feat(vllm): parity with llama.cpp backend - #9328

Merged

mudler merged 16 commits into

feat/vllm-parity

Apr 13, 2026

Commits on Apr 12, 2026

fix(schema): serialize ToolCallID and Reasoning in Messages.ToProto
mudler
committed
40b1c6f
refactor(config): introduce backend hook system and migrate llama-cpp defaults
mudler
committed
a30719f
feat(config): add vLLM parser defaults hook and importer auto-detection
mudler
committed
6cf8263
feat(vllm): wire native tool/reasoning parsers + chat deltas + logprobs
mudler
committed
6786f05
feat(vllm): CPU support + shared utils + vllm-omni feature parity
mudler
committed
b215843
fix(vllm): CPU build compatibility with vllm 0.14.1
mudler
committed
c2f73a9
fix(vllm): tool parser constructor compat + e2e tool calling test
mudler
committed
c99188f
ci(backend): build cpu-vllm container image
mudler
committed
034a60b
test(e2e-backends): add tools capability + HF model name support
mudler
committed
e7f4061
ci(test-extra): run vllm e2e tests on CPU
mudler
committed
c7f444d
fix(vllm): build from source on CI to avoid SIGILL on prebuilt wheel
mudler
committed
329df11
ci(vllm): use bigger-runner instead of source build
mudler
committed
ea2bbab
ci(vllm): install make + build deps on bigger-runner
mudler
committed
c4dc495
ci(vllm): install libnuma1 + libgomp1 on bigger-runner
mudler
committed
017bdee
feat(vllm): bundle libnuma/libgomp via package.sh
mudler
committed
d74cd56

Commits on Apr 13, 2026

ci(vllm): disable tests-vllm-grpc job (heterogeneous runners)
mudler
committed
cd56a05