feat(generation): remote vLLM/NIM generation backend (GPU-free) - #575

Draft

binaryaaron wants to merge 16 commits into

mainfrom

agonzales/remote-generation-backend

Commits on Jul 10, 2026

feat(generation): add remote vLLM endpoint backend

binaryaaron
and
claude
committed
56ff5a8
fix(generation): use vLLM structured_outputs field for remote backend

binaryaaron
and
claude
committed
58baef4
feat(generation): support structural_tag for the remote backend

binaryaaron
and
claude
committed
809352f
feat(tools): add vllm_debug standalone uv script

binaryaaron
and
claude
committed
f2a2447
docs(working-memory): chronicle remote backend work + serving roadmap

binaryaaron
and
claude
committed
2aca78c
docs(working-memory): chronicle the modular state-machine architecture arc

binaryaaron
and
claude
committed
5ce236b
feat(tools): default vllm_debug serve to CUDA graphs (eager opt-in)

binaryaaron
and
claude
committed
56a5045
chore: apply copyright headers to working-memory notes

binaryaaron
and
claude
committed
8b458ff
feat(generation): add remote dialect (vllm|openai) for strict OpenAI servers

binaryaaron
and
claude
committed
70ed3f0
docs(working-memory): add NIM/runtime validation, dialect, perf to chronicle

binaryaaron
and
claude
committed
361c1a1
docs: self-review fixes — module map + measured quality/perf in chronicle

binaryaaron
and
claude
committed
39d7cfb
feat(generation): harden remote backend with retries and defensive parsing
binaryaaron
committed
aacb958
fix(tools): harden vllm_debug structured-output input parsing
binaryaaron
committed
2a26b3f
test(smoke): add remote generation backend smoke tests
binaryaaron
committed
78559fd
chore(working-memory): relocate chronicle notes to the safe-synthesizer KB
binaryaaron
committed
b6f88d9
refactor(llm): define typed local model host
binaryaaron
committed
c226c79

Navigation Menu

Uh oh!

feat(generation): remote vLLM/NIM generation backend (GPU-free) - #575

feat(generation): remote vLLM/NIM generation backend (GPU-free)#575
binaryaaron wants to merge 16 commits into
mainfrom
agonzales/remote-generation-backend

Commits on Jul 10, 2026

feat(generation): add remote vLLM endpoint backend

fix(generation): use vLLM structured_outputs field for remote backend

feat(generation): support structural_tag for the remote backend

feat(tools): add vllm_debug standalone uv script

docs(working-memory): chronicle remote backend work + serving roadmap

docs(working-memory): chronicle the modular state-machine architecture arc

feat(tools): default vllm_debug serve to CUDA graphs (eager opt-in)

chore: apply copyright headers to working-memory notes

feat(generation): add remote dialect (vllm|openai) for strict OpenAI servers

docs(working-memory): add NIM/runtime validation, dialect, perf to chronicle

docs: self-review fixes — module map + measured quality/perf in chronicle

feat(generation): harden remote backend with retries and defensive parsing

fix(tools): harden vllm_debug structured-output input parsing

test(smoke): add remote generation backend smoke tests

chore(working-memory): relocate chronicle notes to the safe-synthesizer KB

refactor(llm): define typed local model host