Repository navigation
chore: sync with internal pipette-clients at df54b1a4 - #32
Merged
Merged
Conversation
ykhrustalev
requested review from
iamstuffed,
m3m0ryh0l3 and
ovenmitt
as code owners
November 16, 2026 20:27
**Problem** This public repository lagged 25 commits behind the internal one, from `be6ca299` (the #30 sync) to `df54b1a4`. **Solution** - Reports readiness timeouts, crashes and out of memory as final failures, and a readiness timeout as `device_not_ready` - Adds `--cdn-proxy` (`PIPETTE_CDN_PROXY`) to fetch models through a caching proxy by `Host`, on the CLI and the iOS app - Advertises cached artifacts in worker claims, and leaves server starts out of `attempt_timeout_seconds` - Fetches the CUDA runtime archive beside a Windows CUDA llama.cpp build, and ships an arm64 Windows MSI of the GPU keepalive - Stops llama.cpp programs with SIGTERM before SIGKILL, and drops llama-bench's own warm-up in prefill, peak-memory and CPU-only decode cells - Replaces cargo-machete with cargo's unused-dependency lints, and builds the llama.cpp integration tests as one binary - Pins `pipette-plan-types` to public `pipette-mgmt` at `06471959`, which carries `device_not_ready` - Keeps the public scrubs: no private llama.cpp variants, private CI steps or runners, TensorRT Edge-LLM docs, private repository links, lab host names or release tags - Leaves the private proxy's repository name out of the CDN proxy docs, and TensorRT Edge-LLM out of the new attempt-clock docs **Testing** - `cargo nextest run --workspace --locked --no-fail-fast`: 6,954 passed, 17 skipped - `cargo fmt --all --check`, `cargo clippy --workspace --all-targets --locked -- -D warnings`, `cargo check --locked` in `tools/port-vectors` and `./ci/lint.sh` are clean - The tree differs from internal `df54b1a4` only by the scrubs of the earlier syncs, the `pipette-mgmt` pin and the doc lines above - Not run locally: the Android Gradle build, the iOS app build and the Windows-only `gpu-keepalive` job Co-authored-by: m3m0ryh0l3 <5114673+m3m0ryh0l3@users.noreply.github.com>
ykhrustalev
force-pushed
the
ykhrustalev/public-sync-2026-11-16
branch
from
November 16, 2026 21:46
c9b4f52 to
a70d0e1
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
This public repository lagged 25 commits behind the internal one, from
be6ca299(the #30 sync) todf54b1a4.Solution
failures, and a readiness timeout as
device_not_ready--cdn-proxy(PIPETTE_CDN_PROXY) to fetch models through acaching proxy by
Host, on the CLI and the iOS appout of
attempt_timeout_secondsbuild, and ships an arm64 Windows MSI of the GPU keepalive
llama-bench's own warm-up in prefill, peak-memory and CPU-only decode
cells
builds the llama.cpp integration tests as one binary
pipette-plan-typesto publicpipette-mgmtat06471959,which carries
device_not_readysteps or runners, TensorRT Edge-LLM docs, private repository links, lab
host names or release tags
and TensorRT Edge-LLM out of the new attempt-clock docs
Testing
cargo nextest run --workspace --locked --no-fail-fast: 6,954 passed,17 skipped
cargo fmt --all --check,cargo clippy --workspace --all-targets --locked -- -D warnings,cargo check --lockedintools/port-vectorsand./ci/lint.share cleandf54b1a4only by the scrubs of theearlier syncs, the
pipette-mgmtpin and the doc lines aboveWindows-only
gpu-keepalivejobCo-authored-by: m3m0ryh0l3 5114673+m3m0ryh0l3@users.noreply.github.com