6d3ece945a7845aa0a6d2bbe6ca23b7a94d`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp312-cp312-win_amd64.whl
SHA-256 checksum: a185cfb821ec1abbf1f808e4e100b1de464cd496f608b930992e0d51fc45f6cb Verifying attestations

Release files / shad0wllm-0.1.0-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl

Download URL	shad0wllm-0.1.0-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
Size	67.1 kB
Tags	CPython 3.12 Linux glibc 2.27+ x86-64 Linux glibc 2.28+ x86-64
SHA-256 checksum How to use checksums	`5bb39574ae79fdff9436ea3fc83270dea3fba0b9b3119237f010589ab9851ca8`
BLAKE2b-256 checksum How to use checksums	`9a0055bb5c28f3c41acb6bcdece1da1f1626c54dfcac2732afdec6629d7979f1`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
SHA-256 checksum: 5bb39574ae79fdff9436ea3fc83270dea3fba0b9b3119237f010589ab9851ca8 Verifying attestations

Release files / shad0wllm-0.1.0-cp312-cp312-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl

Download URL	shad0wllm-0.1.0-cp312-cp312-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl
Size	68.5 kB
Tags	CPython 3.12 Linux glibc 2.27+ ARM64 Linux glibc 2.28+ ARM64
SHA-256 checksum How to use checksums	`0f136461a2e17eba700db4a73677897bceb23b97196ac56f7ffe38e88f30e281`
BLAKE2b-256 checksum How to use checksums	`6f80be432d96c6871319265afb44f65fdade76cd2c89e3cb0a0819200f949347`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp312-cp312-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl
SHA-256 checksum: 0f136461a2e17eba700db4a73677897bceb23b97196ac56f7ffe38e88f30e281 Verifying attestations

Release files / shad0wllm-0.1.0-cp312-cp312-macosx_11_0_arm64.whl

Download URL	shad0wllm-0.1.0-cp312-cp312-macosx_11_0_arm64.whl
Size	52.0 kB
Tags	CPython 3.12 macOS 11.0+ ARM64
SHA-256 checksum How to use checksums	`20dc893299de8050017190a80057d552b070519b225c90c0364c2142bef3f74a`
BLAKE2b-256 checksum How to use checksums	`b412a079b18cddf2a5344ef62850ca4edb459a3669d706f120b322a985b9fd47`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp312-cp312-macosx_11_0_arm64.whl
SHA-256 checksum: 20dc893299de8050017190a80057d552b070519b225c90c0364c2142bef3f74a Verifying attestations

Release files / shad0wllm-0.1.0-cp312-cp312-macosx_10_13_x86_64.whl

Download URL	shad0wllm-0.1.0-cp312-cp312-macosx_10_13_x86_64.whl
Size	51.7 kB
Tags	CPython 3.12 macOS 10.13+ x86-64
SHA-256 checksum How to use checksums	`311974a379d3e2d2aa679d9e27fb15b4087ee9f30279ee3cb3859fa774b854ec`
BLAKE2b-256 checksum How to use checksums	`04829c5dd3d2a4dbb6b6f65eb8ac55e979dd07e9da2b792759cf6272784d8d2b`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp312-cp312-macosx_10_13_x86_64.whl
SHA-256 checksum: 311974a379d3e2d2aa679d9e27fb15b4087ee9f30279ee3cb3859fa774b854ec Verifying attestations

Release files / shad0wllm-0.1.0-cp311-cp311-win_amd64.whl

Download URL	shad0wllm-0.1.0-cp311-cp311-win_amd64.whl
Size	54.5 kB
Tags	CPython 3.11 Windows x86-64
SHA-256 checksum How to use checksums	`2f47720ba6e665a026ebf9d249b4eb4acc46dd21ba35fcb96f1269265a1ab051`
BLAKE2b-256 checksum How to use checksums	`baf82fc3cec359430d41ba98ee6af835b0365f914276b9a271153d35ee5dbe4b`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp311-cp311-win_amd64.whl
SHA-256 checksum: 2f47720ba6e665a026ebf9d249b4eb4acc46dd21ba35fcb96f1269265a1ab051 Verifying attestations

Release files / shad0wllm-0.1.0-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl

Download URL	shad0wllm-0.1.0-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
Size	67.0 kB
Tags	CPython 3.11 Linux glibc 2.27+ x86-64 Linux glibc 2.28+ x86-64
SHA-256 checksum How to use checksums	`ac4bf283cbac7392e1d6de2f39ff77ada9216c8bdede699a3666e1daff7962cd`
BLAKE2b-256 checksum How to use checksums	`73adace0ffc4b73d08dd4a998820b4a48c03f90a076907f3112dfd9d3ddacb8a`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp311-cp311-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
SHA-256 checksum: ac4bf283cbac7392e1d6de2f39ff77ada9216c8bdede699a3666e1daff7962cd Verifying attestations

Release files / shad0wllm-0.1.0-cp311-cp311-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl

Download URL	shad0wllm-0.1.0-cp311-cp311-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl
Size	68.4 kB
Tags	CPython 3.11 Linux glibc 2.27+ ARM64 Linux glibc 2.28+ ARM64
SHA-256 checksum How to use checksums	`ace71217ea9a34c7eb82891660141424d83892fedf06c499f3fa4f6e9d687d94`
BLAKE2b-256 checksum How to use checksums	`16cc13f26c14bba539f2864a1c0acebc1843d73446474ffda38af36e8bdc072a`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp311-cp311-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl
SHA-256 checksum: ace71217ea9a34c7eb82891660141424d83892fedf06c499f3fa4f6e9d687d94 Verifying attestations

Release files / shad0wllm-0.1.0-cp311-cp311-macosx_11_0_arm64.whl

Download URL	shad0wllm-0.1.0-cp311-cp311-macosx_11_0_arm64.whl
Size	52.0 kB
Tags	CPython 3.11 macOS 11.0+ ARM64
SHA-256 checksum How to use checksums	`850ce0e50b8343cfe05cf4252f2165e0f5a3d0b65699f45eb9d1a0fb5a1d1866`
BLAKE2b-256 checksum How to use checksums	`44ea24446db8d26aaa8fc081bc40bef7a8cc41d580401023478ffad03a725aff`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp311-cp311-macosx_11_0_arm64.whl
SHA-256 checksum: 850ce0e50b8343cfe05cf4252f2165e0f5a3d0b65699f45eb9d1a0fb5a1d1866 Verifying attestations

Release files / shad0wllm-0.1.0-cp311-cp311-macosx_10_9_x86_64.whl

Download URL	shad0wllm-0.1.0-cp311-cp311-macosx_10_9_x86_64.whl
Size	51.8 kB
Tags	CPython 3.11 macOS 10.9+ x86-64
SHA-256 checksum How to use checksums	`c01ee9b87219eaf4d7fa68e51326086b874f598bee05d6ead24251bafb13b6a9`
BLAKE2b-256 checksum How to use checksums	`1a918e6ba8260ea0ff0be636ea9f1a4546b8556937b715a2edd1407712acc3ef`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp311-cp311-macosx_10_9_x86_64.whl
SHA-256 checksum: c01ee9b87219eaf4d7fa68e51326086b874f598bee05d6ead24251bafb13b6a9 Verifying attestations

Release files / shad0wllm-0.1.0-cp310-cp310-win_amd64.whl

Download URL	shad0wllm-0.1.0-cp310-cp310-win_amd64.whl
Size	54.5 kB
Tags	CPython 3.10 Windows x86-64
SHA-256 checksum How to use checksums	`05799489cc5a980d13722b9a93048b4cf652e29c40d50376cb1b92c65e888b76`
BLAKE2b-256 checksum How to use checksums	`8d06b16fa427a3aab3d68b46e5285afc42c5430957b2758ef9a660ff071cb93b`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp310-cp310-win_amd64.whl
SHA-256 checksum: 05799489cc5a980d13722b9a93048b4cf652e29c40d50376cb1b92c65e888b76 Verifying attestations

Release files / shad0wllm-0.1.0-cp310-cp310-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl

Download URL	shad0wllm-0.1.0-cp310-cp310-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
Size	66.9 kB
Tags	CPython 3.10 Linux glibc 2.27+ x86-64 Linux glibc 2.28+ x86-64
SHA-256 checksum How to use checksums	`dd38d8d902b128048fbca183eee5465b25aa571dbac69af6decb7ce780aacda1`
BLAKE2b-256 checksum How to use checksums	`4d3fba63afb6bd778fae7b772fff2b504a5d79fa12bd0decce636ce488a95f04`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp310-cp310-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl
SHA-256 checksum: dd38d8d902b128048fbca183eee5465b25aa571dbac69af6decb7ce780aacda1 Verifying attestations

Release files / shad0wllm-0.1.0-cp310-cp310-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl

Download URL	shad0wllm-0.1.0-cp310-cp310-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl
Size	68.4 kB
Tags	CPython 3.10 Linux glibc 2.27+ ARM64 Linux glibc 2.28+ ARM64
SHA-256 checksum How to use checksums	`42355fd1bfa03cb947b3042509b1d36f775a03f598caf87b50959f15476b7ec0`
BLAKE2b-256 checksum How to use checksums	`166675791cbdeca89000e98f5a6102b7551cb94df4b75d9542336df60e6c2934`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp310-cp310-manylinux_2_27_aarch64.manylinux_2_28_aarch64.whl
SHA-256 checksum: 42355fd1bfa03cb947b3042509b1d36f775a03f598caf87b50959f15476b7ec0 Verifying attestations

Release files / shad0wllm-0.1.0-cp310-cp310-macosx_11_0_arm64.whl

Download URL	shad0wllm-0.1.0-cp310-cp310-macosx_11_0_arm64.whl
Size	52.0 kB
Tags	CPython 3.10 macOS 11.0+ ARM64
SHA-256 checksum How to use checksums	`2f9b723ced5d525b6891b5b1c64d49c046d5674ed733298f77b2f8c92f22abec`
BLAKE2b-256 checksum How to use checksums	`82a63fb13ba82ec3ba9a24dba4b825168de4e9c95530e85a4c00df0ffea4c4be`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp310-cp310-macosx_11_0_arm64.whl
SHA-256 checksum: 2f9b723ced5d525b6891b5b1c64d49c046d5674ed733298f77b2f8c92f22abec Verifying attestations

Release files / shad0wllm-0.1.0-cp310-cp310-macosx_10_9_x86_64.whl

Download URL	shad0wllm-0.1.0-cp310-cp310-macosx_10_9_x86_64.whl
Size	51.8 kB
Tags	CPython 3.10 macOS 10.9+ x86-64
SHA-256 checksum How to use checksums	`480ec4fa44704a3dbde40289f3f8e5b3756577c25b783d2c5038f6dbbbe272fc`
BLAKE2b-256 checksum How to use checksums	`60601b481f78f3d49983d6fbffbb4cdab6153f3648537bb6acfc3b985dba71e6`
Upload date	Oct 7, 2026
Uploaded using Trusted Publishing? What is trusted publishing?	Yes
Uploaded via	`twine/7.0.0 CPython/3.13.14`

Provenance

Identity

Publishing platform: GitHub Actions
Publishing repository: github.com/2796gaurav/shad0w
Publishing commit: github.com/2796gaurav/shad0w/tree/6b7aff6bb3249a865624d0e97ffaf86871005101

Workflow

Publishing configuration: github.com/2796gaurav/shad0w/blob/6b7aff6bb3249a865624d0e97ffaf86871005101/.github/workflows/release.yml
Publishing logs: github.com/2796gaurav/shad0w/actions/runs/37576524549/attempts/1

Artifact

Subject: shad0wllm-0.1.0-cp310-cp310-macosx_10_9_x86_64.whl
SHA-256 checksum: 480ec4fa44704a3dbde40289f3f8e5b3756577c25b783d2c5038f6dbbbe272fc Verifying attestations