Hugging Bay Release Template
Copy a Hugging Bay release manifest template with source URLs, SHA-256 hashes, runtime notes, cross-source provenance, signed trust-bundle inputs, and reviewed fallback metadata fields.
Copy this template when a lab or creator wants to release an open AI artifact with namespace claim, provenance, hashes, runtime notes, signed manifests, and optional reviewed peer fallback metadata.
What labs fill in
- owner matches the approved publisher namespace
- sourceUrl is HTTPS and points to the official public release
- license is redistributable and consistent across official, HF, ModelScope, and GitHub rows
- sourceFiles include exact sizeBytes and SHA-256 for every hosted file
- runtimeNotes answer whether users can run the release on Mac, consumer GPU, server GPU, or local runners
- crossSourceEvidence links to the official release, Hugging Face, ModelScope, GitHub, or release blog when available
- reviewed peer fallback metadata stays pending until hosted-file hashes, source URL, license, scan, and reviewer evidence pass
Template commands
- fetchTemplate:
curl -fsSL https://huggingbay.xyz/api/publishers/release-template?kind=llm > hugging-bay-release-template.json - apply:
bay publisher-apply application.json --api https://huggingbay.xyz - createUploadSession:
bay upload-session hugging-bay-release-template.json --files-dir ./release-files --write-draft release-draft.json --token <publisher-token> --api https://huggingbay.xyz - publishSmallRelease:
bay publish hugging-bay-release-template.json --files-dir ./release-files --token <publisher-token> --api https://huggingbay.xyz - checkRelease:
bay check your-open-lab/your-open-model --token <publisher-token> --api https://huggingbay.xyz - fetchReleaseBundle:
curl -fsSL https://huggingbay.xyz/api/releases/your-open-lab-your-open-model
Why release with Hugging Bay
- one canonical launch page that humans, answer engines, and agents can cite
- signed manifests and SHA-256 evidence before users download files
- same-as-upstream provenance from cross-source URLs and hosted hash evidence
- local-run commands for Ollama, LM Studio, vLLM, Transformers, and llama.cpp where applicable
- metadata coverage for Western and Eastern release surfaces such as Hugging Face, ModelScope, GitHub, and official blogs
Machine-readable template
Release template JSON Launch kit JSON Publisher workflow JSON Apply to publish
{
"owner": "your-open-lab",
"name": "your-open-model",
"type": "llm",
"license": "apache-2.0",
"visibility": "draft",
"sourceUrl": "https://your-lab.org/releases/your-open-model",
"summary": "One-sentence release summary for search, cards, answer engines, and social shares.",
"description": "What this release is, who it is for, and what changed from prior versions.\n\nInclude intended use, limitations, training/eval notes when public, and any safety or license caveats.",
"provenance": "Publisher attests these files are official public release artifacts and may be indexed and distributed under the declared license.",
"tags": [
"open-weights",
"local-run",
"transformers"
],
"worksWith": [
"transformers",
"ollama",
"lmstudio",
"vllm",
"llamacpp"
],
"upstream": {
"official": "https://your-lab.org/releases/your-open-model",
"huggingFace": "https://huggingface.co/your-open-lab/your-open-model",
"modelScope": "https://modelscope.cn/models/your-open-lab/your-open-model",
"github": "https://github.com/your-open-lab/your-open-model",
"blog": "https://your-lab.org/blog/your-open-model"
},
"runtimeNotes": {
"recommendedQuant": "Q4_K_M or BF16 depending on hardware",
"minimumVramGb": 8,
"recommendedVramGb": 24,
"mac": "Runs through llama.cpp-compatible GGUF when provided.",
"gpu": "Use vLLM or Transformers for safetensors/BF16 rows.",
"commands": [
{
"tool": "ollama",
"label": "Ollama",
"command": "ollama run your-open-model"
},
{
"tool": "lmstudio",
"label": "LM Studio",
"command": "Search Hugging Bay for the release, then import the hosted GGUF or upstream repo."
},
{
"tool": "vllm",
"label": "vLLM",
"command": "vllm serve your-open-lab/your-open-model --trust-remote-code false"
},
{
"tool": "llamacpp",
"label": "llama.cpp",
"command": "./llama-cli -m your-open-model.Q4_K_M.gguf -st -p \"Hello\""
}
]
},
"sourceFiles": [
{
"path": "README.md",
"sizeBytes": 25000,
"sha256": "replace-with-64-character-sha256",
"contentType": "text/markdown"
},
{
"path": "config.json",
"sizeBytes": 4096,
"sha256": "replace-with-64-character-sha256",
"contentType": "application/json"
},
{
"path": "model.safetensors",
"sizeBytes": 1234567890,
"sha256": "replace-with-64-character-sha256",
"contentType": "application/octet-stream"
}
],
"crossSourceEvidence": [
{
"source": "official",
"url": "https://your-lab.org/releases/your-open-model",
"note": "Official announcement names the license, version, and file list."
},
{
"source": "huggingface",
"url": "https://huggingface.co/your-open-lab/your-open-model",
"note": "HF repo mirrors the same release name and license."
},
{
"source": "modelscope",
"url": "https://modelscope.cn/models/your-open-lab/your-open-model",
"note": "ModelScope row for Chinese/Eastern discoverability when available."
},
{
"source": "github",
"url": "https://github.com/your-open-lab/your-open-model",
"note": "Inference code, tags, and license."
}
],
"reviewedPeerFallback": {
"requestOnlyAfterHosted": true,
"magnetUri": "magnet:?xt=urn:btih:<info-hash>&dn=your-open-model",
"sourceUrl": "https://your-lab.org/releases/your-open-model/your-open-model.torrent",
"filePath": "model.safetensors",
"useCase": "community resilience fallback for an already hosted, hash-verified public release",
"reviewerEvidenceRequired": true
}
}