generic-ml-wrapper 0.1.0__tar.gz → 0.3.0__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.
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/.gitignore +3 -0
- generic_ml_wrapper-0.3.0/CHANGELOG.md +110 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/CONTRIBUTING.md +3 -3
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/GOVERNANCE.md +23 -20
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/PKG-INFO +49 -6
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/README.md +48 -5
- generic_ml_wrapper-0.3.0/RELEASE.md +38 -0
- generic_ml_wrapper-0.3.0/ROADMAP.md +173 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/SECURITY.md +1 -1
- generic_ml_wrapper-0.3.0/VERSION +1 -0
- generic_ml_wrapper-0.3.0/docs/CLI.md +257 -0
- generic_ml_wrapper-0.3.0/docs/CLIENTS.md +135 -0
- generic_ml_wrapper-0.3.0/docs/CONFIGURATION.md +288 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/docs/DESIGN.md +96 -21
- generic_ml_wrapper-0.3.0/docs/TROUBLESHOOTING.md +113 -0
- generic_ml_wrapper-0.3.0/docs/USER_GUIDE.md +153 -0
- generic_ml_wrapper-0.3.0/docs/WORKFLOWS.md +173 -0
- generic_ml_wrapper-0.3.0/hatch_build.py +53 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/pyproject.toml +8 -1
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/__init__.py +1 -1
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/adapter/inbound/cli/app.py +770 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/adapter/outbound/bootstrap/filesystem_layout_seeder.py +197 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/adapter/outbound/bootstrap/path_client_detector.py +24 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/adapter/outbound/bootstrap/tty_client_chooser.py +49 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/adapter/outbound/bootstrap/tty_persona_chooser.py +52 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/caller/claude_cli_caller.py +1 -1
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/caller/cursor_cli_caller.py +1 -1
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/caller/default_provider.py +10 -2
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/caller/status_line_config.py +21 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/adapter/outbound/compress/__init__.py +3 -0
- generic_ml_wrapper-0.1.0/src/generic_ml_wrapper/adapter/outbound/interceptor/compressor.py → generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/adapter/outbound/compress/cache_backed_compressor.py +21 -18
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/adapter/outbound/hook/__init__.py +3 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/adapter/outbound/hook/session_logger.py +61 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/adapter/outbound/persona/__init__.py +3 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/adapter/outbound/persona/filesystem_persona_source.py +68 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/adapter/outbound/plugin/__init__.py +3 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/adapter/outbound/plugin/filesystem_plugin_source.py +102 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/adapter/outbound/workflow/filesystem_workflow_source.py +320 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/domain/model/client_catalog.py +77 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/domain/model/context_source.py +91 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/domain/model/learned.py +56 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/domain/model/persona.py +26 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/domain/model/plugin.py +20 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/domain/model/rules.py +94 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/domain/service/greeting.py +72 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/domain/service/hook.py +75 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/domain/service/hook_runner.py +49 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/domain/service/persona_parser.py +57 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/port/inbound/check_client_ready.py +46 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/port/inbound/first_run_init.py +37 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/port/inbound/list_personas.py +23 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/port/inbound/list_plugins.py +23 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/port/inbound/render_greeting.py +20 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/port/outbound/client_chooser.py +23 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/port/outbound/client_detector.py +20 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/port/outbound/context_compressor.py +29 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/port/outbound/hook.py +25 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/outbound/layout_seeder.py +8 -1
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/port/outbound/persona_chooser.py +27 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/port/outbound/persona_source.py +42 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/port/outbound/plugin_source.py +64 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/outbound/workflow_source.py +15 -4
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/usecase/check_client_ready.py +56 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/usecase/first_run_init.py +71 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/usecase/launch.py +70 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/usecase/list_personas.py +33 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/usecase/list_plugins.py +33 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/usecase/new_workflow.py +8 -6
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/application/usecase/render_greeting.py +75 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/usecase/start_job.py +24 -9
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/wiring/composition.py +141 -1
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/common/config.py +387 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/common/paths.py +8 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/resources/personas/_floor.md +13 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/resources/personas/butler.md +18 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/resources/personas/companion.md +18 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/resources/personas/mentor.md +18 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/resources/personas/plain.md +12 -0
- generic_ml_wrapper-0.3.0/src/generic_ml_wrapper/resources/personas/terse.md +17 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/resources/workflows/create-workflow/workflow.md +25 -4
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_authoring_separation.py +2 -0
- generic_ml_wrapper-0.3.0/tests/test_bootstrap.py +101 -0
- generic_ml_wrapper-0.3.0/tests/test_cache_backed_compressor.py +95 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_callers.py +23 -0
- generic_ml_wrapper-0.3.0/tests/test_check_client_ready.py +57 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_cli.py +439 -1
- generic_ml_wrapper-0.3.0/tests/test_config.py +180 -0
- generic_ml_wrapper-0.3.0/tests/test_context_source.py +54 -0
- generic_ml_wrapper-0.3.0/tests/test_docs_consistency.py +70 -0
- generic_ml_wrapper-0.3.0/tests/test_exit_handling.py +118 -0
- generic_ml_wrapper-0.3.0/tests/test_filesystem_persona_source.py +50 -0
- generic_ml_wrapper-0.3.0/tests/test_filesystem_plugin_source.py +78 -0
- generic_ml_wrapper-0.3.0/tests/test_filesystem_workflow_source.py +343 -0
- generic_ml_wrapper-0.3.0/tests/test_first_run_init.py +123 -0
- generic_ml_wrapper-0.3.0/tests/test_greeting.py +59 -0
- generic_ml_wrapper-0.3.0/tests/test_hook_runner.py +76 -0
- generic_ml_wrapper-0.3.0/tests/test_learned.py +26 -0
- generic_ml_wrapper-0.3.0/tests/test_list_personas_usecase.py +29 -0
- generic_ml_wrapper-0.3.0/tests/test_list_plugins_usecase.py +26 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_list_workflows_usecase.py +2 -1
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_new_workflow_usecase.py +8 -4
- generic_ml_wrapper-0.3.0/tests/test_path_client_detector.py +36 -0
- generic_ml_wrapper-0.3.0/tests/test_persona_parser.py +38 -0
- generic_ml_wrapper-0.3.0/tests/test_render_greeting_usecase.py +88 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_rule_cleaner.py +19 -0
- generic_ml_wrapper-0.3.0/tests/test_session_logger.py +64 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_start_job_usecase.py +63 -4
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_status_line_config.py +16 -0
- generic_ml_wrapper-0.3.0/tests/test_tty_client_chooser.py +55 -0
- generic_ml_wrapper-0.3.0/tests/test_tty_persona_chooser.py +58 -0
- generic_ml_wrapper-0.3.0/tests/test_version.py +24 -0
- generic_ml_wrapper-0.1.0/CHANGELOG.md +0 -45
- generic_ml_wrapper-0.1.0/VERSION +0 -1
- generic_ml_wrapper-0.1.0/src/generic_ml_wrapper/adapter/inbound/cli/app.py +0 -369
- generic_ml_wrapper-0.1.0/src/generic_ml_wrapper/adapter/outbound/bootstrap/filesystem_layout_seeder.py +0 -104
- generic_ml_wrapper-0.1.0/src/generic_ml_wrapper/adapter/outbound/workflow/filesystem_workflow_source.py +0 -165
- generic_ml_wrapper-0.1.0/src/generic_ml_wrapper/common/config.py +0 -179
- generic_ml_wrapper-0.1.0/tests/test_bootstrap.py +0 -51
- generic_ml_wrapper-0.1.0/tests/test_compressor.py +0 -88
- generic_ml_wrapper-0.1.0/tests/test_config.py +0 -92
- generic_ml_wrapper-0.1.0/tests/test_filesystem_workflow_source.py +0 -135
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/.coveragerc +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/.gitattributes +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/.github/workflows/ci.yml +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/.github/workflows/pr-hygiene.yml +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/.github/workflows/release.yml +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/.github/workflows/sonar.yml +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/.importlinter +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/.pre-commit-config.yaml +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/AGENTS.md +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/CODE_OF_CONDUCT.md +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/LICENSE +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/NOTICE +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/docs/images/gmlw-favicon.svg +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/docs/images/gmlw-lockup-dark.svg +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/docs/images/gmlw-lockup.svg +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/docs/images/gmlw-statusline.gif +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/docs/images/gmlw-usage.gif +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/docs/tapes/README.md +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/docs/tapes/render.sh +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/docs/tapes/seed.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/docs/tapes/statusline.tape +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/docs/tapes/usage.tape +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/noxfile.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/pyrightconfig.json +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/sonar-project.properties +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/inbound/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/inbound/cli/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/inbound/cli/banner.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/bootstrap/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/caller/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/caller/codex_cli_caller.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/caller/context_file.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/caller/context_opening.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/caller/loader.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/caller/vibe_cli_caller.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/caller/vibe_config.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/credentials/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/credentials/filesystem_credentials_store.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/gateway/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/gateway/anthropic_sse.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/gateway/openai_chat.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/gateway/openai_responses.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/gateway/relay.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/interceptor/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/interceptor/size_logger.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/status/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/status/claude_status_parser.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/status/cursor_status_parser.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/store/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/store/filesystem_transcript_store.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/store/ledger.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/store/sqlite_per_turn_store.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/store/sqlite_session_store.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/store/sqlite_usage_store.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/workflow/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/workspace/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/adapter/outbound/workspace/local_workspace_inspector.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/domain/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/domain/model/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/domain/model/client_status.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/domain/model/identifiers.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/domain/model/run.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/domain/model/session.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/domain/model/turn_usage.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/domain/model/workspace.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/domain/service/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/domain/service/interceptor.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/domain/service/interceptor_chain.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/domain/service/rule_cleaner.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/domain/service/session_naming.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/domain/service/statusline_renderer.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/inbound/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/inbound/bootstrap.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/inbound/export_usage.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/inbound/list_jobs.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/inbound/list_sessions.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/inbound/list_workflows.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/inbound/new_workflow.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/inbound/render_statusline.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/inbound/set_credential.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/inbound/start_job.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/outbound/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/outbound/cli_caller.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/outbound/client_status.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/outbound/credentials_store.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/outbound/interceptor.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/outbound/per_turn_metering.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/outbound/session_store.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/outbound/transcript.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/outbound/usage_store.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/port/outbound/workspace.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/usecase/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/usecase/bootstrap.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/usecase/export_usage.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/usecase/list_jobs.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/usecase/list_sessions.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/usecase/list_workflows.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/usecase/render_statusline.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/usecase/set_credential.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/application/wiring/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/common/__init__.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/common/log.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/common/spec_loader.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/py.typed +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/src/generic_ml_wrapper/resources/workflows/_common/base.md +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/_conformance.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/conftest.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_anthropic_sse.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_banner.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_caller_loader.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_context_file.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_credentials_store.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_cursor_status_parser.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_export_usage_usecase.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_filesystem_transcript_store.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_identifiers.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_interceptor_chain.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_list_jobs_usecase.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_list_sessions_usecase.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_local_workspace_inspector.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_log.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_model_invariants.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_openai_chat.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_openai_responses.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_relay.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_session_naming.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_size_logger.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_smoke.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_sqlite_per_turn_store.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_sqlite_session_store.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_sqlite_usage_store.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_statusline.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_store_conformance.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tests/test_vibe_config.py +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tools/hooks/check-commit-msg.sh +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/tools/hooks/guard-branch.sh +0 -0
- {generic_ml_wrapper-0.1.0 → generic_ml_wrapper-0.3.0}/uv.lock +0 -0
|
@@ -0,0 +1,110 @@
|
|
|
1
|
+
# Changelog
|
|
2
|
+
|
|
3
|
+
All notable changes to this project are documented here. The format follows
|
|
4
|
+
[Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and the project adheres
|
|
5
|
+
to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
|
|
6
|
+
|
|
7
|
+
## [Unreleased]
|
|
8
|
+
|
|
9
|
+
## [0.3.0] - 2026-07-18
|
|
10
|
+
|
|
11
|
+
The lifecycle release — hooks that act at the seams around a run, and rule capture that
|
|
12
|
+
works everywhere.
|
|
13
|
+
|
|
14
|
+
### Added
|
|
15
|
+
- **Lifecycle action hooks.** gmlw already hooks *content* (the interceptor chain); it now
|
|
16
|
+
hooks *actions* at two seams bracketing the client run. A `[[hooks]]` entry binds a
|
|
17
|
+
`HookPort` spec (a `module:Class` / `/path.py:Class`, or a plugin id) to a `phase` —
|
|
18
|
+
`pre-launch` (after the context is compiled and the caller resolved, before the client
|
|
19
|
+
starts) or `post-session` (after the client exits, with its exit code) — with an optional
|
|
20
|
+
`client` scope, under the same trusted-code boundary as `[[interceptors]]` and `[callers]`.
|
|
21
|
+
Hooks are best-effort: a failing hook never breaks a launch or its teardown. Both launch
|
|
22
|
+
paths (`start`, `workflow new`) route through one `run_with_hooks` sequence. Ships the
|
|
23
|
+
built-in `SessionLogger` as a reference hook.
|
|
24
|
+
- **Always-on rule lifecycle.** Rule capture is no longer workflow-only: the `rules`
|
|
25
|
+
context source (now active by default in a plain start, config-overridable) leads with
|
|
26
|
+
a capture directive so a demanded correction becomes a draft rule in **any** session.
|
|
27
|
+
The directive carries the full loop — offer to record a durable, reusable reflex;
|
|
28
|
+
**dedup** against the existing rules and update/supersede a match instead of stacking a
|
|
29
|
+
near-duplicate; and, when a rule is **mechanically enforceable**, offer to realise it as
|
|
30
|
+
a script or check rather than a standing reminder.
|
|
31
|
+
|
|
32
|
+
### Fixed
|
|
33
|
+
- Docs: removed a stray tool artifact from `docs/CLIENTS.md`; corrected the workflow
|
|
34
|
+
compression default in `docs/CONFIGURATION.md` (every source defaults to
|
|
35
|
+
`compression = false`); documented `[companion] name`. Added a doc-consistency guard
|
|
36
|
+
against leaked tool-artifact tags.
|
|
37
|
+
|
|
38
|
+
## [0.2.0] - 2026-07-16
|
|
39
|
+
|
|
40
|
+
The companion release — everything a session inherits, and the ergonomics around it.
|
|
41
|
+
|
|
42
|
+
### Added
|
|
43
|
+
- **Mode-aware context packaging.** A `[startup.<mode>]` matrix picks which sources
|
|
44
|
+
compose (persona · profile · learned · company · rules · workflow base/steps) for
|
|
45
|
+
each mode (default / workflow / authoring), with typed per-source compression
|
|
46
|
+
(`[compress.prompts]`).
|
|
47
|
+
- **Personas.** Selectable tone with a universal floor (`gmlw persona list`), a free
|
|
48
|
+
local host greeting at launch, and a first-run persona choice — configured under
|
|
49
|
+
`[companion]`.
|
|
50
|
+
- **Learned notebook.** A portable, user-owned notebook (`profile/me/learned.md`) the
|
|
51
|
+
client mirrors into, read into every client so what one learns they all inherit;
|
|
52
|
+
negatives are first-class.
|
|
53
|
+
- **Rule format.** A domain-neutral `Rule / When / Signals / Strength / Origin`
|
|
54
|
+
(+ optional `Precedence`), captured as a draft during a workflow.
|
|
55
|
+
- **Formal plugins folder.** Reference a caller by id (`~/.gmlw/plugins/<id>/` with a
|
|
56
|
+
`plugin.toml`); `gmlw plugins list`.
|
|
57
|
+
- **Cursor allowance block.** The status line renders cursor's plan pools from an
|
|
58
|
+
optional local cache when the client does not pipe them.
|
|
59
|
+
- **First-run init.** Detect installed clients and seed a filled config with a default.
|
|
60
|
+
- **JSON output** for the listing/reporting commands (`--json`).
|
|
61
|
+
- **Ergonomics.** Client and working-directory preflight, implicit `gmlw <job>`
|
|
62
|
+
(shorthand for `start`), a friendly no-job message, and auto-help for an incomplete
|
|
63
|
+
sub-command.
|
|
64
|
+
|
|
65
|
+
### Changed
|
|
66
|
+
- Documentation overhaul: task-oriented guides (`docs/USER_GUIDE.md`, `docs/CLI.md`,
|
|
67
|
+
`docs/CONFIGURATION.md`, `docs/CLIENTS.md`, `docs/WORKFLOWS.md`,
|
|
68
|
+
`docs/TROUBLESHOOTING.md`), a client capability matrix in the README, and `DESIGN.md`
|
|
69
|
+
synchronised with the code; `GOVERNANCE.md` corrected. Supported Python stated as
|
|
70
|
+
3.11–3.14 to match CI.
|
|
71
|
+
|
|
72
|
+
## [0.1.0] - 2026-07-14
|
|
73
|
+
|
|
74
|
+
First public release — a metering wrapper around ML coding CLIs.
|
|
75
|
+
|
|
76
|
+
### Added
|
|
77
|
+
- **Jobs & sessions.** Enter at a **job** you tag; the wrapper mints a named,
|
|
78
|
+
resumable **session** on the client and persists it. `gmlw start`, `jobs`,
|
|
79
|
+
`sessions`, `export`.
|
|
80
|
+
- **Four clients** driven the same way: **claude**, **cursor**, **codex**, **vibe**.
|
|
81
|
+
Which one is config-driven (`[client]`) or `--client`.
|
|
82
|
+
- **Metering relay.** A local, capability-URL-authenticated relay records **per-turn
|
|
83
|
+
tokens and cost** for the metered clients (claude/codex/vibe) into a SQLite ledger;
|
|
84
|
+
`gmlw export` reports per-turn rows, per-model totals, and per-session cost.
|
|
85
|
+
- **Status line** for the clients that host one (claude, cursor): git · folder ·
|
|
86
|
+
model · context% · a client-specific allowance block, plus a per-session and
|
|
87
|
+
per-job usage footer.
|
|
88
|
+
- **Workflows.** Author a small operating context once (`gmlw workflow new`, a warm
|
|
89
|
+
create-workflow interview) and launch a job with it; context is compiled from a
|
|
90
|
+
shared base + your profile + rules + the workflow steps, through an interceptor
|
|
91
|
+
chain (with opt-in context compression via `generic-ml-cache`).
|
|
92
|
+
- **Durable provenance.** The exact compiled context is written per session
|
|
93
|
+
(`contexts/<job>/<session>.context.md`); an **opt-in transcript** keeps each metered
|
|
94
|
+
call's request/response/usage.
|
|
95
|
+
- **Credentials.** Per-workflow secrets stored `0600` and injected into the client's
|
|
96
|
+
environment at launch (`gmlw creds set`).
|
|
97
|
+
- **Storage.** A single SQLite ledger (`~/.gmlw/ledger.db`, WAL) for jobs, sessions,
|
|
98
|
+
per-turn usage, and session costs.
|
|
99
|
+
- **Safety.** Validated `JobId` + filesystem containment under an owner-only `~/.gmlw`;
|
|
100
|
+
never overwriting an unparseable client-settings or credentials file.
|
|
101
|
+
|
|
102
|
+
### Engineering
|
|
103
|
+
- Hexagonal (ports & adapters), enforced by `import-linter`; strict `ruff` + `pyright`
|
|
104
|
+
over `src` and `tests`; `nox` gates mirrored by CI across Python 3.11–3.14; a
|
|
105
|
+
server-side no-AI-attribution check and branch protection.
|
|
106
|
+
|
|
107
|
+
[Unreleased]: https://github.com/danielslobozian/generic-ml-wrapper/compare/v0.3.0...HEAD
|
|
108
|
+
[0.3.0]: https://github.com/danielslobozian/generic-ml-wrapper/compare/v0.2.0...v0.3.0
|
|
109
|
+
[0.2.0]: https://github.com/danielslobozian/generic-ml-wrapper/compare/v0.1.0...v0.2.0
|
|
110
|
+
[0.1.0]: https://github.com/danielslobozian/generic-ml-wrapper/releases/tag/v0.1.0
|
|
@@ -18,7 +18,7 @@ By participating you agree to abide by the [Code of Conduct](CODE_OF_CONDUCT.md)
|
|
|
18
18
|
|
|
19
19
|
## Development setup
|
|
20
20
|
|
|
21
|
-
You need Python 3.11–3.
|
|
21
|
+
You need Python 3.11–3.14 and [`uv`](https://docs.astral.sh/uv/). The gates run through
|
|
22
22
|
[`nox`](https://nox.thea.codes/) with the `uv` backend, from the committed `uv.lock`.
|
|
23
23
|
|
|
24
24
|
```bash
|
|
@@ -41,7 +41,7 @@ The gates are defined once in [`noxfile.py`](noxfile.py); CI is a thin caller of
|
|
|
41
41
|
so what runs locally is byte-for-byte what runs in CI.
|
|
42
42
|
|
|
43
43
|
```bash
|
|
44
|
-
nox # the default gate: lint · imports · typecheck · tests (3.11–3.
|
|
44
|
+
nox # the default gate: lint · imports · typecheck · tests (3.11–3.14)
|
|
45
45
|
nox -s green # the whole gate in one env: lint · format · imports · pyright · coverage
|
|
46
46
|
nox -s tests # just the test matrix
|
|
47
47
|
nox -s coverage # tests with the 80% coverage floor
|
|
@@ -56,7 +56,7 @@ nox -s coverage # tests with the 80% coverage floor
|
|
|
56
56
|
The suite is **offline and token-free**: no real client is launched, no network call is
|
|
57
57
|
made, and no model is invoked — external boundaries are faked, and `~/.gmlw` is
|
|
58
58
|
redirected to a temp dir for every test. CI runs it on Linux, macOS, and Windows across
|
|
59
|
-
Python 3.11–3.
|
|
59
|
+
Python 3.11–3.14; make sure it passes locally before opening a PR.
|
|
60
60
|
|
|
61
61
|
## Coding guidelines
|
|
62
62
|
|
|
@@ -35,7 +35,7 @@ project's scope, it lands; if a maintainer raises a concern, it is discussed unt
|
|
|
35
35
|
resolved rather than forced through.
|
|
36
36
|
|
|
37
37
|
For larger or contested decisions — anything that changes the public CLI, the
|
|
38
|
-
|
|
38
|
+
ledger schema, the metering behaviour, or the project's scope — the maintainers
|
|
39
39
|
decide together, and the reasoning is recorded in the relevant issue or pull
|
|
40
40
|
request so it can be revisited later. Where maintainers disagree and cannot reach
|
|
41
41
|
consensus, the original author has the final say. This is a practical arrangement
|
|
@@ -47,25 +47,28 @@ to a more shared model as the maintainer team grows.
|
|
|
47
47
|
Some decisions are effectively already made, and proposals that cut against them
|
|
48
48
|
will usually be declined — not out of inflexibility, but because they are the
|
|
49
49
|
point of the project. These are documented in [`docs/DESIGN.md`](docs/DESIGN.md)
|
|
50
|
-
and [`
|
|
51
|
-
|
|
52
|
-
- **
|
|
53
|
-
|
|
54
|
-
|
|
55
|
-
|
|
56
|
-
|
|
57
|
-
|
|
58
|
-
|
|
50
|
+
and [`ROADMAP.md`](ROADMAP.md); the load-bearing ones are:
|
|
51
|
+
|
|
52
|
+
- **gmlw wraps a client; it never reimplements one.** The judgment stays in the
|
|
53
|
+
client. gmlw adds only the deterministic parts an *application* needs — session
|
|
54
|
+
identity, launch, context compilation, persistence, metering. The four client
|
|
55
|
+
adapters (claude / cursor / codex / vibe) are the core, and adding or improving
|
|
56
|
+
client support is *in* scope, not declined.
|
|
57
|
+
- **Model calls sink to [`generic-ml-cache`](https://github.com/danielslobozian/generic-ml-cache).**
|
|
58
|
+
The wrapper implements no record/replay cache or model-call layer of its own; the
|
|
59
|
+
optional context compressor talks to that sibling. Proposals to reinvent it here
|
|
59
60
|
will be declined.
|
|
60
|
-
- **The
|
|
61
|
-
|
|
62
|
-
- **
|
|
63
|
-
|
|
64
|
-
- **
|
|
65
|
-
|
|
66
|
-
|
|
67
|
-
|
|
68
|
-
|
|
61
|
+
- **The workflow is optional.** `gmlw start <job>` with no workflow is already the
|
|
62
|
+
whole wrapper; the workflow is an enrichment, never the identity of the product.
|
|
63
|
+
- **Metering is not a toggle for a metered client.** A metered client always runs
|
|
64
|
+
through the local relay; there is no plain, un-metered caller path.
|
|
65
|
+
- **Public-clean by construction.** No personal data, employer, job prefixes, or
|
|
66
|
+
internal hosts in the repo — real content lives only under `~/.gmlw`, and
|
|
67
|
+
`secret-audit.sh` gates every publish.
|
|
68
|
+
- **Dependencies point inward.** The hexagon (domain ← usecase ← ports; adapters
|
|
69
|
+
depend on ports) is enforced by import-linter, not just convention.
|
|
70
|
+
|
|
71
|
+
The full set is `docs/DESIGN.md` §15 ("Design invariants — do not re-litigate").
|
|
69
72
|
|
|
70
73
|
A maintainer may still decide that a stance should change — but doing so is a
|
|
71
74
|
deliberate, recorded decision, not an incidental side effect of another change.
|
|
@@ -73,7 +76,7 @@ deliberate, recorded decision, not an incidental side effect of another change.
|
|
|
73
76
|
## Proposing a change
|
|
74
77
|
|
|
75
78
|
Anyone may propose a change by opening an issue. For features, check
|
|
76
|
-
[`
|
|
79
|
+
[`ROADMAP.md`](ROADMAP.md) first — your idea may already be planned, or
|
|
77
80
|
deliberately out of scope. The roadmap is a living document: it reflects current
|
|
78
81
|
intent, not a binding contract, and the maintainers decide what lands in which
|
|
79
82
|
version. For how to actually submit work, see [`CONTRIBUTING.md`](CONTRIBUTING.md).
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
Metadata-Version: 2.4
|
|
2
2
|
Name: generic-ml-wrapper
|
|
3
|
-
Version: 0.
|
|
3
|
+
Version: 0.3.0
|
|
4
4
|
Summary: A wrapper around an ML coding CLI: enter at a job, get a named resumable session, optionally driven by a workflow.
|
|
5
5
|
Project-URL: Homepage, https://github.com/danielslobozian/generic-ml-wrapper
|
|
6
6
|
Project-URL: Repository, https://github.com/danielslobozian/generic-ml-wrapper
|
|
@@ -52,7 +52,7 @@ Run **claude**, **cursor**, **codex**, or **vibe** exactly as you know them —
|
|
|
52
52
|
[](LICENSE)
|
|
53
53
|
[](https://github.com/danielslobozian/generic-ml-wrapper/actions/workflows/ci.yml)
|
|
54
54
|
[](https://sonarcloud.io/summary/overall?id=danielslobozian_generic-ml-wrapper)
|
|
55
|
-
[](pyproject.toml)
|
|
56
56
|
|
|
57
57
|
[](src/generic_ml_wrapper/adapter/outbound/caller/claude_cli_caller.py)
|
|
58
58
|
[](src/generic_ml_wrapper/adapter/outbound/caller/cursor_cli_caller.py)
|
|
@@ -61,7 +61,7 @@ Run **claude**, **cursor**, **codex**, or **vibe** exactly as you know them —
|
|
|
61
61
|
|
|
62
62
|
<br>
|
|
63
63
|
|
|
64
|
-
[Why a wrapper](#why-a-wrapper) • [The job](#the-job) • [Install](#install) • [Workflows](#workflows) • [What it records](#what-it-records) • [Design](docs/DESIGN.md)
|
|
64
|
+
[Why a wrapper](#why-a-wrapper) • [Clients](#clients-at-a-glance) • [The job](#the-job) • [Install](#install) • [Workflows](#workflows) • [What it records](#what-it-records) • [Guides](#guides) • [Design](docs/DESIGN.md)
|
|
65
65
|
|
|
66
66
|
</div>
|
|
67
67
|
|
|
@@ -86,6 +86,19 @@ The judgment stays in the client. The deterministic parts — session identity,
|
|
|
86
86
|
<img src="docs/images/gmlw-usage.gif" alt="gmlw export — per-turn tokens and cost for a job" width="760">
|
|
87
87
|
</div>
|
|
88
88
|
|
|
89
|
+
## Clients at a glance
|
|
90
|
+
|
|
91
|
+
Four clients, one surface — but they are **not** identical. What each supports today:
|
|
92
|
+
|
|
93
|
+
| Client | Metering | Resume | Status line | Context delivery |
|
|
94
|
+
|---|:---:|:---:|:---:|---|
|
|
95
|
+
| **claude** (`claude`) | ✅ | ✅ | ✅ | native `--append-system-prompt-file` |
|
|
96
|
+
| **cursor** (`cursor-agent`) | ❌ | ✅ | ✅ | context-file instruction |
|
|
97
|
+
| **codex** (`codex`) | ✅ | ❌ | ❌ | initial instruction |
|
|
98
|
+
| **vibe** (`vibe`) | ✅ | ❌ | ❌ | initial instruction |
|
|
99
|
+
|
|
100
|
+
Cursor isn't metered by the open-source wrapper (its usage isn't on an interceptable API); Codex and Vibe don't resume. See [`docs/CLIENTS.md`](docs/CLIENTS.md) for the per-client detail and setup.
|
|
101
|
+
|
|
89
102
|
## The job
|
|
90
103
|
|
|
91
104
|
A **job** is the one concept everything hangs off. It's the piece of work you're tagging — a ticket, a refactor, an investigation — and it is the primary key of the whole ledger.
|
|
@@ -110,7 +123,7 @@ Nothing about the client changes — you get its full TUI. The wrapper owns the
|
|
|
110
123
|
|
|
111
124
|
## Install
|
|
112
125
|
|
|
113
|
-
Requires Python 3.11–3.
|
|
126
|
+
Requires Python 3.11–3.14 and [`uv`](https://docs.astral.sh/uv/), plus **one supported coding CLI already installed and logged in** (`claude`, `cursor-agent`, `codex`, or `vibe`). The console script is `gmlw`.
|
|
114
127
|
|
|
115
128
|
```sh
|
|
116
129
|
uv tool install generic-ml-wrapper # or: uv sync --extra dev (from a clone)
|
|
@@ -141,19 +154,49 @@ Everything lives under `~/.gmlw/`, owner-only, on your machine:
|
|
|
141
154
|
| **Context** | `contexts/<job>/<session>.context.md` | the exact compiled context the session launched with |
|
|
142
155
|
| **Transcript** *(opt-in)* | `transcripts/<job>/<session>/call_NNN.{in.json,out.sse,usage.json}` | every call's request, response, and usage — a portable, self-contained folder |
|
|
143
156
|
|
|
144
|
-
For **Claude Code**, the wrapper renders a rich status line straight into the client's own status bar — the git branch, folder, model, context %, and live cost, plus the job's running total. (The status-line seam is client-agnostic; Claude
|
|
157
|
+
For **Claude Code** and **Cursor**, the wrapper renders a rich status line straight into the client's own status bar — the git branch, folder, model, context %, and live cost, plus the job's running total. (The status-line seam is client-agnostic; Claude and Cursor are wired today, and other parsers can be added.)
|
|
145
158
|
|
|
146
159
|
<div align="center">
|
|
147
160
|
<img src="docs/images/gmlw-statusline.gif" alt="the gmlw status line rendered for claude and cursor" width="760">
|
|
148
161
|
</div>
|
|
149
162
|
|
|
163
|
+
## Your operating context, carried across clients
|
|
164
|
+
|
|
165
|
+
Beyond metering, `gmlw` builds a portable operating context that follows you from one client to the next — none of it locked inside a single tool:
|
|
166
|
+
|
|
167
|
+
- **Profile & company** — who you are and your project's conventions (`profile/me`, `profile/company`), composed into every session.
|
|
168
|
+
- **Learned notebook** — what your tools notice about how you work, in one file they all mirror into; negatives ("what to avoid") are first-class.
|
|
169
|
+
- **Personas** — a selectable tone with a free, local greeting at launch (`gmlw persona list`).
|
|
170
|
+
- **Rules** — corrections you've demanded, captured as reusable reflexes (`~/.gmlw/rules/*.rule.md`).
|
|
171
|
+
- **Mode-aware packaging** — a `[startup.<mode>]` matrix decides which sources compose for a plain session, a workflow, or authoring, with optional typed compression.
|
|
172
|
+
- **Plugins & overrides** — swap a client's caller by id (`gmlw plugins list`, `[callers]`).
|
|
173
|
+
|
|
174
|
+
Workflows (above) are *optional* and separate from this personal layer. Every listing/reporting command also speaks `--json` for automation.
|
|
175
|
+
|
|
176
|
+
## What gmlw does not do
|
|
177
|
+
|
|
178
|
+
- It is **not a sandbox** — it launches the real client with your credentials (see [`SECURITY.md`](SECURITY.md)).
|
|
179
|
+
- It does **not** meter Cursor, and does **not** resume Codex or Vibe (see the [matrix](#clients-at-a-glance)).
|
|
180
|
+
- It does **not** call models itself — the optional context compressor records through [`generic-ml-cache`](https://github.com/danielslobozian/generic-ml-cache).
|
|
181
|
+
- It ships **no** compression prompts — the compressor stays inert until you configure one.
|
|
182
|
+
|
|
183
|
+
## Guides
|
|
184
|
+
|
|
185
|
+
- [User guide](docs/USER_GUIDE.md) — task-oriented, first launch through daily work.
|
|
186
|
+
- [CLI reference](docs/CLI.md) — every command, flag, and `--json` option.
|
|
187
|
+
- [Configuration](docs/CONFIGURATION.md) — every `config.toml` section, with examples.
|
|
188
|
+
- [Clients](docs/CLIENTS.md) — per-client behaviour, setup, and limitations.
|
|
189
|
+
- [Workflows](docs/WORKFLOWS.md) — authoring, layout, scripts, rules, credentials.
|
|
190
|
+
- [Troubleshooting](docs/TROUBLESHOOTING.md) — failures, diagnostics, reset/recovery.
|
|
191
|
+
- [Design](docs/DESIGN.md) · [Security](SECURITY.md) · [Contributing](CONTRIBUTING.md)
|
|
192
|
+
|
|
150
193
|
## Develop
|
|
151
194
|
|
|
152
195
|
The gates are defined once in [`noxfile.py`](noxfile.py); CI is a thin caller of them, so what runs locally is byte-for-byte what runs in CI.
|
|
153
196
|
|
|
154
197
|
```sh
|
|
155
198
|
uv sync --extra dev # or: nox -s dev (builds the IDE .venv)
|
|
156
|
-
nox # lint · imports · typecheck · tests (3.11–3.
|
|
199
|
+
nox # lint · imports · typecheck · tests (3.11–3.14)
|
|
157
200
|
nox -s green # the whole gate in one env (lint · format · imports · pyright · coverage)
|
|
158
201
|
```
|
|
159
202
|
|
|
@@ -16,7 +16,7 @@ Run **claude**, **cursor**, **codex**, or **vibe** exactly as you know them —
|
|
|
16
16
|
[](LICENSE)
|
|
17
17
|
[](https://github.com/danielslobozian/generic-ml-wrapper/actions/workflows/ci.yml)
|
|
18
18
|
[](https://sonarcloud.io/summary/overall?id=danielslobozian_generic-ml-wrapper)
|
|
19
|
-
[](pyproject.toml)
|
|
20
20
|
|
|
21
21
|
[](src/generic_ml_wrapper/adapter/outbound/caller/claude_cli_caller.py)
|
|
22
22
|
[](src/generic_ml_wrapper/adapter/outbound/caller/cursor_cli_caller.py)
|
|
@@ -25,7 +25,7 @@ Run **claude**, **cursor**, **codex**, or **vibe** exactly as you know them —
|
|
|
25
25
|
|
|
26
26
|
<br>
|
|
27
27
|
|
|
28
|
-
[Why a wrapper](#why-a-wrapper) • [The job](#the-job) • [Install](#install) • [Workflows](#workflows) • [What it records](#what-it-records) • [Design](docs/DESIGN.md)
|
|
28
|
+
[Why a wrapper](#why-a-wrapper) • [Clients](#clients-at-a-glance) • [The job](#the-job) • [Install](#install) • [Workflows](#workflows) • [What it records](#what-it-records) • [Guides](#guides) • [Design](docs/DESIGN.md)
|
|
29
29
|
|
|
30
30
|
</div>
|
|
31
31
|
|
|
@@ -50,6 +50,19 @@ The judgment stays in the client. The deterministic parts — session identity,
|
|
|
50
50
|
<img src="docs/images/gmlw-usage.gif" alt="gmlw export — per-turn tokens and cost for a job" width="760">
|
|
51
51
|
</div>
|
|
52
52
|
|
|
53
|
+
## Clients at a glance
|
|
54
|
+
|
|
55
|
+
Four clients, one surface — but they are **not** identical. What each supports today:
|
|
56
|
+
|
|
57
|
+
| Client | Metering | Resume | Status line | Context delivery |
|
|
58
|
+
|---|:---:|:---:|:---:|---|
|
|
59
|
+
| **claude** (`claude`) | ✅ | ✅ | ✅ | native `--append-system-prompt-file` |
|
|
60
|
+
| **cursor** (`cursor-agent`) | ❌ | ✅ | ✅ | context-file instruction |
|
|
61
|
+
| **codex** (`codex`) | ✅ | ❌ | ❌ | initial instruction |
|
|
62
|
+
| **vibe** (`vibe`) | ✅ | ❌ | ❌ | initial instruction |
|
|
63
|
+
|
|
64
|
+
Cursor isn't metered by the open-source wrapper (its usage isn't on an interceptable API); Codex and Vibe don't resume. See [`docs/CLIENTS.md`](docs/CLIENTS.md) for the per-client detail and setup.
|
|
65
|
+
|
|
53
66
|
## The job
|
|
54
67
|
|
|
55
68
|
A **job** is the one concept everything hangs off. It's the piece of work you're tagging — a ticket, a refactor, an investigation — and it is the primary key of the whole ledger.
|
|
@@ -74,7 +87,7 @@ Nothing about the client changes — you get its full TUI. The wrapper owns the
|
|
|
74
87
|
|
|
75
88
|
## Install
|
|
76
89
|
|
|
77
|
-
Requires Python 3.11–3.
|
|
90
|
+
Requires Python 3.11–3.14 and [`uv`](https://docs.astral.sh/uv/), plus **one supported coding CLI already installed and logged in** (`claude`, `cursor-agent`, `codex`, or `vibe`). The console script is `gmlw`.
|
|
78
91
|
|
|
79
92
|
```sh
|
|
80
93
|
uv tool install generic-ml-wrapper # or: uv sync --extra dev (from a clone)
|
|
@@ -105,19 +118,49 @@ Everything lives under `~/.gmlw/`, owner-only, on your machine:
|
|
|
105
118
|
| **Context** | `contexts/<job>/<session>.context.md` | the exact compiled context the session launched with |
|
|
106
119
|
| **Transcript** *(opt-in)* | `transcripts/<job>/<session>/call_NNN.{in.json,out.sse,usage.json}` | every call's request, response, and usage — a portable, self-contained folder |
|
|
107
120
|
|
|
108
|
-
For **Claude Code**, the wrapper renders a rich status line straight into the client's own status bar — the git branch, folder, model, context %, and live cost, plus the job's running total. (The status-line seam is client-agnostic; Claude
|
|
121
|
+
For **Claude Code** and **Cursor**, the wrapper renders a rich status line straight into the client's own status bar — the git branch, folder, model, context %, and live cost, plus the job's running total. (The status-line seam is client-agnostic; Claude and Cursor are wired today, and other parsers can be added.)
|
|
109
122
|
|
|
110
123
|
<div align="center">
|
|
111
124
|
<img src="docs/images/gmlw-statusline.gif" alt="the gmlw status line rendered for claude and cursor" width="760">
|
|
112
125
|
</div>
|
|
113
126
|
|
|
127
|
+
## Your operating context, carried across clients
|
|
128
|
+
|
|
129
|
+
Beyond metering, `gmlw` builds a portable operating context that follows you from one client to the next — none of it locked inside a single tool:
|
|
130
|
+
|
|
131
|
+
- **Profile & company** — who you are and your project's conventions (`profile/me`, `profile/company`), composed into every session.
|
|
132
|
+
- **Learned notebook** — what your tools notice about how you work, in one file they all mirror into; negatives ("what to avoid") are first-class.
|
|
133
|
+
- **Personas** — a selectable tone with a free, local greeting at launch (`gmlw persona list`).
|
|
134
|
+
- **Rules** — corrections you've demanded, captured as reusable reflexes (`~/.gmlw/rules/*.rule.md`).
|
|
135
|
+
- **Mode-aware packaging** — a `[startup.<mode>]` matrix decides which sources compose for a plain session, a workflow, or authoring, with optional typed compression.
|
|
136
|
+
- **Plugins & overrides** — swap a client's caller by id (`gmlw plugins list`, `[callers]`).
|
|
137
|
+
|
|
138
|
+
Workflows (above) are *optional* and separate from this personal layer. Every listing/reporting command also speaks `--json` for automation.
|
|
139
|
+
|
|
140
|
+
## What gmlw does not do
|
|
141
|
+
|
|
142
|
+
- It is **not a sandbox** — it launches the real client with your credentials (see [`SECURITY.md`](SECURITY.md)).
|
|
143
|
+
- It does **not** meter Cursor, and does **not** resume Codex or Vibe (see the [matrix](#clients-at-a-glance)).
|
|
144
|
+
- It does **not** call models itself — the optional context compressor records through [`generic-ml-cache`](https://github.com/danielslobozian/generic-ml-cache).
|
|
145
|
+
- It ships **no** compression prompts — the compressor stays inert until you configure one.
|
|
146
|
+
|
|
147
|
+
## Guides
|
|
148
|
+
|
|
149
|
+
- [User guide](docs/USER_GUIDE.md) — task-oriented, first launch through daily work.
|
|
150
|
+
- [CLI reference](docs/CLI.md) — every command, flag, and `--json` option.
|
|
151
|
+
- [Configuration](docs/CONFIGURATION.md) — every `config.toml` section, with examples.
|
|
152
|
+
- [Clients](docs/CLIENTS.md) — per-client behaviour, setup, and limitations.
|
|
153
|
+
- [Workflows](docs/WORKFLOWS.md) — authoring, layout, scripts, rules, credentials.
|
|
154
|
+
- [Troubleshooting](docs/TROUBLESHOOTING.md) — failures, diagnostics, reset/recovery.
|
|
155
|
+
- [Design](docs/DESIGN.md) · [Security](SECURITY.md) · [Contributing](CONTRIBUTING.md)
|
|
156
|
+
|
|
114
157
|
## Develop
|
|
115
158
|
|
|
116
159
|
The gates are defined once in [`noxfile.py`](noxfile.py); CI is a thin caller of them, so what runs locally is byte-for-byte what runs in CI.
|
|
117
160
|
|
|
118
161
|
```sh
|
|
119
162
|
uv sync --extra dev # or: nox -s dev (builds the IDE .venv)
|
|
120
|
-
nox # lint · imports · typecheck · tests (3.11–3.
|
|
163
|
+
nox # lint · imports · typecheck · tests (3.11–3.14)
|
|
121
164
|
nox -s green # the whole gate in one env (lint · format · imports · pyright · coverage)
|
|
122
165
|
```
|
|
123
166
|
|
|
@@ -0,0 +1,38 @@
|
|
|
1
|
+
# Release checklist
|
|
2
|
+
|
|
3
|
+
Steps to cut a `generic-ml-wrapper` release. The point is to keep the version and the
|
|
4
|
+
docs in lockstep; the checks under [`tests/test_docs_consistency.py`](tests/test_docs_consistency.py)
|
|
5
|
+
enforce the parts that can be enforced.
|
|
6
|
+
|
|
7
|
+
## 1. Version & metadata (these must all agree)
|
|
8
|
+
|
|
9
|
+
- [ ] `pyproject.toml` — `version`
|
|
10
|
+
- [ ] `src/generic_ml_wrapper/__init__.py` — `__version__`
|
|
11
|
+
- [ ] `VERSION`
|
|
12
|
+
- [ ] `CHANGELOG.md` — move `[Unreleased]` items under a new `[x.y.z] - YYYY-MM-DD`, and update the link refs at the bottom
|
|
13
|
+
- [ ] `SECURITY.md` — the stated current version
|
|
14
|
+
- [ ] `ROADMAP.md` — move the shipped milestone into "Shipped"
|
|
15
|
+
|
|
16
|
+
`test_version_is_consistent_across_sources` fails if `__version__`, `VERSION`, and
|
|
17
|
+
`pyproject.toml` disagree.
|
|
18
|
+
|
|
19
|
+
## 2. Supported Python
|
|
20
|
+
|
|
21
|
+
- [ ] `pyproject.toml` classifiers, `noxfile.py` `PYTHON_VERSIONS`, the README badge/text, and `CONTRIBUTING.md` all state the same range
|
|
22
|
+
|
|
23
|
+
## 3. Gate & audit
|
|
24
|
+
|
|
25
|
+
- [ ] `nox -s green` passes (lint · format · imports · pyright · coverage)
|
|
26
|
+
- [ ] the full CI matrix is green (Python 3.11–3.14 × Linux / macOS / Windows)
|
|
27
|
+
- [ ] `./secret-audit.sh .` reports clean
|
|
28
|
+
|
|
29
|
+
## 4. Docs
|
|
30
|
+
|
|
31
|
+
- [ ] README capability matrix and the [`docs/`](docs/) guides reflect the release
|
|
32
|
+
- [ ] relative links resolve (`test_relative_markdown_links_resolve`) and the CLI reference covers every command (`test_cli_reference_documents_every_command`)
|
|
33
|
+
|
|
34
|
+
## 5. Tag & publish
|
|
35
|
+
|
|
36
|
+
- [ ] merge to `main`; tag `vX.Y.Z`
|
|
37
|
+
- [ ] create the GitHub Release — `release.yml` publishes to PyPI via Trusted Publishing (no token)
|
|
38
|
+
- [ ] confirm the new version on PyPI and that `uv tool install generic-ml-wrapper` pulls it
|
|
@@ -0,0 +1,173 @@
|
|
|
1
|
+
<!-- SPDX-FileCopyrightText: 2026 Daniel Slobozian -->
|
|
2
|
+
<!-- SPDX-License-Identifier: Apache-2.0 -->
|
|
3
|
+
|
|
4
|
+
# Roadmap
|
|
5
|
+
|
|
6
|
+
Where `generic-ml-wrapper` has been and where it's going. Direction, not a
|
|
7
|
+
commitment — see [CHANGELOG.md](CHANGELOG.md) for what has actually shipped.
|
|
8
|
+
|
|
9
|
+
The through-line: turn a metering wrapper into something a person *wants* to open —
|
|
10
|
+
a companion that carries **who you are and how you work across every client**, and
|
|
11
|
+
stays a wrapper (it extends each client; it never reimplements one).
|
|
12
|
+
|
|
13
|
+
## Shipped
|
|
14
|
+
|
|
15
|
+
### 0.1.0
|
|
16
|
+
The metering foundation — jobs & sessions, four clients (claude / cursor / codex /
|
|
17
|
+
vibe), a per-turn metering relay into a SQLite ledger, a client-aware status line,
|
|
18
|
+
and the workflow system. On PyPI.
|
|
19
|
+
|
|
20
|
+
### 0.2.0 — the companion
|
|
21
|
+
Everything a session inherits, and the ergonomics around it.
|
|
22
|
+
|
|
23
|
+
- **First-run init** — detect installed clients, seed a filled config with a default.
|
|
24
|
+
- **Config-driven, mode-aware context packaging** — a `[startup.<mode>]` matrix picks
|
|
25
|
+
which sources compose (persona · profile · learned · company · rules · workflow
|
|
26
|
+
base/steps) for each mode (default / workflow / authoring), with **typed per-source
|
|
27
|
+
compression** (`[compress.prompts]`).
|
|
28
|
+
- **Personas** — selectable tone with a universal floor; `gmlw persona list`; a free,
|
|
29
|
+
local **host greeting** at launch; a first-run persona choice.
|
|
30
|
+
- **Learned** — a portable, user-owned notebook the client mirrors into, read into
|
|
31
|
+
every client so what one learns they all inherit; **negatives are first-class**.
|
|
32
|
+
- **Rule format** — a domain-neutral `Rule / When / Signals / Strength / Origin`
|
|
33
|
+
(+ optional `Precedence`), with in-session capture during a workflow.
|
|
34
|
+
- **Formal plugins folder** — reference a caller by id (`~/.gmlw/plugins/<id>/`),
|
|
35
|
+
`gmlw plugins list`.
|
|
36
|
+
- **Cursor allowance block** — the status line renders cursor's plan pools from an
|
|
37
|
+
optional local cache when the client doesn't pipe them.
|
|
38
|
+
- **Ergonomics** — client & working-directory preflight (clear guidance instead of a
|
|
39
|
+
cryptic crash), implicit `gmlw <job>` (shorthand for `start`), a friendly no-job
|
|
40
|
+
message, and auto-help for an incomplete sub-command.
|
|
41
|
+
- **Create-workflow** — the interviewer assesses each step's *codeability* and offers
|
|
42
|
+
to script the mechanical ones, keeping the AI for the judgment steps.
|
|
43
|
+
|
|
44
|
+
## Planned
|
|
45
|
+
|
|
46
|
+
### 0.3.0 — lifecycle action hooks — done (Unreleased)
|
|
47
|
+
gmlw already hooks **content**: the interceptor chain transforms context sections at
|
|
48
|
+
compile time and wire traffic at relay time (e.g. anonymisation on `request`). It now
|
|
49
|
+
also hooks **actions** — "after this phase, run something" — at two lifecycle seams
|
|
50
|
+
that bracket the client run:
|
|
51
|
+
|
|
52
|
+
- **`pre-launch`** — after the context is compiled and the caller resolved, before the
|
|
53
|
+
client starts. Knows the client, the working directory, and the session. For
|
|
54
|
+
per-client setup: deploying skills/rules into a client's *native* mechanism, writing
|
|
55
|
+
MCP config, warming a cache.
|
|
56
|
+
- **`post-session`** — after the client exits. Knows the exit code and the session.
|
|
57
|
+
For cleanup, notification, archival, or roll-up.
|
|
58
|
+
|
|
59
|
+
Config mirrors the existing extension points — a `[[hooks]]` entry with a `phase`, a
|
|
60
|
+
`spec` (a plugin id or `path.py:Class`), and an optional `client` scope — under the
|
|
61
|
+
same trusted-code boundary as `[[interceptors]]`, `[callers]`, and plugins. Hooks are
|
|
62
|
+
best-effort and never break a launch. Both launch paths (`start`, `workflow new`) route
|
|
63
|
+
through one shared launch sequence, so the seams bracket every run. Shipped with a
|
|
64
|
+
built-in `SessionLogger` reference hook.
|
|
65
|
+
|
|
66
|
+
**Still open — the flagship example hook:** a **cross-client skills/rules deployer** that
|
|
67
|
+
consumes a git repo of skills and installs them, per client, as faithfully as each
|
|
68
|
+
client's format allows. It is its own initiative (per-client format knowledge), built on
|
|
69
|
+
the seam above; the infrastructure is done and waiting for it.
|
|
70
|
+
|
|
71
|
+
#### Rule lifecycle — done (Unreleased)
|
|
72
|
+
0.2.0 shipped rule *capture*, but only inside a workflow (the directive lived in the
|
|
73
|
+
workflow base) and without dedup or a path to code. The rule loop is now rounded out —
|
|
74
|
+
the directive moved out of the workflow base to the head of the always-on `rules`
|
|
75
|
+
context source (verbatim, gmlw's voice), and the source is active by default in a plain
|
|
76
|
+
start:
|
|
77
|
+
|
|
78
|
+
- **Rule proposal in normal usage** — the "offer to record a rule" directive is
|
|
79
|
+
always-on, so a demanded correction becomes a draft rule wherever it happens, not only
|
|
80
|
+
inside a workflow.
|
|
81
|
+
- **Existing-rule check (dedup / update-not-duplicate)** — before proposing, the client
|
|
82
|
+
reads the existing rules and updates or supersedes a matching one instead of stacking a
|
|
83
|
+
near-duplicate, mirroring the learned notebook's supersede-on-contradiction.
|
|
84
|
+
- **Rule → code feasibility** — a captured rule is judged for *mechanical
|
|
85
|
+
enforceability* and, if so, offered as a script/check rather than a reminder — the
|
|
86
|
+
step-codeability logic (create-workflow) generalised from workflow steps to rules, and
|
|
87
|
+
a natural future `pre-launch` hook consumer.
|
|
88
|
+
|
|
89
|
+
### 0.4.0 — discoverability & first-run: getting captivated fast
|
|
90
|
+
Reframed around the real metric for a new user: **time to a first live session.**
|
|
91
|
+
Captivation lives in the model session, not the wrapper — so the job is to remove
|
|
92
|
+
everything between a stranger and that first response, then reveal depth *gradually*
|
|
93
|
+
(progressive disclosure). Because the wrapper cedes the screen to the client, discovery
|
|
94
|
+
lives in the thin surfaces **around** a run — pre-launch, run-and-return commands, the
|
|
95
|
+
return, and ambient context pushed *into* the client — never a persistent UI over a
|
|
96
|
+
live session.
|
|
97
|
+
|
|
98
|
+
- **First-run onboarding** — on first launch, a greeting instead of the help menu:
|
|
99
|
+
*"looks like your first time — set up (about a minute), or skip? (`gmlw init`
|
|
100
|
+
anytime)."* Setup establishes the two things that shape every later run:
|
|
101
|
+
- **A working client** — the one hard requirement (no client, no session). *Fast path*
|
|
102
|
+
when a client is already installed and authenticated: detect it and reach a session
|
|
103
|
+
in seconds. *Guided path* when none is: a subscription→client mapping (*"do you pay
|
|
104
|
+
for Claude / ChatGPT / Cursor / Mistral?"* → the matching supported client CLI), a
|
|
105
|
+
link/script to install, and **guide-and-verify auth** — print the exact login
|
|
106
|
+
command, then poll readiness until it goes green (guide, don't drive the login).
|
|
107
|
+
- **A default persona** — the experience-defining choice (otherwise every launch needs
|
|
108
|
+
a flag). Picked from **static previews** (sample lines per persona), not a live
|
|
109
|
+
model quiz, so choosing costs nothing.
|
|
110
|
+
- **Bare `gmlw` is first-run-aware** — first time → the greeting/onboarding; thereafter →
|
|
111
|
+
a grouped capability index (*launch / inspect / author*), with `gmlw help <topic>` for
|
|
112
|
+
the core concepts (job vs workflow, start vs run, personas, cost). `--help` keeps the
|
|
113
|
+
argparse view; every listing ends with a next-action footer.
|
|
114
|
+
- **Config registry** — one typed source of truth (a `pydantic-settings` model) for
|
|
115
|
+
every setting: key, type, default, allowed values, description. It replaces the
|
|
116
|
+
hand-rolled accessors, and every surface (help, `config`, per-workflow settings)
|
|
117
|
+
renders from it — so the config file no longer has to be pre-populated to be found.
|
|
118
|
+
- **`config` commands** — `config list / get / set`, rendering the registry with inline
|
|
119
|
+
allowed-value validation.
|
|
120
|
+
- **`--version`** — surface the running version.
|
|
121
|
+
- **Progressive disclosure** — depth is revealed over time, not dumped up front:
|
|
122
|
+
- **Exit receipt** — on the return (client exit), a persistent summary: this session's
|
|
123
|
+
and the job's cost, the resume/report commands, and one usage-driven, suppressible
|
|
124
|
+
tip (the channel that surfaces features as they become relevant). Also the home for
|
|
125
|
+
**authoring-cost visibility** — workflow authoring is already a metered session;
|
|
126
|
+
surface it in listings/export instead of hiding it, keeping the work/authoring split.
|
|
127
|
+
- **Ambient capability card** — an optional, off-by-default context injection so the
|
|
128
|
+
client itself answers "how do I …" gmlw questions mid-session. Kept minimal and
|
|
129
|
+
counted against the context budget.
|
|
130
|
+
- **Greeting → context** — retire the launch-time host greeting (structurally invisible:
|
|
131
|
+
the client clears the screen) and inject it as a context instruction the client
|
|
132
|
+
renders in-band, with live metadata (time, name, recent activity).
|
|
133
|
+
- **`workflow edit`** — update an existing workflow, not only create.
|
|
134
|
+
- **Robustness** — clean interrupt/exit: catch Ctrl+C / an interrupt so leaving a client
|
|
135
|
+
session ends cleanly instead of surfacing the child's raw error.
|
|
136
|
+
|
|
137
|
+
### 0.5.0 — the workflow, first-class
|
|
138
|
+
Two ways people relate to a workflow: *applied to a job* (a ticket it treats), or *the
|
|
139
|
+
recurring job itself* (a repeatable extraction). Make both first-class, and make
|
|
140
|
+
authoring one a guided conversation — because most people don't arrive with a clear
|
|
141
|
+
model of their own process.
|
|
142
|
+
|
|
143
|
+
- **`run <workflow>`** — launch a workflow directly; the job defaults to the workflow
|
|
144
|
+
name and sessions accumulate under it. `start <job> [--workflow]` stays for the
|
|
145
|
+
applied case. A trigger-gated pre-launch chooser fills only *missing* arguments and
|
|
146
|
+
always echoes the equivalent one-liner, so interactive use teaches the fast path
|
|
147
|
+
(full argv never gains a prompt).
|
|
148
|
+
- **Name at the end, not the start** — `workflow new` no longer demands a name up front;
|
|
149
|
+
forcing one presumes the user already knows the shape. The authoring conversation
|
|
150
|
+
shapes the workflow and *proposes* the name at convergence. (Distinct from
|
|
151
|
+
`run <workflow>`, where an existing, recurring workflow's name has already earned it.)
|
|
152
|
+
- **Facilitative + constructive authoring** — the create-workflow conversation as a
|
|
153
|
+
blended consultant. Two axes: *facilitate* (a **parking lot** for tangents so nothing
|
|
154
|
+
is lost, reflective listening, a **diverge → converge** phase model, and **process
|
|
155
|
+
leveling** to answer "step or its own workflow?") and *contribute* (start in inquiry
|
|
156
|
+
and **move to expert when warranted**, propose the upstream/downstream stages the
|
|
157
|
+
author omitted, surface **implications** they haven't hit) — bounded by guardrails
|
|
158
|
+
against railroading a novice and a **consent gate** on anything personal. Authoring
|
|
159
|
+
keeps distilled state (a draft plus the parking lot) as files in the workflow folder,
|
|
160
|
+
so it survives context compaction.
|
|
161
|
+
- **Per-workflow persona, composed as voice over method** — a workflow can carry its own
|
|
162
|
+
persona instead of the single global tone. Persona is **manner**; the authoring
|
|
163
|
+
behavior above is **method**. They compose by layer: the method lives in the mode
|
|
164
|
+
floor (guaranteed, persona-independent), and the persona colors delivery on top — it
|
|
165
|
+
can voice the invariants but not break them. Personas carry an **authoring-friendliness**
|
|
166
|
+
expectation, so a voice built around brutal efficiency can't sabotage the facilitation.
|
|
167
|
+
|
|
168
|
+
## Parked
|
|
169
|
+
|
|
170
|
+
- **External source connectors** — let a workflow pull from external systems (APIs,
|
|
171
|
+
cloud storage, platforms) rather than manual input. Large; its own initiative.
|
|
172
|
+
- Relay extraction to a standalone project (shelved — no second consumer yet).
|
|
173
|
+
- Workspace layout feature.
|