Skip to main content

abstraction-model, in Python

Whether two strings from two hosts name one model. Hugging Face, Ollama, LM Studio and a directory on disk each write the same weights a different way, so an application that has already downloaded a model cannot tell that the one it is about to fetch is the same file.

Every token that is not known packaging is kept. The reverse rule — a list of noise to strip — collapses a fine-tune onto the model it was tuned from, and then a request for one is answered by the other. One module, standard library only, importing nothing of ours.

This page is the Python package. The Go implementation, the store side and what is UNPROVEN are on the repository.

Install

Not on PyPI, and the name on PyPI is not ours.

git clone https://github.com/openabstractions/abstraction-model
pip install ./abstraction-model/python

Python 3.9 or later. abstraction_model.py is one file with no imports of ours, so copying it into a _vendor/ directory of your own is an equally complete installation.

An example that runs

import abstraction_model as model

a = "bartowski/Qwen2.5-0.5B-Instruct-GGUF/Qwen2.5-0.5B-Instruct-IQ2_M.gguf"
b = "qwen2.5:0.5b-instruct-q4_K_M"

print(model.parse(a))
print(model.parse(b))
print("same model:", model.same(a, b))
print("same file: ", model.parse(a) == model.parse(b))

Two quantisations of one model are the same model and are still two files on disk, and that is the distinction the layer exists to make: same answers the routing question, equality of the parsed Name answers the storage one.

What an application calls

call what it does
parse(s) a Name with family and quant
family(s) the family alone: what survives packaging, host prefixes and file extensions
same(a, b) whether two host identifiers name one model for routing
shard(name) the shard a multi-file model's part belongs to, when the name says

What may break

  • This module answers about names and nothing else. It does not download, does not hash and never touches a store.
  • The algorithm is a judgment, not a standard. It is compared against the Go implementation over a shared corpus of identifiers, so the two agree; there is no third party to be right against.
  • No conformance verdict. No scenario in the suite cites this layer yet — what is proven and what is not.
  • Not on any package index, and no release carries an API stability promise. Pin a commit you have read.

Apache-2.0. See LICENSE.

Metadata

Release files for abstraction-model 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 abstraction-model 0.1.0
File Size Uploaded Provenance Actions
abstraction_model-0.1.0.tar.gz 8.5 kB With provenance Details for abstraction_model-0.1.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for abstraction-model 0.1.0
File Interpreter ABI Platform Actions
abstraction_model-0.1.0-py3-none-any.whl Python 3 none any Details for abstraction_model-0.1.0-py3-none-any.whl

Total release size: 17.6 kB

Release files / abstraction_model-0.1.0.tar.gz

Download URL abstraction_model-0.1.0.tar.gz
Size 8.5 kB
Tags Source
SHA-256 checksum
How to use checksums
a02aa09cb2147004682a324646fefe36873a0d4b83ebb713dedf43a84f50629a
BLAKE2b-256 checksum
How to use checksums
75105ced279cd433e747e46ae041aa325208329cbe58637e3628eb6932cdf7ee
Upload date
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.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Sep 10, 2026.

Transparency log

Release files / abstraction_model-0.1.0-py3-none-any.whl

Download URL abstraction_model-0.1.0-py3-none-any.whl
Size 9.1 kB
Tags Python 3
SHA-256 checksum
How to use checksums
3283f73665de720115455ba64798ccbe8bb87dc77d0a3a9ee84d3292ee64305b
BLAKE2b-256 checksum
How to use checksums
e61acda46e6c53bb800f194c80e2c93a9201dc93e54c285c019740f6d1d59177
Upload date
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.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Sep 10, 2026.

Transparency log

Security for abstraction-model 0.1.0

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.

Full provenance

Both files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for abstraction_model-0.1.0.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Sep 10, 2026.

Transparency log

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Release history Release notifications | RSS feed

This release

0.1.0 This release

2 release files Full provenance

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page