| # Hugging Face Public Release Guide |
|
|
| This package is pre-optimized for a public research repository. |
|
|
| ## One-click Windows publication |
|
|
| 1. Extract the ZIP to a normal folder. |
| 2. Double-click `PUBLISH_TO_HUGGINGFACE.bat`. |
| 3. Press Enter to use the default repository name `heaven-vector-compression-engine`, or type another name. |
| 4. Paste a Hugging Face token with **write** permission when prompted. |
| 5. The script creates/reuses a **public** repository under the account authenticated by that token and uploads the full release. |
|
|
| The token is not embedded in the ZIP and is not written into project files. |
|
|
| ## Why the repository is search/agent friendly |
|
|
| - Hugging Face YAML metadata is at the top of `README.md`. |
| - The title contains both the full project name and `HVCE`. |
| - Search terms are represented in metadata and prose without claiming unsupported performance. |
| - `AI_CONTEXT.md`, `llms.txt`, `research_manifest.json`, `codemeta.json`, `CITATION.cff`, and `CITATION.bib` provide high-signal machine-readable/agent-friendly entry points. |
| - Benchmark data and the generator are stored alongside the code. |
| - Claim boundaries and negative results are first-class files. |
|
|
| ## Suggested repository description |
|
|
| **Very early experimental lossless-compression and archiving research prototype combining exact generative recipes, reversible transforms, causal residual prediction, solid packing, deduplication and cross-file state reuse. May compress poorly; broad superiority is not claimed.** |
|
|
| ## Suggested search keywords |
|
|
| `lossless compression`, `data compression`, `archiver`, `hybrid compression`, `generative compression`, `delta compression`, `deduplication`, `reversible transforms`, `causal prediction`, `information theory`, `experimental`, `research prototype`. |
|
|
