alab-cli 0.1.0__tar.gz → 0.1.2__tar.gz

{alab_cli-0.1.0 → alab_cli-0.1.2}/PKG-INFO

@@ -1,12 +1,13 @@
 Metadata-Version: 2.4
 Name: alab-cli
-Version: 0.1.0
+Version: 0.1.2
 Summary: Local agent-first experiment workbench CLI
 Author: ALab contributors
 License-Expression: GPL-3.0-or-later
 Project-URL: Homepage, https://github.com/bebetterest/ALab
 Project-URL: Repository, https://github.com/bebetterest/ALab
 Project-URL: Issues, https://github.com/bebetterest/ALab/issues
+Project-URL: Changelog, https://github.com/bebetterest/ALab/blob/main/CHANGELOG.md
 Requires-Python: >=3.11
 Description-Content-Type: text/markdown
 License-File: LICENSE
@@ -23,13 +24,22 @@ Dynamic: license-file
   <img src="docs/assets/readme-header.png" alt="Hand-drawn ALab virtual experiment workbench banner" width="100%">
 </p>
 
+> 🤖 Most of the code is completed by [Codex](https://openai.com/codex/).
+>
+> 📮 Feedback: [bebetterest@outlook.com](mailto:bebetterest@outlook.com)
+>
+> [English](README.md) | [中文](README_cn.md)
+
 ALab is a local, agent-first Python CLI workbench for iterative experiments. It lets external agents work in isolated Git worktrees, run repeatable evaluations, submit final results, and inspect visible prior experiment evidence through explicit collaboration boundaries.
 
-ALab V1 is intentionally local-only: no server, sync service, web UI, built-in agent launcher, or account system. It owns the local project records, source snapshots, experiment lifecycle, runner execution, logs, artifacts, and visibility rules; agents remain external CLI operators.
+ALab V1 is intentionally local-only: no hosted service, sync service, remote web UI, built-in agent launcher, or account system. It owns the local project records, source snapshots, experiment lifecycle, runner execution, logs, artifacts, visibility rules, and a root-only local read-only dashboard; agents remain external CLI operators.
 
 ## Highlights
 
 - Local CLI workbench for projects, sources, experiments, runs, submissions, logs, artifacts, annotations, and audits.
+- HOME-level feedback capture so any ALab role can leave local suggestions, questions, or bug reports without needing project credentials.
+- Root-only local dashboard for browser-based read-only inspection of global/project/experiment/run/log/artifact/audit/feedback/system state.
+- Markdown evidence reports for project owners or visible experiment contexts through `alab report`.
 - Context-aware command surface: `alab help` and command preflight show only what the current project, experiment, inspection checkout, token, or explicit key can use.
 - Git-backed experiment isolation: each experiment is an isolated branch/worktree with a worktree token for run and submit operations.
 - Reproducible project setup: project config controls runner, reward, artifact capture, environment, secrets, mutable paths, and visibility.
@@ -198,6 +208,9 @@ Useful next commands:
 
 ```sh
 alab help
+alab feedback --kind suggestion --body "Describe a suggestion, question, or bug for the project owner."
+alab --key <root-key> dashboard --no-open
+alab --key <admin-or-root-key> report --project <project-id> --out ./alab-project-report.md
 alab observe experiments list
 alab observe runs list --exp <exp-id>
 alab observe experiments best
@@ -211,6 +224,9 @@ alab observe experiments best
 - **Experiment**: isolated Git branch/worktree bound to exactly one source and one config version.
 - **Run**: one evaluator execution for an experiment commit, with status, reward, logs, artifacts, and warning codes.
 - **Submit**: closes an experiment with final summary, feedback, final run, final commit, and explicit refs.
+- **Feedback**: HOME-level plaintext notes for suggestions, questions, bugs, or other agent observations, stored as one directory per entry under `feedback/`.
+- **Dashboard**: root-only local browser panel for read-only inspection. It binds to `127.0.0.1`, uses a random browser session token, serves bounded paginated read models for large homes, and writes no ALab records.
+- **Reports**: Markdown evidence exports for project-level owner handoffs or visible experiment contexts, without raw secrets, hidden log contents, or artifact bytes.
 - **Inspection checkout**: read-only checkout for observing/exporting scoped experiment evidence without becoming submit-capable.
 
 ## Configuration
@@ -228,7 +244,14 @@ See [docs/spec_runners_adapters.md](docs/spec_runners_adapters.md), [docs/spec_p
 
 ## Examples
 
-- [examples/skydiscover_circle_packing_codex](examples/skydiscover_circle_packing_codex/): complete SkyDiscover circle-packing example with setup, single Codex worker run, project-controller loop, report collection, and real-run notes.
+See [examples](examples/) for the runnable example matrix. The current examples
+cover a local scoring loop, a Dockerized clinic-order fulfillment planner with
+artifact export, a Harbor hidden-verifier incident classifier, a collaborative
+incident triage lifecycle workflow, a generated dashboard showcase home, and
+the SkyDiscover circle-packing Codex single-worker protocol. The same examples area also includes
+`examples/templates/`, a copyable multi-instance TSP template library for
+local, Docker, Harbor, SkyDiscover Python, and SkyDiscover Docker runner
+projects.
 
 The repository also includes Codex-facing role skills under [skills](skills/). They are external runbooks for operating ALab through the CLI as an experiment worker, project controller, or global admin; they do not add an embedded agent launcher to ALab.
 
@@ -245,10 +268,10 @@ Opt-in integration gates are excluded from the default suite:
 
 ```sh
 ALAB_RUN_REAL_DOCKER=1 UV_CACHE_DIR=.uv-cache UV_DEFAULT_INDEX=https://pypi.org/simple uv run pytest -m real_docker
-ALAB_RUN_REAL_SKYDISCOVER_PYTHON=1 UV_CACHE_DIR=.uv-cache uv run pytest -m real_skydiscover_python
+ALAB_RUN_REAL_SKYDISCOVER_PYTHON=1 UV_CACHE_DIR=.uv-cache UV_DEFAULT_INDEX=https://pypi.org/simple uv run pytest -m real_skydiscover_python
 ALAB_RUN_NETWORKED_SKYDISCOVER_PYTHON=1 UV_CACHE_DIR=.uv-cache UV_DEFAULT_INDEX=https://pypi.org/simple uv run pytest -m networked_skydiscover_python
 ALAB_RUN_NATIVE_SKYDISCOVER_PYTHON=1 UV_CACHE_DIR=.uv-cache UV_DEFAULT_INDEX=https://pypi.org/simple uv run pytest -m native_skydiscover_python
-ALAB_RUN_LIVE_SKYDISCOVER_CATALOG=1 UV_CACHE_DIR=.uv-cache uv run pytest -m live_skydiscover_catalog
+ALAB_RUN_LIVE_SKYDISCOVER_CATALOG=1 UV_CACHE_DIR=.uv-cache UV_DEFAULT_INDEX=https://pypi.org/simple uv run pytest -m live_skydiscover_catalog
 ```
 
 Notes:
@@ -256,7 +279,7 @@ Notes:
 - `uv.lock` is tracked because CI and local validation use `uv run --locked`.
 - Keep local cache/output paths ignored (`.uv-cache/`, `.pytest_cache/`, `.ruff_cache/`, `.alab-demo/`, `.env`).
 - GitHub Actions runs the default lint and pytest suite on pull requests and pushes to `main`; real Docker and live/networked SkyDiscover gates remain manual workflow inputs.
-- Pushes to `main` check PyPI for the current `pyproject.toml` package version; if that exact version is missing, CI builds and publishes through PyPI Trusted Publishing, otherwise it skips publishing.
+- Pushes to `main` check PyPI for the current `pyproject.toml` package version; if that exact version is missing, CI builds and publishes through PyPI Trusted Publishing, then creates a `v<version>` GitHub Release from the matching [CHANGELOG.md](CHANGELOG.md) section. If PyPI already has that version, publishing and release creation are skipped.
 - The PyPI `alab-cli` project must trust repository `bebetterest/ALab`, workflow `ci.yml`, and environment `pypi` before the first automated publish can succeed.
 
 ## Security And Data Model
@@ -268,6 +291,7 @@ ALab V1 is a local collaboration boundary, not a multi-user security product:
 - Credential verifiers are stored; raw credential secrets are not.
 - Project records are local plaintext SQLite/filesystem data.
 - `secret_env` values are local plaintext, redacted from rendered logs where configured, and never exported by config commands.
+- The dashboard is loopback-only and root-only. Its API can read hidden/full logs and artifact bytes for that root session, but it must not render raw keys, raw tokens, credential verifier material, or raw `secret_env` values.
 - Artifact exports are exact captured bytes and are not automatically redacted.
 
 ## Repository Structure
@@ -280,6 +304,8 @@ ALab V1 is a local collaboration boundary, not a multi-user security product:
 ├── docs/
 │   ├── assets/
 │   │   └── readme-header.png
+│   ├── README.md
+│   ├── README_cn.md
 │   ├── blueprint.md
 │   ├── blueprint_cn.md
 │   ├── completion_audit.md
@@ -289,7 +315,15 @@ ALab V1 is a local collaboration boundary, not a multi-user security product:
 │   ├── progress_pipeline.md
 │   └── progress_pipeline_cn.md
 ├── examples/
-│   └── skydiscover_circle_packing_codex/
+│   ├── README.md
+│   ├── README_cn.md
+│   ├── collaboration_observe_lifecycle/
+│   ├── dashboard_showcase/
+│   ├── docker_file_reward_artifacts/
+│   ├── harbor_verifier_minimal/
+│   ├── local_agent_scoreboard/
+│   ├── skydiscover_circle_packing_codex/
+│   └── templates/
 ├── skills/
 │   ├── alab-experiment-worker/
 │   ├── alab-project-controller/
@@ -304,6 +338,8 @@ ALab V1 is a local collaboration boundary, not a multi-user security product:
 │   └── test_runner_skydiscover.py
 ├── LICENSE
 ├── .env.example
+├── CHANGELOG.md
+├── CHANGELOG_cn.md
 ├── pyproject.toml
 ├── uv.lock
 ├── README.md
@@ -316,10 +352,12 @@ Local-only agent notes such as `AGENTS.md` and `CORE.md` are intentionally git-i
 
 - English documentation is canonical.
 - Synchronized Chinese documents use the `*_cn.md` naming pattern.
+- [docs/README.md](docs/README.md) explains the documentation structure and reading order.
 - [docs/blueprint.md](docs/blueprint.md) is the V1 product overview.
-- [docs/spec_cli.md](docs/spec_cli.md), [docs/spec_storage_auth_context.md](docs/spec_storage_auth_context.md), [docs/spec_project_source_experiment.md](docs/spec_project_source_experiment.md), [docs/spec_lifecycle.md](docs/spec_lifecycle.md), [docs/spec_runners_adapters.md](docs/spec_runners_adapters.md), [docs/spec_observe_collaboration.md](docs/spec_observe_collaboration.md), and [docs/spec_tests.md](docs/spec_tests.md) define subsystem contracts.
+- [docs/spec_cli.md](docs/spec_cli.md), [docs/spec_storage_auth_context.md](docs/spec_storage_auth_context.md), [docs/spec_project_source_experiment.md](docs/spec_project_source_experiment.md), [docs/spec_lifecycle.md](docs/spec_lifecycle.md), [docs/spec_runners_adapters.md](docs/spec_runners_adapters.md), [docs/spec_observe_collaboration.md](docs/spec_observe_collaboration.md), [docs/spec_dashboard.md](docs/spec_dashboard.md), and [docs/spec_tests.md](docs/spec_tests.md) define subsystem contracts.
 - [docs/progress.md](docs/progress.md), [docs/progress_pipeline.md](docs/progress_pipeline.md), [docs/progress_closed_gaps.md](docs/progress_closed_gaps.md), and [docs/progress_log.md](docs/progress_log.md) track current state, active queues, closed gaps, and history.
 - [docs/completion_audit.md](docs/completion_audit.md) tracks requirement-level evidence.
+- [CHANGELOG.md](CHANGELOG.md) records release-facing changes and is used by CI for GitHub Release notes.
 
 ## Contributing
 

{alab_cli-0.1.0 → alab_cli-0.1.2}/README.md

@@ -4,13 +4,22 @@
   <img src="docs/assets/readme-header.png" alt="Hand-drawn ALab virtual experiment workbench banner" width="100%">
 </p>
 
+> 🤖 Most of the code is completed by [Codex](https://openai.com/codex/).
+>
+> 📮 Feedback: [bebetterest@outlook.com](mailto:bebetterest@outlook.com)
+>
+> [English](README.md) | [中文](README_cn.md)
+
 ALab is a local, agent-first Python CLI workbench for iterative experiments. It lets external agents work in isolated Git worktrees, run repeatable evaluations, submit final results, and inspect visible prior experiment evidence through explicit collaboration boundaries.
 
-ALab V1 is intentionally local-only: no server, sync service, web UI, built-in agent launcher, or account system. It owns the local project records, source snapshots, experiment lifecycle, runner execution, logs, artifacts, and visibility rules; agents remain external CLI operators.
+ALab V1 is intentionally local-only: no hosted service, sync service, remote web UI, built-in agent launcher, or account system. It owns the local project records, source snapshots, experiment lifecycle, runner execution, logs, artifacts, visibility rules, and a root-only local read-only dashboard; agents remain external CLI operators.
 
 ## Highlights
 
 - Local CLI workbench for projects, sources, experiments, runs, submissions, logs, artifacts, annotations, and audits.
+- HOME-level feedback capture so any ALab role can leave local suggestions, questions, or bug reports without needing project credentials.
+- Root-only local dashboard for browser-based read-only inspection of global/project/experiment/run/log/artifact/audit/feedback/system state.
+- Markdown evidence reports for project owners or visible experiment contexts through `alab report`.
 - Context-aware command surface: `alab help` and command preflight show only what the current project, experiment, inspection checkout, token, or explicit key can use.
 - Git-backed experiment isolation: each experiment is an isolated branch/worktree with a worktree token for run and submit operations.
 - Reproducible project setup: project config controls runner, reward, artifact capture, environment, secrets, mutable paths, and visibility.
@@ -179,6 +188,9 @@ Useful next commands:
 
 ```sh
 alab help
+alab feedback --kind suggestion --body "Describe a suggestion, question, or bug for the project owner."
+alab --key <root-key> dashboard --no-open
+alab --key <admin-or-root-key> report --project <project-id> --out ./alab-project-report.md
 alab observe experiments list
 alab observe runs list --exp <exp-id>
 alab observe experiments best
@@ -192,6 +204,9 @@ alab observe experiments best
 - **Experiment**: isolated Git branch/worktree bound to exactly one source and one config version.
 - **Run**: one evaluator execution for an experiment commit, with status, reward, logs, artifacts, and warning codes.
 - **Submit**: closes an experiment with final summary, feedback, final run, final commit, and explicit refs.
+- **Feedback**: HOME-level plaintext notes for suggestions, questions, bugs, or other agent observations, stored as one directory per entry under `feedback/`.
+- **Dashboard**: root-only local browser panel for read-only inspection. It binds to `127.0.0.1`, uses a random browser session token, serves bounded paginated read models for large homes, and writes no ALab records.
+- **Reports**: Markdown evidence exports for project-level owner handoffs or visible experiment contexts, without raw secrets, hidden log contents, or artifact bytes.
 - **Inspection checkout**: read-only checkout for observing/exporting scoped experiment evidence without becoming submit-capable.
 
 ## Configuration
@@ -209,7 +224,14 @@ See [docs/spec_runners_adapters.md](docs/spec_runners_adapters.md), [docs/spec_p
 
 ## Examples
 
-- [examples/skydiscover_circle_packing_codex](examples/skydiscover_circle_packing_codex/): complete SkyDiscover circle-packing example with setup, single Codex worker run, project-controller loop, report collection, and real-run notes.
+See [examples](examples/) for the runnable example matrix. The current examples
+cover a local scoring loop, a Dockerized clinic-order fulfillment planner with
+artifact export, a Harbor hidden-verifier incident classifier, a collaborative
+incident triage lifecycle workflow, a generated dashboard showcase home, and
+the SkyDiscover circle-packing Codex single-worker protocol. The same examples area also includes
+`examples/templates/`, a copyable multi-instance TSP template library for
+local, Docker, Harbor, SkyDiscover Python, and SkyDiscover Docker runner
+projects.
 
 The repository also includes Codex-facing role skills under [skills](skills/). They are external runbooks for operating ALab through the CLI as an experiment worker, project controller, or global admin; they do not add an embedded agent launcher to ALab.
 
@@ -226,10 +248,10 @@ Opt-in integration gates are excluded from the default suite:
 
 ```sh
 ALAB_RUN_REAL_DOCKER=1 UV_CACHE_DIR=.uv-cache UV_DEFAULT_INDEX=https://pypi.org/simple uv run pytest -m real_docker
-ALAB_RUN_REAL_SKYDISCOVER_PYTHON=1 UV_CACHE_DIR=.uv-cache uv run pytest -m real_skydiscover_python
+ALAB_RUN_REAL_SKYDISCOVER_PYTHON=1 UV_CACHE_DIR=.uv-cache UV_DEFAULT_INDEX=https://pypi.org/simple uv run pytest -m real_skydiscover_python
 ALAB_RUN_NETWORKED_SKYDISCOVER_PYTHON=1 UV_CACHE_DIR=.uv-cache UV_DEFAULT_INDEX=https://pypi.org/simple uv run pytest -m networked_skydiscover_python
 ALAB_RUN_NATIVE_SKYDISCOVER_PYTHON=1 UV_CACHE_DIR=.uv-cache UV_DEFAULT_INDEX=https://pypi.org/simple uv run pytest -m native_skydiscover_python
-ALAB_RUN_LIVE_SKYDISCOVER_CATALOG=1 UV_CACHE_DIR=.uv-cache uv run pytest -m live_skydiscover_catalog
+ALAB_RUN_LIVE_SKYDISCOVER_CATALOG=1 UV_CACHE_DIR=.uv-cache UV_DEFAULT_INDEX=https://pypi.org/simple uv run pytest -m live_skydiscover_catalog
 ```
 
 Notes:
@@ -237,7 +259,7 @@ Notes:
 - `uv.lock` is tracked because CI and local validation use `uv run --locked`.
 - Keep local cache/output paths ignored (`.uv-cache/`, `.pytest_cache/`, `.ruff_cache/`, `.alab-demo/`, `.env`).
 - GitHub Actions runs the default lint and pytest suite on pull requests and pushes to `main`; real Docker and live/networked SkyDiscover gates remain manual workflow inputs.
-- Pushes to `main` check PyPI for the current `pyproject.toml` package version; if that exact version is missing, CI builds and publishes through PyPI Trusted Publishing, otherwise it skips publishing.
+- Pushes to `main` check PyPI for the current `pyproject.toml` package version; if that exact version is missing, CI builds and publishes through PyPI Trusted Publishing, then creates a `v<version>` GitHub Release from the matching [CHANGELOG.md](CHANGELOG.md) section. If PyPI already has that version, publishing and release creation are skipped.
 - The PyPI `alab-cli` project must trust repository `bebetterest/ALab`, workflow `ci.yml`, and environment `pypi` before the first automated publish can succeed.
 
 ## Security And Data Model
@@ -249,6 +271,7 @@ ALab V1 is a local collaboration boundary, not a multi-user security product:
 - Credential verifiers are stored; raw credential secrets are not.
 - Project records are local plaintext SQLite/filesystem data.
 - `secret_env` values are local plaintext, redacted from rendered logs where configured, and never exported by config commands.
+- The dashboard is loopback-only and root-only. Its API can read hidden/full logs and artifact bytes for that root session, but it must not render raw keys, raw tokens, credential verifier material, or raw `secret_env` values.
 - Artifact exports are exact captured bytes and are not automatically redacted.
 
 ## Repository Structure
@@ -261,6 +284,8 @@ ALab V1 is a local collaboration boundary, not a multi-user security product:
 ├── docs/
 │   ├── assets/
 │   │   └── readme-header.png
+│   ├── README.md
+│   ├── README_cn.md
 │   ├── blueprint.md
 │   ├── blueprint_cn.md
 │   ├── completion_audit.md
@@ -270,7 +295,15 @@ ALab V1 is a local collaboration boundary, not a multi-user security product:
 │   ├── progress_pipeline.md
 │   └── progress_pipeline_cn.md
 ├── examples/
-│   └── skydiscover_circle_packing_codex/
+│   ├── README.md
+│   ├── README_cn.md
+│   ├── collaboration_observe_lifecycle/
+│   ├── dashboard_showcase/
+│   ├── docker_file_reward_artifacts/
+│   ├── harbor_verifier_minimal/
+│   ├── local_agent_scoreboard/
+│   ├── skydiscover_circle_packing_codex/
+│   └── templates/
 ├── skills/
 │   ├── alab-experiment-worker/
 │   ├── alab-project-controller/
@@ -285,6 +318,8 @@ ALab V1 is a local collaboration boundary, not a multi-user security product:
 │   └── test_runner_skydiscover.py
 ├── LICENSE
 ├── .env.example
+├── CHANGELOG.md
+├── CHANGELOG_cn.md
 ├── pyproject.toml
 ├── uv.lock
 ├── README.md
@@ -297,10 +332,12 @@ Local-only agent notes such as `AGENTS.md` and `CORE.md` are intentionally git-i
 
 - English documentation is canonical.
 - Synchronized Chinese documents use the `*_cn.md` naming pattern.
+- [docs/README.md](docs/README.md) explains the documentation structure and reading order.
 - [docs/blueprint.md](docs/blueprint.md) is the V1 product overview.
-- [docs/spec_cli.md](docs/spec_cli.md), [docs/spec_storage_auth_context.md](docs/spec_storage_auth_context.md), [docs/spec_project_source_experiment.md](docs/spec_project_source_experiment.md), [docs/spec_lifecycle.md](docs/spec_lifecycle.md), [docs/spec_runners_adapters.md](docs/spec_runners_adapters.md), [docs/spec_observe_collaboration.md](docs/spec_observe_collaboration.md), and [docs/spec_tests.md](docs/spec_tests.md) define subsystem contracts.
+- [docs/spec_cli.md](docs/spec_cli.md), [docs/spec_storage_auth_context.md](docs/spec_storage_auth_context.md), [docs/spec_project_source_experiment.md](docs/spec_project_source_experiment.md), [docs/spec_lifecycle.md](docs/spec_lifecycle.md), [docs/spec_runners_adapters.md](docs/spec_runners_adapters.md), [docs/spec_observe_collaboration.md](docs/spec_observe_collaboration.md), [docs/spec_dashboard.md](docs/spec_dashboard.md), and [docs/spec_tests.md](docs/spec_tests.md) define subsystem contracts.
 - [docs/progress.md](docs/progress.md), [docs/progress_pipeline.md](docs/progress_pipeline.md), [docs/progress_closed_gaps.md](docs/progress_closed_gaps.md), and [docs/progress_log.md](docs/progress_log.md) track current state, active queues, closed gaps, and history.
 - [docs/completion_audit.md](docs/completion_audit.md) tracks requirement-level evidence.
+- [CHANGELOG.md](CHANGELOG.md) records release-facing changes and is used by CI for GitHub Release notes.
 
 ## Contributing
 

{alab_cli-0.1.0 → alab_cli-0.1.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "alab-cli"
-version = "0.1.0"
+version = "0.1.2"
 description = "Local agent-first experiment workbench CLI"
 readme = "README.md"
 requires-python = ">=3.11"
@@ -22,6 +22,7 @@ dependencies = [
 Homepage = "https://github.com/bebetterest/ALab"
 Repository = "https://github.com/bebetterest/ALab"
 Issues = "https://github.com/bebetterest/ALab/issues"
+Changelog = "https://github.com/bebetterest/ALab/blob/main/CHANGELOG.md"
 
 [project.scripts]
 alab = "alab.cli:main"
@@ -36,7 +37,7 @@ dev = [
 package = true
 
 [tool.setuptools.package-data]
-alab = ["migrations/*.sql"]
+alab = ["migrations/*.sql", "dashboard_static/*", "dashboard_static/vendor/*"]
 
 [tool.pytest.ini_options]
 testpaths = ["tests"]

{alab_cli-0.1.0 → alab_cli-0.1.2}/src/alab/__init__.py

@@ -2,4 +2,4 @@
 
 __all__ = ["__version__"]
 
-__version__ = "0.1.0"
+__version__ = "0.1.2"

{alab_cli-0.1.0 → alab_cli-0.1.2}/src/alab/cli.py

@@ -16,7 +16,7 @@ from .errors import AlabError, error_exit_code
 from .home import resolve_home
 from .registry import COMMANDS, CommandSpec, match_command
 from .rendering import ResultBlock, error_block, render_text
-from .services import GlobalOptions, Request
+from .services import GlobalOptions, LongRunningResult, Request
 
 
 @dataclass
@@ -48,6 +48,7 @@ GLOBAL_PUBLIC: set[PathTuple] = {
     ("config", "validate"),
     ("context", "show"),
     ("context", "repair"),
+    ("feedback",),
 }
 
 GLOBAL_CONFIG_REPAIR: set[PathTuple] = {
@@ -56,6 +57,7 @@ GLOBAL_CONFIG_REPAIR: set[PathTuple] = {
     ("config", "set"),
     ("config", "reset"),
     ("config", "validate"),
+    ("feedback",),
 }
 
 PUBLIC_PROJECT: set[PathTuple] = {
@@ -86,6 +88,7 @@ OBSERVE_READ: set[PathTuple] = {
     ("exp", "search"),
     ("exp", "show"),
     ("exp", "best"),
+    ("report",),
     ("observe", "experiments", "list"),
     ("observe", "experiments", "search"),
     ("observe", "experiments", "show"),
@@ -586,7 +589,13 @@ def run(argv: list[str]) -> int:
         enforce_global_config_valid(spec.path, base_req)
         req = build_request(parsed)
         preflight(spec, req, rest)
-        blocks = spec.handler(rest, req)
+        result = spec.handler(rest, req)
+        if isinstance(result, LongRunningResult):
+            blocks = _with_context_token_warnings(req, result.blocks)
+            sys.stdout.write(render_text(blocks))
+            sys.stdout.flush()
+            return result.run()
+        blocks = result
         blocks = _with_context_token_warnings(req, blocks)
         sys.stdout.write(render_text(blocks))
         return infer_result_exit_code(blocks)