Publish dataset card, citation, and provenance manifest

CITATION.cff

@@ -0,0 +1,23 @@
+cff-version: 1.2.0
+message: "If you use this collection, please cite it as below."
+title: "llama-cpp-wheels: the complete prebuilt llama-cpp-python wheel collection"
+abstract: >-
+  A continuously maintained collection of prebuilt llama-cpp-python wheels
+  spanning every release (0.1.30 onward) across Linux, Windows and macOS,
+  all supported Python versions, and CPU / CUDA / Metal / Vulkan / OpenCL /
+  BLAS backends. Built and kept current by an automated CI "factory".
+type: dataset
+authors:
+  - name: "AIencoder"
+url: "https://huggingface.co/datasets/AIencoder/llama-cpp-wheels"
+repository: "https://huggingface.co/datasets/AIencoder/llama-cpp-wheels"
+license: MIT
+keywords:
+  - llama.cpp
+  - llama-cpp-python
+  - python-wheels
+  - prebuilt-binaries
+  - cuda
+  - metal
+  - inference
+date-released: 2026-10-31

README.md

@@ -264,4 +264,24 @@ The non-mirror wheels are compiled automatically by a GitHub Actions "factory":
 
 *Maintained by [AIencoder](https://huggingface.co/AIencoder). Wheels are builds of
 [abetlen/llama-cpp-python](https://github.com/abetlen/llama-cpp-python); all credit for llama-cpp-python and
-llama.cpp to their authors.*
+llama.cpp to their authors.*
+
+## Citation & attribution
+
+This collection is **curated, built, and continuously maintained by [AIencoder](https://huggingface.co/AIencoder)**. The wheels themselves are builds of [llama-cpp-python](https://github.com/abetlen/llama-cpp-python) by Andrei Betlen and [llama.cpp](https://github.com/ggerganov/llama.cpp) by Georgi Gerganov, both under the MIT license — their copyright and license notices are retained inside each wheel. The **selection, build automation, and layout of this collection** are my own contribution.
+
+The license does not *require* you to credit this collection, but if it saved you time, a citation is appreciated and helps keep it discoverable:
+
+> AIencoder. *llama-cpp-wheels: the complete prebuilt llama-cpp-python wheel collection.* Hugging Face, 2026. https://huggingface.co/datasets/AIencoder/llama-cpp-wheels
+
+```bibtex
+@misc{aiencoder_llama_cpp_wheels,
+  author       = {{AIencoder}},
+  title        = {llama-cpp-wheels: the complete prebuilt llama-cpp-python wheel collection},
+  year         = {2026},
+  howpublished = {Hugging Face dataset},
+  url          = {https://huggingface.co/datasets/AIencoder/llama-cpp-wheels}
+}
+```
+
+Each wheel's provenance (SHA256 + Xet hash) is recorded in [`MANIFEST.tsv`](./MANIFEST.tsv), a dated receipt of the full collection.