Repository navigation
refactor: remove inert --checkpointing flag and enable_checkpointing option - #200
Conversation
…option Nothing read TaskOptions.enable_checkpointing and no checkpoint manager exists, yet the CLI, REST, MCP and docs advertised rollback. Remove the flag, the stored option, the API/MCP setters and echoes, and the docs/examples. Old state.json files that contain the key still load (pydantic ignores unknown fields); a test pins this. Dz-Task-Id: tsk_8429202d6299d5283bfe1faf989d3c63
✅ Reviewed — nothing blocking0 actionable comment(s) · grounded on your code ⏱ 27s wall clock · 🤖 developerz.ai — automated review, running on your model and your box. |
There was a problem hiding this comment.
Review summary — 23 file(s), 0 finding(s).
Completed review removes the inert checkpointing CLI, state, REST and MCP option and its documentation, with guards for legacy-state loading and CLI rejection. No merge-blocking defects found in the supplied diff.
- test files were not shown to the reviewer — this diff touched no path recognised as a test, so no assertion was available to read the change against
🤖 developerz.ai review — automated, running on your model and your box. What is this? · Reviewed by custom/gpt-6.1-sol on box hv-dedicated-6 — 809.2k in / 923 out over 6 calls, on your flat-rate plan: no per-token charge (~$1.63 at OpenRouter list rates).
There was a problem hiding this comment.
Review summary — 23 file(s), 0 finding(s).
completed Removed the inert checkpointing option across CLI, REST, MCP, persisted models and documentation. Legacy-state loading and key removal on save are covered by a regression test; no merge-blocking defects found in the supplied diff.
- test files were not shown to the reviewer — this diff touched no path recognised as a test, so no assertion was available to read the change against
🤖 developerz.ai review — automated, running on your model and your box. What is this? · Reviewed by custom/gpt-6.1-sol on box hv-dedicated-dev-din — 809.7k in / 880 out over 6 calls, on your flat-rate plan: no per-token charge (~$1.63 at OpenRouter list rates).
|
Resolve merge conflict on #200
Work landed on #200 — watching it toward merge. 20m · 13 tool calls · 2 messages Sessions: coding run · review 1 · review 2 · review 3 🤖 developerz.ai, an automated agent, not a human — hv-dedicated-dev-din (developerz-ai[bot]) · model |
|
File fence widened for a merge conflict resolution. The
Why: Resolved docs remove enable_checkpointing, but auto-merged tests/api/test_doc_field_lists.py still requires it in PATCH /config and GET /status (lines 425-445). The scoped docs test will fail until its obsolete assertions are removed. The platform granted it under the same area lock a new task takes (no other in-flight task holds these files), after checking the path is not a secret, CI configuration or a human-merge path. Posted automatically by the developerz bot (AI-authored). |
…e-the-inert-checkpointing-flag-and-its--8429202d # Conflicts: # docs/api-reference.md # docs/mcp-tools.md # src/claude_task_master/api/models_task.py # src/claude_task_master/cli_commands/workflow_start.py # src/claude_task_master/core/control_ops.py # src/claude_task_master/core/state.py # src/claude_task_master/mcp/server_specs.py # src/claude_task_master/mcp/tool_handlers_control.py
…e-the-inert-checkpointing-flag-and-its--8429202d
…204) ## What get_context_window (core/agent_models.py) had no caller anywhere in the repo; it was only re-exported from the core facade. Delete the helper, drop the core.__init__ import and __all__ entry, and reword the two comments that pointed at it (the MODEL_CONTEXT_WINDOWS comment and the ContextWindowsConfig docstring). MODEL_CONTEXT_WINDOWS and MODEL_CONTEXT_WINDOWS_STANDARD are retained: they are still exported and tested. CHANGELOG Unreleased records the removal of the public name. ## Acceptance - Repo-wide grep for get_context_window returns zero hits in src/, tests/, docs/ and CHANGELOG.md - python -c "import claude_task_master.core" succeeds and every name in core.__all__ resolves via hasattr - CHANGELOG.md [Unreleased] has a ### Removed section naming get_context_window - uv run pytest tests/core/test_agent_models.py passes unchanged (constants/facade tests green) - bin/check is green (full gate run by the harness after the team finishes) ## Changes - `src/claude_task_master/core/agent_models.py` (edit) — Delete the dead get_context_window function (lines 186-211). Reword the comment at lines 106-109 that introduces MODEL_CONTEXT_WINDOWS: drop the 'for get_context_window' framing and the sentence 'The helper can read config.json ...', keep the inert-at-runtime warning. KEEP MODEL_CONTEXT_WINDOWS, MODEL_CONTEXT_WINDOWS_STANDARD and DEFAULT_COMPACT_THRESHOLD_PERCENT — they are still facade-exported and tested. Do not touch other functions. - `src/claude_task_master/core/__init__.py` (edit) — Remove get_context_window from the agent_models import block (line 27) and from __all__ (line 222). Nothing else in the import list changes; grep proved no in-repo consumer of the facade export. - `src/claude_task_master/core/config.py` (edit) — Docstring-only edit in ContextWindowsConfig (lines 120-127): drop the clause 'and readable by ``get_context_window``' so the repo has no reference to the deleted name. The config model, its fields and CLAUDETM_CONTEXT_* env loading stay exactly as they are. - `CHANGELOG.md` (edit) — Under ## [Unreleased], add a '### Removed' section ABOVE the existing '### Fixed' (Keep a Changelog order) with one entry: removed the dead public helper get_context_window from claude_task_master.core / core.agent_models; it had no caller and its return value set no compaction threshold or SDK window; constants MODEL_CONTEXT_WINDOWS etc. are retained. Match the file's bold-lead entry style; no links added. - `README.md` (edit) — [docs] minor — README.md:346 says the inert context_windows values can still be "read by the context-window helper"; that helper is get_context_window, which this diff deletes (was core/agent_models.py:189-211), so the sentence was true before the diff and is false after it — the same clause was corrected in config.py:122-123 but README was missed. A reader who believes it goes hunting for a helper that no longer exists or assumes the values still have a runtime consumer. Fix: change "Values can still be loaded from configuration, emitted by profiles, and read by the context-window h [trimmed] ## Verification `./bin/check` passed (exit 0) in 9m 55s. What the editors ran to see the change working, and what they saw: - `uv run python -c "import claude_task_master.core as c; assert all(hasattr(c, n) for n in c.__all__)" && uv run pytest tests/core/test_agent_models.py -q` (core): Import and __all__ hasattr assert passed silently, then '64 passed in 1.64s' from tests/core/test_agent_models.py. - `grep 'get_context_window' over src/claude_task_master/core/` (core): Zero hits — no remaining references in agent_models.py, __init__.py or config.py. - `python verify_docs_links.py` (docs): exit 0 — 49 markdown files scanned, CHANGELOG.md extracted with 116 anchors, "All 299 links verified successfully!" - `code_changes on CHANGELOG.md` (docs): Uncommitted diff adds exactly the '### Removed' section (naming get_context_window, above '### Fixed') under '## [Unreleased]' — the section my brief required. - `repo-wide grep for get_context_window` (docs): Exactly one hit: CHANGELOG.md line 12 (the changelog entry itself, as the naming criterion requires); zero hits in src/, tests/, docs/ — the first pass's code deletion is confirmed in the tree. - `repo-wide grep for the retained names` (docs): MODEL_CONTEXT_WINDOWS and MODEL_CONTEXT_WINDOWS_STANDARD at agent_models.py:109/118 plus facade exports in core/__init__.py, DEFAULT_COMPACT_THRESHOLD_PERCENT at agent_models.py:186, ContextWindowsConfig at config.py:119 — the invariant my entry asserts holds. - `python verify_docs_links.py (my scoped check)` (docs): Exit 0: 'All 299 links verified successfully!' across 49 markdown files, CHANGELOG.md included (116 anchors, no broken links). - `python verify_docs_links.py` (docs): exit 0 — 'All 299 links verified successfully!' over 49 markdown files, README.md included - `repo-wide grep for get_context_window in src/ and for 'context-window helper' in README.md` (docs): zero hits for the helper name in src/ (deletion already in tree); README.md:346 was the only line naming the 'context-window helper' and now no longer does ## Notes Handoffs this run could not complete: - `CHANGELOG.md` — Deleted the dead get_context_window helper (agent_models.py), its facade export (core/__init__.py import + __all__), and its last docstring mention (config.py ContextWindowsConfig).; The acceptance requires CHANGELOG.md [Unreleased] to gain a '### Removed' section naming get_context_window (Keep a Changelog 1.1.0: '### Removed' sits above '### Fixed' inside '## [Unreleased]'). CHANGELOG.md is not in my write set, so I did not touch it. (the editor sent to make it neither wrote it nor said why not) Part 2/4 of goal `rtk_9f27cc6e57494e1aa71249859e371c59` - #200 - developerz-ai/claude-task-master (no PR yet) - #195 <!-- dz-task:tsk_930a71a5c4d33b3f04b6f3e58161aa0c --> --- 🤖 Opened by **developerz.ai** for task `tsk_930a71a5c4d33b3f04b6f3e58161aa0c` on developerz-ai/claude-task-master. Raised automatically by the `raw_task_triage` lane. <!-- dz-stack:start --> > 🤖 developerz.ai, an automated agent, not a human — this PR is one slice of a stacked plan. > Dispatched automatically by the `raw_task_triage` lane > Model `glm-5.3-flash` on `api.z.ai` via `native`, funded by `account_key` > Ran on box `hv-dedicated-dev-din` > Cost ~$0.10 for 788.6k in / 32.7k out (estimate: list price x reported tokens, cached input at its cached rate) > Run timeline [`run_c1593d007aea4410a1150a39db663a96`](https://app.developerz.ai/fleet/runs/run_c1593d007aea4410a1150a39db663a96) > Part 2/4 of plan [`rtk_9f27cc6e57494e1aa71249859e371c59`](https://app.developerz.ai/fleet/plans/rtk_9f27cc6e57494e1aa71249859e371c59) <!-- dz-stack:end --> Co-authored-by: developerz-ai[bot] <286432749+developerz-ai[bot]@users.noreply.github.com>
There was a problem hiding this comment.
Review summary — 9 file(s), 0 finding(s).
Completed review removes checkpointing from CLI, REST/MCP configuration surfaces and related documentation guards. No concrete defects found in the supplied diff; reported head CI checks are green.
- test files were not shown to the reviewer — this diff touched no path recognised as a test, so no assertion was available to read the change against
🤖 developerz.ai review — automated, running on your model and your box. What is this? · Reviewed by custom/gpt-6.1-sol on box hv-dedicated-dev-din — 103.2k in / 213 out over 2 calls, on your flat-rate plan: no per-token charge (~$0.21 at OpenRouter list rates).
## What The GraphQL review-thread formatter _format_pr_comments had no call site anywhere in the codebase or tests. Remove its definition (client_pr_helpers.py:408-434), its __all__ entry (client_pr_helpers.py:477), and the backward-compat re-export import and __all__ entry in client_pr.py (lines 17 and 317) as one atomic change — ruff selects pyflakes (F822), so the definition and the exports must go together. The five live siblings (_get_comment_resolved_map, _build_pr_status_query, _parse_pr_status_response, _parse_check_contexts, _format_pr_comments_from_rest) are unchanged. CHANGELOG Unreleased gains a Removed note. ## Acceptance - Repo-wide grep for _format_pr_comments matches only historical CHANGELOG prose and the cli.py comment naming _format_pr_comments_from_rest — no definition, __all__ entry, or import of _format_pr_comments remains in src/ - _format_pr_comments_from_rest, _get_comment_resolved_map, _build_pr_status_query, _parse_pr_status_response and _parse_check_contexts remain defined in client_pr_helpers.py and listed in both module __all__ lists (client_pr_helpers.py:472-478 → 5 entries, client_pr.py:309-319 → 8 entries) - uv run ruff check src/claude_task_master/github/client_pr.py src/claude_task_master/github/client_pr_helpers.py is clean (F/F822 selected via pyproject.toml:115) - CHANGELOG.md Unreleased has a Removed note for _format_pr_comments and all historical entries are byte-identical - bin/check is green (harness full gate) ## Changes - `src/claude_task_master/github/client_pr_helpers.py` (edit) — Delete the dead GraphQL formatter _format_pr_comments (lines 408-434, keeping exactly two blank lines before _format_pr_comments_from_rest at 437) and drop "_format_pr_comments", from __all__ (line 477). The five sibling helpers (lines 13, 97, 226, 370, 437) are untouched. - `src/claude_task_master/github/client_pr.py` (edit) — Remove _format_pr_comments, from the client_pr_helpers import block (line 17) and "_format_pr_comments", from __all__ (line 317). Keep the block's `# noqa: F401 — re-exported for backward compat` comment and all five sibling names; the live calls at lines 118, 149, 303, 306 are untouched. - `CHANGELOG.md` (edit) — Add one bullet under the existing Unreleased '### Removed' section (after the get_context_window entry, line 12): the dead _format_pr_comments GraphQL formatter (github/client_pr_helpers.py) was removed — no call site anywhere; removed together with both __all__ entries and the client_pr backward-compat re-export. Historical release notes below are NOT edited. ## Verification `./bin/check` passed (exit 0) in 3m 19s. What the editors ran to see the change working, and what they saw: - `uv run ruff check src/claude_task_master/github/client_pr.py src/claude_task_master/github/client_pr_helpers.py` (github): EXIT=0 — All checks passed! (F selected via pyproject.toml:115, so F822 would fail on any leftover __all__ entry/import of the deleted name; proves the three removals landed atomically) - `grep _format_pr_comments over src/claude_task_master/github/client_pr*.py` (github): only _format_pr_comments_from_rest remains: client_pr.py:17 (import), :305 (live call), :316 (__all__); client_pr_helpers.py:408 (def), :448 (__all__) — no bare _format_pr_comments token left Part 3/4 of goal `rtk_9f27cc6e57494e1aa71249859e371c59` - #200 - #204 - #195 <!-- dz-task:tsk_a2b402fa733df2802883bdbd46e33a72 --> --- 🤖 Opened by **developerz.ai** for task `tsk_a2b402fa733df2802883bdbd46e33a72` on developerz-ai/claude-task-master. Raised automatically by the `raw_task_triage` lane. <!-- dz-stack:start --> > 🤖 developerz.ai, an automated agent, not a human — this PR is one slice of a stacked plan. > Dispatched automatically by the `raw_task_triage` lane > Model `glm-5.3-flash` on `api.z.ai` via `native`, funded by `account_key` > Ran on box `hv-dedicated-reviewer-1` > Cost ~$0.04 for 266.8k in / 21.5k out (estimate: list price x reported tokens, cached input at its cached rate) > Run timeline [`run_15db90b89057460cb959e5caa13dfb2c`](https://app.developerz.ai/fleet/runs/run_15db90b89057460cb959e5caa13dfb2c) > Part 3/4 of plan [`rtk_9f27cc6e57494e1aa71249859e371c59`](https://app.developerz.ai/fleet/plans/rtk_9f27cc6e57494e1aa71249859e371c59) <!-- dz-stack:end --> Co-authored-by: developerz-ai[bot] <286432749+developerz-ai[bot]@users.noreply.github.com>
…e-the-inert-checkpointing-flag-and-its--8429202d # Conflicts: # CHANGELOG.md
…e-the-inert-checkpointing-flag-and-its--8429202d
There was a problem hiding this comment.
Review summary — 1 file(s), 0 finding(s).
completed — The diff adds a changelog entry announcing checkpointing removal and legacy-state compatibility. No concrete defect is established by the supplied code; the implementation and compatibility test are outside this diff and remain unverified.
- test files were not shown to the reviewer — this diff touched no path recognised as a test, so no assertion was available to read the change against
🤖 developerz.ai review — automated, running on your model and your box. What is this? · Reviewed by custom/gpt-6.1-sol on box hv-dedicated-reviewer-1 — 17.9k in / 131 out over 1 call, on your flat-rate plan: no per-token charge (~$0.04 at OpenRouter list rates).
What
Nothing read TaskOptions.enable_checkpointing and no checkpoint manager exists, yet the CLI, REST, MCP and docs advertised rollback. Remove the flag, the stored option, the API/MCP setters and echoes, and the docs/examples. Old state.json files that contain the key still load (pydantic ignores unknown fields); a test pins this.
Acceptance
claudetm start --helpno longer lists --checkpointing and no help text, docs or example presents checkpointing as a featureChanges
src/claude_task_master/core/state_models.py(edit) — Deleteenable_checkpointing: bool = Falsefrom TaskOptions (line 41). Pydantic default extra='ignore' keeps old state.json loading; do NOT add extra='forbid'.src/claude_task_master/core/state.py(edit) — Removeenable_checkpointingline from update_options docstring (line ~416).src/claude_task_master/core/control_ops.py(edit) — Removeenable_checkpointingline from update_config docstring (line ~248).src/claude_task_master/cli_commands/workflow_start.py(edit) — Remove theenable_checkpointingtyper.Option (lines 99-103,--checkpointing) and theenable_checkpointing=enable_checkpointingkwarg to TaskOptions (line 256). Check for any docstring/example mention.src/claude_task_master/api/models_task.py(edit) — Remove enable_checkpointing from the config-update request model (docstring line 117, Field ~156-159) and from the response options model (docstring line 291, field line 304).src/claude_task_master/api/routes_info.py(edit) — Removeenable_checkpointing=state.options.enable_checkpointing,(line 164) from the options response construction.src/claude_task_master/api/routes_task.py(edit) — Removeenable_checkpointing=False, # Default to False(line 115) from TaskOptions construction.src/claude_task_master/mcp/server_specs.py(edit) — Remove the enable_checkpointing docstring line (line 195) in the update-config tool spec string.src/claude_task_master/mcp/tool_handlers_control.py(edit) — Remove the enable_checkpointing parameter (line 188), docstring line (210) and theif enable_checkpointing is not None: kwargs[...]block (237-238) from the update_config handler. Check server_specs/other callers pass no such kwarg.CHANGELOG.md(edit) — Under## [Unreleased](line 8), add a### Removedsection noting the inertclaudetm start --checkpointingflag, TaskOptions.enable_checkpointing, and the REST/MCP config field were removed (nothing read them; no checkpoint manager existed); old state.json files containing the key still load (ignored). Place it before/after### Fixed.docs/api-reference.md(edit) — Remove"enable_checkpointing": false,from the status example JSON (line 154) and theenable_checkpointingbullet (line 419).docs/mcp-tools.md(edit) — Remove theenable_checkpointingparameter table row (line 443).examples/README.md(edit) — Remove the whole## Checkpointing (Experimental)section (lines 105-111) that advertises --checkpointing rollback; keep surrounding sections' spacing; make sure no TOC link points to it.examples/08-advanced-workflows.md(edit) — Remove section### 5. Checkpointing for Safety(lines 337-344) including the code block; renumber later sibling headings if any and fix any TOC/link references.tests/core/test_state.py(edit) — Removeenable_checkpointingfrom the model_dump expected dict (line 89) and the state fixture (line ~2463). ADD a test (failure-first style) that a state.json whoseoptionscontains"enable_checkpointing": truestill loads via StateManager.load_state, the loaded options has no such attribute, and a save/reload round-trip writes noenable_checkpointingkey.tests/cli_commands/test_workflow.py(edit) — Change_UNDOCUMENTED_START_FLAGSto{"--help"}and drop the stale 'being removed' comment (lines 16-17). Optionally add a test thatstart --helpno longer lists --checkpointing and that passing it fails with a usage error.tests/cli/test_mailbox_commands.py(edit) — Removeenable_checkpointing=False,(line 61) from the TaskOptions construction.tests/cli_commands/test_control.py(edit) — Removeenable_checkpointing=False,(line 46) from the TaskOptions construction.tests/api/test_routes_config.py(edit) — Remove every"enable_checkpointing"entry from fixtures (lines 38,97,161,201,234,248,289) and the two assertions (75, 265).tests/api/test_routes_control.py(edit) — Remove every"enable_checkpointing": False,fixture line (about 20 occurrences, lines 35..791).tests/api/test_routes_status.py(edit) — Remove every"enable_checkpointing": False,fixture line (about 25 occurrences) and the assertion at line 58.tests/api/test_auth.py(edit) — Remove"enable_checkpointing": False,at lines 98 and 173.tests/integration/test_api_cli_integration.py(edit) — Remove"enable_checkpointing": False,at lines 48 and 104.Verification
./bin/checkpassed (exit 0) in 2m 36s.What the editors ran to see the change working, and what they saw:
uv run pytest -p no:cacheprovider -n0 -q on the 5 owned test files(tests-api): 119 passed in 11.27s; grep for 'checkpoint' in those files returns nothinguv run pytest -p no:cacheprovider -n0 -q tests/core/test_state.py tests/cli_commands/test_workflow.py tests/cli_commands/test_control.py tests/cli/test_mailbox_commands.py(tests-core): 4 failed, 291 passed. The 4 failures are the new or updated guards: test_model_dump, test_load_state_with_removed_enable_checkpointing_key, test_every_start_flag_is_in_readme_table ("README 'Start Options' table is missing: --checkpointing") and test_checkpointing_flag_removed. They fail because src/ still defines the flag when I ran: workflow_start.py:99 and state_models.py:41. The source owner's edits had not landed yet. They should pass once those are removed. I did not re-run after that.uv run ruff check <scoped src files + src/claude_task_master/api>(src): All checks passed!grep -rn checkpoint src examples docs/*.md (excluding CHANGELOG)(src): No text hits in source or docs; only stale .pyc binary matches.uv run pytest -p no:cacheprovider -n0 tests/core/test_state.py tests/cli_commands/test_workflow.py tests/cli_commands/test_control.py tests/cli/test_mailbox_commands.py(tests-core): 295 passed in 7.11s, EXIT=0Notes
Premises of the request the team disproved:
Handoffs this run could not complete:
tests/core/test_state.py— TaskOptions.enable_checkpointing removed; the file still mentions checkpointing; Must add or keep a test that loads a state.json whose options contain enable_checkpointing and loads cleanly, with a round-trip save omitting the key. Drop any assertions on the field. (the editor sent to make it neither wrote it nor said why not)tests/cli_commands/test_workflow.py— The --checkpointing option is gone fromstart; Lines ~16-17 exempt --checkpointing from the README-table test; remove that exemption. Also drop any test that passes --checkpointing or reads enable_checkpointing. (the editor sent to make it neither wrote it nor said why not)Part 1/4 of goal
rtk_9f27cc6e57494e1aa71249859e371c59🤖 Opened by developerz.ai for task
tsk_8429202d6299d5283bfe1faf989d3c63on developerz-ai/claude-task-master.Raised automatically by the
raw_task_triagelane.