Releases after v0.1.4 differ from the stock configuration in the ways listed
in CHANGES.md. From v0.2.0 the serving image is built by this
repository: Tinfoil's image plus one file that requires the metrics key on
vLLM's /metrics.
The v0.1.4 configuration serves Audn's Penclaw checkpoint using the
Maple Kimi v0.0.3 configuration.
It changes the primary checkpoint repository and revision, Modelwrap tuple,
pack schema, and vLLM model path. It uses the same eight B300 GPUs, serving
image, Inferact DSpark draft, and native vLLM bearer authentication. Through
v0.1.8 it answered to the stock API model name kimi-k3-1-pro. From v0.2.0
the API model name is kimi-k3-1-pro-uncensored. The signed release
identifies the actual checkpoint served under either name.
The snapshot manifest records the exact Audn checkpoint
revision and Hugging Face's checksums. The measured configuration pins each
checkpoint revision, Modelwrap tuple, and schema, along with the serving image
digest. The guest mounts the packs read-only and dm-verity authenticates
blocks as they are read against the pinned pack roots. Auditors can compare
the manifest's file hashes when independently validating the gated snapshot.
Pinned Hugging Face metadata ·
Tinfoil model weight verification
- Read
tinfoil-config.ymlat the published tag and check the primary and draft checkpoint revisions, Modelwrap tuples and schemas, runtime image digest, and paths passed to--modeland--speculative-config. - Verify the signed deployment artifacts and use the Tinfoil CLI with the exact release tag and deployment digest to verify the live enclave and TLS key.
- Auditors with Hugging Face access can download the pinned snapshot, verify its files, rebuild the schema-2 pack, and compare its root with the measured config.
The release, pack commitment, and snapshot manifest are public. Audn approval is required to download the primary tensor bytes and independently reproduce that commitment. The checkpoint is distributed under the Kimi K3.1-Pro license. Checkpoint license · Modelwrap reproduction
Pass the full root_offset_uuid tuple from the completed schema-2 wrap to
python3 render_config.py --mpk TUPLE. Run python3 render_config.py --check
before publishing. The renderer verifies that the template matches the pinned
stock configuration outside the four primary pack fields and the differences
listed in EDITS. The source names the image without a digest. The release
workflow builds the image, writes its digest into the configuration on a
release-only commit, tags that commit and measures it. The renderer checks the
configuration on the default branch and again, with the pinned digest, at the
release tag.
The Dockerfile starts from the exact image the stock
configuration pins and adds maple_metrics_auth.py
through scripts/install_metrics_auth.py.
The installer checks the vLLM version and the three places in vLLM's source the
guard depends on, and fails the build if any has changed. Tests in tests/
cover the guard and check that the Dockerfile, the configuration and the guard
agree.