Metadata-Version: 2.4
Name: fraqtl
Version: 0.1.0
Summary: fraQtl — compression and inference optimization for LLMs. Meta-package: installs the diagnostic and points to the fraQtl toolchain.
Project-URL: Homepage, https://www.fraqtl.ai
Project-URL: Models, https://huggingface.co/fraQtl
Author-email: "Samuel M.A Salfati" <contact@fraqtl.ai>
License: Apache-2.0
Keywords: compression,inference,kv-cache,llm,quantization
Requires-Python: >=3.9
Requires-Dist: fraqtl-diagnostic
Description-Content-Type: text/markdown

# fraQtl

Compression and inference optimization for LLMs — with receipts.

This is the umbrella package for the fraQtl toolchain. Installing it gives you
[`fraqtl-diagnostic`](https://pypi.org/project/fraqtl-diagnostic/) and a
`fraqtl` command that describes the rest of the toolchain:

| Package | What it does |
|---|---|
| [`fraqtl-diagnostic`](https://pypi.org/project/fraqtl-diagnostic/) | Fingerprint any transformer's compression potential in ~30 minutes |
| [`fraqtl-runtime`](https://pypi.org/project/fraqtl-runtime/) | Drop-in loader for fraQtl-compressed Hugging Face checkpoints |
| [`cairn-security-agent-audit`](https://pypi.org/project/cairn-security-agent-audit/) | Audit AI-agent traces for safely reusable work, with measured false-hit rates |

**Receipts before claims:** long-context KV-cache compression benchmarked at
128K context under multi-user concurrency, retrieval-gated, with a public
repro kit — see [huggingface.co/fraQtl](https://huggingface.co/fraQtl) and
[fraqtl.ai](https://www.fraqtl.ai).

Contact: contact@fraqtl.ai
