Skip to content

build(llm-router): bump llm-d-router to v0.11.0 and patch vllm-router for EPD #761

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
thxCode merged 2 commits into from
Oct 14, 2026
Merged

build(llm-router): bump llm-d-router to v0.11.0 and patch vllm-router for EPD #761

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3ea7b0a
build(llm-router): bump llm-d-router to v0.11.0
thxCode Oct 14, 2026
48566f7
build(llm-router): carry vllm-router EPD routing as a patch
thxCode Oct 14, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 15 additions & 2 deletions pack/llm-router/Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,8 +29,8 @@ ARG UBUNTU_VERSION=24.04
ARG GO_VERSION=1.26-bookworm
ARG RUST_VERSION=1.95-bookworm

ARG LLMD_ROUTER_VERSION=v0.10.0
ARG VLLM_ROUTER_VERSION=v0.1.15
ARG LLMD_ROUTER_VERSION=v0.11.0
ARG VLLM_ROUTER_VERSION=v0.1.16.1
ARG SGLANG_GATEWAY_VERSION=gateway-v0.3.1

######################## llm-d endpoint picker ########################
Expand Down Expand Up @@ -150,6 +150,19 @@ EOF
# the same names but different contents, and the failure would be a confusing reject
# rather than a missing patch.
#
# This stage also carries upstream PR 266 -- commit 516f1cb, the encoder-prefill-decode
# routing -- as 0001-epd-routing.patch, a cherry-pick onto v0.1.16.1. That tag is the
# last vllm-project/router release cut before 516f1cb landed, so its tree does not
# contain the commit and the feature is carried here rather than waited out. The three
# patches above it (lazy Mooncake, the read-only observer endpoint, the Kubernetes
# discovery readiness-loss deselect) rebase onto the tree that cherry-pick produces,
# which is why they travel with it.
#
# THE IMAGE MOVES TO THE FIRST UPSTREAM TAG THAT CONTAINS THE COMMIT: when upstream
# cuts a tag whose tree carries 516f1cb, bump VLLM_ROUTER_VERSION to that tag and delete
# 0001. The other three patches stay until their own WHEN TO REMOVE condition holds, and
# each patch header names the tag it is written against under AGAINST.
#
# Applied with git apply, which fails the build loudly on a version bump whose hunks no
# longer match -- that failure is the tripwire, and git is already in the base stage so no
# package is added for this.
Expand Down
Loading
Loading