{ "agent_trace": { "codex_logs_available_local_only": true, "raw_logs_published": false, "summary_path": "agent-summary.json" }, "codex_version": "codex-cli 0.131.0", "collected_at": "2026-05-19T20:23:04.608679+00:00", "docker_cpus": 16, "docker_memory": "30g", "eval": { "error_code": null, "failed_tests": [ "eval.tests.test_oranda_comprehensive.test_no_arguments_shows_help", "eval.tests.test_oranda_comprehensive.test_generate_ci_help", "eval.tests.test_oranda_comprehensive.test_serve_without_build", "eval.tests.test_oranda_comprehensive.test_generate_ci_requires_tty", "eval.tests.test_basic_invocation.test_no_args_shows_help", "eval.tests.test_cargo_detection.test_build_cargo_and_config_preference", "eval.tests.test_cargo_detection.test_build_with_cargo_toml", "eval.tests.test_advanced_config.test_build_with_changelog_enabled", "eval.tests.test_error_cases.test_build_with_nonexistent_readme_path", "eval.tests.test_error_cases.test_invalid_theme_name", "eval.tests.test_generate_command.test_generate_ci_custom_output_path_requires_tty", "eval.tests.test_generate_command.test_generate_ci_requires_tty", "eval.tests.test_generate_command.test_generate_ci_help", "eval.tests.test_generate_command.test_generate_ci_with_site_path_requires_tty", "eval.tests.test_hidden_commands.test_config_schema_command", "eval.tests.test_hidden_commands.test_config_schema_help", "eval.tests.test_hidden_commands.test_config_schema_with_output_file", "eval.tests.test_dev_command.test_dev_no_first_build_flag", "eval.tests.test_dev_command.test_dev_with_config_starts", "eval.tests.test_error_cases.test_serve_requires_public_dir", "eval.tests.test_hidden_commands.test_generate_css_help", "eval.tests.test_serve_command.test_serve_verbose_flag", "eval.tests.test_generate_command.test_generate_ci_explicit_github_requires_tty", "eval.tests.test_serve_command.test_serve_without_public_dir", "eval.tests.test_error_cases.test_build_empty_directory", "eval.tests.test_subcommand_help.test_generate_ci_help_content", "tests.test_build.test_build_with_package_json_extracts_metadata", "tests.test_build.test_build_workspace_empty_members_list", "tests.test_build.test_build_creates_artifacts_js", "tests.test_config.test_marketing_social_twitter", "tests.test_config.test_styles_theme_light", "tests.test_config.test_build_no_config_uses_defaults", "tests.test_build.test_build_default_theme_is_dark", "tests.test_build.test_build_json_only_workspace", "tests.test_build.test_build_github_repo_link_rendered", "tests.test_build.test_build_og_meta_tags_present", "tests.test_config.test_styles_additional_css", "tests.test_config.test_styles_logo_custom", "tests.test_build.test_build_empty_directory_creates_minimal_site", "tests.test_build.test_build_creates_favicon", "tests.test_artifact_inference.test_tarball_with_target_triple_inference_and_labeling", "tests.test_artifact_inference.test_sha256_files_not_treated_as_installers", "tests.test_build.test_build_workspace_with_custom_dist_dir", "tests.test_config.test_invalid_project_type_error", "tests.test_build.test_build_with_oranda_json_overrides_config", "tests.test_config.test_build_empty_config", "tests.test_config.test_styles_favicon_custom", "tests.test_build.test_build_workspace_multiple_members_all_written", "tests.test_build.test_build_multiple_files_in_static_dir", "tests.test_config.test_styles_theme_axodark", "tests.test_build.test_build_permissions_policy_header", "tests.test_build.test_build_workspace_generates_index_page", "tests.test_build.test_build_static_dir_copied_to_output", "tests.test_config.test_build_path_prefix_option", "tests.test_build.test_build_with_cargo_toml_extracts_metadata", "tests.test_config.test_marketing_analytics_plausible", "tests.test_config.test_project_name_override", "tests.test_config.test_styles_multiple_additional_css", "tests.test_config.test_invalid_json_syntax_error", "tests.test_build.test_build_workspace_member_data_collected", "tests.test_config.test_project_multiple_fields", "tests.test_build.test_build_theme_light_applied", "tests.test_build.test_build_single_project_with_custom_dist_dir", "tests.test_config.test_components_changelog_with_config", "tests.test_build.test_build_changelog_link_in_navigation", "tests.test_build.test_build_workspace_with_generate_index_false", "tests.test_build.test_build_empty_cargo_toml", "tests.test_config.test_marketing_analytics_fathom", "tests.test_build.test_build_invalid_json_shows_helpful_error", "tests.test_config.test_project_license_in_footer", "tests.test_artifact_inference.test_msi_bundle_detection_and_labeling", "tests.test_artifact_inference.test_tarball_label_not_zip_label", "tests.test_artifact_inference.test_installer_script_detection_shell_and_powershell", "tests.test_config.test_project_description_in_metadata", "tests.test_config.test_marketing_analytics_plausible_default_script_url", "tests.test_build.test_build_html_structure_valid", "tests.test_config.test_marketing_analytics_google", "tests.test_build.test_build_workspace_member_with_missing_path", "tests.test_build.test_build_workspace_success_message_uses_workspace_dist_dir", "tests.test_config.test_marketing_analytics_umami" ], "failed_tests_omitted": 237, "public_eval_path": "eval.json", "summary_path": "eval-summary.json", "test_branch_errors": {}, "test_records": 978, "warnings": [] }, "host_machine": "x86_64", "host_system": "Linux", "inference_mode": "paper-prompt-nointernet", "instance_id": "axodotdev__oranda.27d60c7", "metrics": { "almost_resolved": "False", "cached_input_tokens": "3290496", "calls": "81", "codex_version": "codex-cli 0.131.0", "created_at": "2026-05-18T23:50:21.002109+00:00", "docker_cpus": "16", "docker_memory": "30g", "error_code": "", "estimated_cost_usd": "", "evaluator_problem": "False", "host_machine": "x86_64", "host_system": "Linux", "inference_mode": "paper-prompt-nointernet", "input_tokens": "3453586", "instance_id": "axodotdev__oranda.27d60c7", "long_context_calls": "0", "model": "gpt-5.5", "n_resolved_tests": "469", "n_system_errors": "0", "n_tests": "767", "n_warnings": "0", "output_tokens": "40471", "paper_compliant": "False", "pricing_source": "", "reasoning_effort": "xhigh", "reasoning_output_tokens": "10671", "resolved": "False", "run_name": "cpx62-paperprompt-xhigh-shard-1-of-5-20260518T232006Z-axodotdev-oranda", "run_version": "20260518T232006Z", "score": "0.6114732724902217", "session_ended_at": "2026-05-19T00:00:26.454Z", "session_started_at": "2026-05-18T23:50:29.042Z", "test_branch_errors": "{}", "total_tokens": "3494057", "wall_clock_seconds": "597" }, "model": "gpt-5.5", "package": { "contents": [ ".", "./.git", "./.git/COMMIT_EDITMSG", "./.git/HEAD", "./.git/branches", "./.git/config", "./.git/description", "./.git/hooks", "./.git/hooks/applypatch-msg.sample", "./.git/hooks/commit-msg.sample", "./.git/hooks/fsmonitor-watchman.sample", "./.git/hooks/post-update.sample", "./.git/hooks/pre-applypatch.sample", "./.git/hooks/pre-commit.sample", "./.git/hooks/pre-merge-commit.sample", "./.git/hooks/pre-push.sample", "./.git/hooks/pre-rebase.sample", "./.git/hooks/pre-receive.sample", "./.git/hooks/prepare-commit-msg.sample", "./.git/hooks/push-to-checkout.sample", "./.git/hooks/update.sample", "./.git/index", "./.git/info", "./.git/info/exclude", "./.git/logs", "./.git/logs/HEAD", "./.git/logs/refs", "./.git/logs/refs/heads", "./.git/logs/refs/heads/master", "./.git/objects", "./.git/objects/25", "./.git/objects/25/a56e8f4073cadf7c62891e71143179672d7efa", "./.git/objects/27", "./.git/objects/27/a336e15d9c1deebf463dd99aa470331c728e1f", "./.git/objects/41", "./.git/objects/41/e30c89e5fa11a6401ff19fc7b0f0d045e03dd0", "./.git/objects/8e", "./.git/objects/8e/782fc2011c3eb128aa01e36392d3d9e36b8d45", "./.git/objects/c8", "./.git/objects/c8/d6ac9565be096a3c2217e996e05b434dd538a1", "./.git/objects/info", "./.git/objects/pack", "./.git/refs", "./.git/refs/heads", "./.git/refs/heads/master", "./.git/refs/tags", "./LICENSE-APACHE", "./LICENSE-MIT", "./README.md", "./compile.sh", "./executable", "./oranda_reimpl.py" ], "contents_omitted": 0, "submission_available_local_only": true }, "paper_compliant": false, "paper_mode": true, "reasoning_effort": "xhigh", "run_name": "cpx62-paperprompt-xhigh-shard-1-of-5-20260518T232006Z-axodotdev-oranda", "run_version": "20260518T232006Z", "usage_audit": { "available_local_artifact": true, "public_path": "usage-audit.json" } }