benchmax 0.2.1__tar.gz → 0.2.2.dev0__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (57) hide show
  1. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/PKG-INFO +1 -1
  2. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/pyproject.toml +1 -1
  3. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/harbor/README.md +3 -1
  4. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/.gitignore +0 -0
  5. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/README.md +0 -0
  6. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/pytest.ini +0 -0
  7. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/auth.py +0 -0
  8. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/bundle.py +0 -0
  9. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/README.md +0 -0
  10. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/__init__.py +0 -0
  11. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/base/README.md +0 -0
  12. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/base/__init__.py +0 -0
  13. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/base/dataset.py +0 -0
  14. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/base/env.py +0 -0
  15. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/base/openai_types.py +0 -0
  16. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/dataset.py +0 -0
  17. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/environment.py +0 -0
  18. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/harbor/__init__.py +0 -0
  19. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/harbor/bundled_agent.py +0 -0
  20. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/harbor/credentials.py +0 -0
  21. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/harbor/dataset.py +0 -0
  22. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/harbor/dep_check.py +0 -0
  23. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/harbor/env.py +0 -0
  24. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/harbor/types.py +0 -0
  25. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/identity.py +0 -0
  26. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/logging.py +0 -0
  27. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/shared_types.py +0 -0
  28. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/rewards/README.md +0 -0
  29. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/rewards/__init__.py +0 -0
  30. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/rewards/adaptive.py +0 -0
  31. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/rewards/deterministic.py +0 -0
  32. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/rewards/diversity.py +0 -0
  33. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/rewards/judge.py +0 -0
  34. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/rewards/prompts.py +0 -0
  35. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/rewards/rubric.py +0 -0
  36. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/rewards/scoring.py +0 -0
  37. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/conftest.py +0 -0
  38. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/bundle/test_artifact.py +0 -0
  39. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/bundle/test_source_capture.py +0 -0
  40. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/envs/test_base_dataset.py +0 -0
  41. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/envs/test_base_env_group.py +0 -0
  42. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/envs/test_contract_types.py +0 -0
  43. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/envs/test_environment_group.py +0 -0
  44. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/envs/test_example_id.py +0 -0
  45. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/fakes/model_server.py +0 -0
  46. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/harbor/test_bundled_agent.py +0 -0
  47. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/harbor/test_harbor_dataset.py +0 -0
  48. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/harbor/test_harbor_env.py +0 -0
  49. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/rewards/conftest.py +0 -0
  50. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/rewards/test_adaptive.py +0 -0
  51. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/rewards/test_deterministic.py +0 -0
  52. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/rewards/test_diversity.py +0 -0
  53. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/rewards/test_diversity_env.py +0 -0
  54. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/rewards/test_judge.py +0 -0
  55. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/rewards/test_rubric.py +0 -0
  56. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/rewards/test_rubric_rewards.py +0 -0
  57. {benchmax-0.2.1 → benchmax-0.2.2.dev0}/tests/unit/test_auth.py +0 -0
{benchmax-0.2.1 → benchmax-0.2.2.dev0}/PKG-INFO RENAMED
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: benchmax
3
- Version: 0.2.1
3
+ Version: 0.2.2.dev0
4
4
  Summary: Platform-independent runtime for grouped LLM environments
5
5
  Author: benchmax Authors
6
6
  Classifier: Operating System :: OS Independent
{benchmax-0.2.1 → benchmax-0.2.2.dev0}/pyproject.toml RENAMED
@@ -1,6 +1,6 @@
1
1
  [project]
2
2
  name = "benchmax"
3
- version = "0.2.1"
3
+ version = "0.2.2.dev0"
4
4
  description = "Platform-independent runtime for grouped LLM environments"
5
5
  readme = "README.md"
6
6
  authors = [{ name = "benchmax Authors" }]
{benchmax-0.2.1 → benchmax-0.2.2.dev0}/src/benchmax/envs/harbor/README.md RENAMED
@@ -50,6 +50,8 @@ for each rollout, `HarborEnv` gives the task to the configured agent and sandbox
50
50
 
51
51
  the rollout request supplies the model endpoint and credential used by the agent. verifier and judge configuration remains separate. because agents usually run inside remote sandboxes, `HarborEnv` requests a publicly reachable model endpoint by default.
52
52
 
53
+ configure the agent harness with compaction and summarization disabled: training assumes a linear chat history where each model call extends the previous messages. a harness that rewrites its context mid-rollout produces a transcript that cannot be trained on.
54
+
53
55
  benchmax owns rollout-group concurrency. `max_concurrent_trials` can additionally cap how many Harbor sandbox trials this environment runs at once.
54
56
 
55
57
  ## sandboxes and credentials
@@ -58,7 +60,7 @@ the Harbor environment configuration selects the sandbox provider. credentials a
58
60
 
59
61
  Modal and Daytona currently use explicit `sandbox_credentials`. these values become part of the environment bundle, so use dedicated, revocable credentials rather than personal ones.
60
62
 
61
- Harbor verifiers currently receive credentials as static environment variables in `TrialVerifierConfig`. runtime credential injection is available to the rollout agent, not to the verifier.
63
+ Harbor verifiers receive credentials as static environment variables in `TrialVerifierConfig`. the only values injected at rollout time are the model endpoint and API key the agent uses; nothing is injected into the verifier.
62
64
 
63
65
  ## bundled custom agents
64
66
 
{benchmax-0.2.1 → benchmax-0.2.2.dev0}/.gitignore RENAMED
File without changes
{benchmax-0.2.1 → benchmax-0.2.2.dev0}/README.md RENAMED
File without changes
{benchmax-0.2.1 → benchmax-0.2.2.dev0}/pytest.ini RENAMED
File without changes