feat: uncapped computer agents, stream-safe responses and provider volumes by abhishekgahlot2 · Pull Request #98 · DeepGym/deepgym

fix: stop on incomplete openai responses and raise their output cap

abhishekgahlot2

committed

d8535cf

feat: model-max output and max effort for computer agents

abhishekgahlot2

committed

8c0b10a

fix: never act on an unfinished anthropic stream

abhishekgahlot2

committed

1eab97c

fix: require message_stop before acting on a stream

abhishekgahlot2

committed

b7744a2

feat: unlimited turns, wall clock and steps when set to none

abhishekgahlot2

committed

8e44e48

fix: scripted qa controls end when an env has no step limit

abhishekgahlot2

committed

5eea61a

feat: mount provider volumes into daytona sessions

abhishekgahlot2

committed

7febe68

test: volume params on the session sandbox builder

abhishekgahlot2

committed

67e2877

docs: volumes are shared across episodes; mount read-only and pin

abhishekgahlot2

committed

541a0ce

Navigation Menu

Repository navigation

feat: uncapped computer agents, stream-safe responses and provider volumes - #98

feat: uncapped computer agents, stream-safe responses and provider volumes#98
abhishekgahlot2 merged 9 commits into
mainfrom
fix/openai-incomplete-response-pr

Commits on Oct 18, 2026

fix: stop on incomplete openai responses and raise their output cap

feat: model-max output and max effort for computer agents

fix: never act on an unfinished anthropic stream

fix: require message_stop before acting on a stream

feat: unlimited turns, wall clock and steps when set to none

fix: scripted qa controls end when an env has no step limit

feat: mount provider volumes into daytona sessions

test: volume params on the session sandbox builder

docs: volumes are shared across episodes; mount read-only and pin