feat(catalogue): add OSWorld2 with OSWorld-V2's own agents, its services on the machine and a null control by sirotaoleg · Pull Request #134 · ThinkFlowLab/nullius

feat(standard): pin the serving providers that may answer a model

pierreleca

committed

c570c02

docs(standard): describe --served-by and the record of who served a call

pierreleca

committed

ebe1a19

docs(standard): state the serving providers' differences in a sentence of their own

pierreleca

committed

76305ac

fix(standard): read who served a call apart from the record, and change no outcome without a pin

pierreleca

committed

f56bb3c

docs(standard): say how the record reads who served a call, and what held cannot tell

pierreleca

committed

a048866

test(standard): run the relay's own command in a bounded process when it must refuse a pin

pierreleca

committed

325e014

docs(standard): say what held reads in a sentence without a contrast

pierreleca

committed

fbcbcdf

feat(builder): choose a model's serving providers on the Model step

pierreleca

committed

8771cf0

docs(builder): describe Served by on the Model step and its list of a model's providers

pierreleca

committed

c0ea2bb

fix(builder): set Served by below its model field's row, clear of its line

pierreleca

committed

c47d9ff

fix(standard): hold a pin to the provider that answered and to the model it was set for

pierreleca

committed

5c57a38

docs(standard): say what a slug's variant is, which error is kept, and where held reads unknown

pierreleca

committed

63d11c8

fix(standard): read who served a reply from the endpoint its routing facts mark as selected

pierreleca

committed

980f22d

docs(standard): say a Responses request asks for the routing facts, a pin no provider serves gets 404, and why held may read unknown

pierreleca

committed

f092fa3

test(standard): check that routing facts marking no single endpoint take no name a chunk gave before

pierreleca

committed

fe81ec1

refactor(builder): drop a comment written twice above the model field's change

pierreleca

committed

7b3f8f3

docs(standard): say in sentences of their own why a Responses request asks for the routing facts and where held reads unknown

pierreleca

committed

d6f6514

Merge branch 'main' into feat/served-by

pierreleca

committed

ace9d29

test(standard): keep a unit list and a serving pin together in a run's record

pierreleca

committed

7cde14d

docs(standard): describe a job's unit list and serving pin together in the field guide

pierreleca

committed

a7317da

Merge branch 'main' into feat/served-by

pierreleca

committed

e0dd012

test(standard): run a serving pin under a machine's constraints

pierreleca

committed

a80ca4e

docs(standard): name a job's constraints beside its unit list and serving pin in the field guide

pierreleca

committed

468406d

Navigation Menu

Repository navigation

feat(catalogue): add OSWorld2 with OSWorld-V2's own agents, its services on the machine and a null control - #134

feat(catalogue): add OSWorld2 with OSWorld-V2's own agents, its services on the machine and a null control#134
sirotaoleg wants to merge 31 commits into
ThinkFlowLab:mainfrom
sirotaoleg:feat/osworld2

Commits on Nov 13, 2026

feat(standard): pin the serving providers that may answer a model

docs(standard): describe --served-by and the record of who served a call

docs(standard): state the serving providers' differences in a sentence of their own

fix(standard): read who served a call apart from the record, and change no outcome without a pin

docs(standard): say how the record reads who served a call, and what held cannot tell

test(standard): run the relay's own command in a bounded process when it must refuse a pin

docs(standard): say what held reads in a sentence without a contrast

feat(builder): choose a model's serving providers on the Model step

docs(builder): describe Served by on the Model step and its list of a model's providers

fix(builder): set Served by below its model field's row, clear of its line

fix(standard): hold a pin to the provider that answered and to the model it was set for

docs(standard): say what a slug's variant is, which error is kept, and where held reads unknown

fix(standard): read who served a reply from the endpoint its routing facts mark as selected

docs(standard): say a Responses request asks for the routing facts, a pin no provider serves gets 404, and why held may read unknown

test(standard): check that routing facts marking no single endpoint take no name a chunk gave before

refactor(builder): drop a comment written twice above the model field's change

docs(standard): say in sentences of their own why a Responses request asks for the routing facts and where held reads unknown

Merge branch 'main' into feat/served-by

test(standard): keep a unit list and a serving pin together in a run's record

docs(standard): describe a job's unit list and serving pin together in the field guide

Merge branch 'main' into feat/served-by

test(standard): run a serving pin under a machine's constraints

docs(standard): name a job's constraints beside its unit list and serving pin in the field guide

Commits on Nov 15, 2026

feat(standard): give the standard what computer-use benchmarks need, on top of --served-by: gated tasks, a definition's pins as its serving pins, a pinned route's refusal as the unit's, limits of t…

feat(standard): let a benchmark run no rate-limited unit again, and keep the score of a unit a guard stopped despite a last 429

fix(standard): keep a guard-stopped unit's score only for the agent's own last 429, not a side call's or a context window's stop

feat(catalogue): bundle OSWorld2 on the standard: its driver with the upstream's own runners, mock websites, GitLab and proxy, proven on test_v2, and how to prepare a machine for it

docs(catalogue): say that a machine's constraints in Settings do not reach OSWorld2's VMs and services

feat(catalogue): run no rate-limited OSWorld2 unit again by itself

feat(catalogue): prove OSWorld2's null control on test_v2 again, on this branch

test(catalogue): run OSWorld2's tests on a machine without its scoring programs, and fake the sweep's clock for the sweep alone