wade-cli 0.48.1__tar.gz → 0.50.2__tar.gz

{wade_cli-0.48.1 → wade_cli-0.50.2}/.wade.yml

@@ -58,6 +58,22 @@ models:
       effort: xhigh
 provider:
   name: github
+bot_review:
+  # This repo runs CodeRabbit + Codex, not Cursor Bugbot. Bugbot is disabled so
+  # expectation-verified completion (#448) doesn't wait arrival_timeout for a
+  # review that never arrives on every review check. arrival_timeout/ack_timeout
+  # keep their defaults (300s / 900s).
+  auto_trigger: false
+  bots:
+  - name: coderabbit
+    trigger: '@coderabbitai review'
+    enabled: true
+  - name: codex
+    trigger: '@codex review'
+    enabled: true
+  - name: bugbot
+    trigger: bugbot run
+    enabled: false
 knowledge:
   enabled: true
   path: KNOWLEDGE.md

{wade_cli-0.48.1 → wade_cli-0.50.2}/CHANGELOG.md

@@ -4,6 +4,60 @@ All notable changes to this project will be documented in this file.
 
 The format is based on [Conventional Commits](https://conventionalcommits.org/).
 
+## [v0.50.2] — 2026-08-19
+
+### Bug Fixes
+
+- list uncommitted files before the PR-merge "work will be lost" prompt (#454) (932c645)
+
+## [v0.50.1] — 2026-08-18
+
+### Chores
+
+- bump crossby to v0.24.3 (#452) (403daa8)
+
+## [v0.50.0] — 2026-08-18
+
+### Features
+
+- harden and align review-pass budget and skip guidance across workflows (#451) (cef9d99)
+
+## [v0.49.1] — 2026-08-18
+
+### Bug Fixes
+
+- don't report review bots done until they've actually reviewed (#449) (01a3192)
+
+## [v0.49.0] — 2026-08-18
+
+### Features
+
+- standardize WADE session summaries with emoji status and decisive dialog prompts (#444) (4c4d730)
+
+## [v0.48.5] — 2026-08-18
+
+### Documentation
+
+- align installed agent guidance with the plan-driven lifecycle (#442) (d5dbcfc)
+
+## [v0.48.4] — 2026-08-18
+
+### Bug Fixes
+
+- wrap long choices in interactive pickers instead of cropping (#446) (6b162b6)
+
+## [v0.48.3] — 2026-08-18
+
+### Documentation
+
+- reconcile developer references with the current codebase (#440) (c20a564)
+
+## [v0.48.2] — 2026-08-17
+
+### Documentation
+
+- refresh public onboarding and product positioning (#438) (3f822e2)
+
 ## [v0.48.1] — 2026-08-17
 
 ### Documentation

{wade_cli-0.48.1 → wade_cli-0.50.2}/CONTRIBUTING.md

@@ -35,14 +35,15 @@ uv pip install -e ".[dev]"
 
 ```
 CLI Layer      →  services, models, config, logging, ui
-Service Layer  →  providers, crossby (AI tool adapters), git, db, models, config, logging
+Service Layer  →  providers, crossby (AI tool adapters), git, models, config, logging
 Provider Layer →  models, config, logging  (no service imports)
 Git Layer      →  models, config, logging  (no service imports)
-DB Layer       →  models, logging          (no config imports)
 Models Layer   →  nothing (leaf)
 ```
 
-AI tool adapters live in the external [`crossby`](https://github.com/ivanviragine/crossby) package, not this repo — see `docs/dev/architecture.md`.
+Never import a higher layer from a lower one, and keep business logic out of `cli/`. One sanctioned exception runs the other way: services may import the lower `skills/` utility layer for skill-file management (`install_skills`, the `*_SKILLS` registries, gitignore/cross-tool constants), never the reverse.
+
+AI tool adapters live in the external [`crossby`](https://github.com/ivanviragine/crossby) package, not this repo — wade's services import `crossby.ai_tools` and `crossby.models.ai` directly. See `docs/dev/architecture.md`.
 
 CLI modules are thin dispatch — parse flags with Typer, call service methods. Business logic lives in `services/`, not `cli/`.
 

{wade_cli-0.48.1 → wade_cli-0.50.2}/KNOWLEDGE.md

@@ -656,3 +656,24 @@ WADE owns zero AI-tool model catalogs — catalog models come from crossby's sta
 uv.lock is gitignored in the WADE repo (.gitignore:47) and NOT tracked. A dependency bump therefore commits only pyproject.toml (and the regenerated lock stays local for reproducibility). CI uses 'uv sync --all-extras' (not --locked/--frozen), so there is no lock-freshness gate. Do not 'git add -f uv.lock' during a dependency change.
 
 ---
+
+## 3f6332a4b61d | 2026-08-18 | implementation | tags: providers, deps, clickup | Issue #441
+
+WADE's dependency-graph and parent-linkage automation is numeric-ID only, so opaque-ID providers (ClickUp) cannot use it. Three places all assume digits: parse_deps_output's _ARROW_RE (deps_service.py, `^\d+ -> \d+`) rejects non-numeric edges; AbstractTaskProvider.find_parent_issue defaults to returning None (providers/base.py) and ClickUpProvider does NOT override it (GitHubProvider does, providers/github.py); and _CHECKLIST_ISSUE_RE / parse_all_issue_refs (models/task.py) match only `#<digits>`. Consequence: for ClickUp, `wade task deps` produces no accepted edges and an epic's `- [ ] #id` checklist is not recognized by done/batch parent flows — the epic is a human-readable tracking doc only. Documented in templates/skills/task/SKILL.md and deps/SKILL.md scope notes (PR #442).
+## 82ad4c583c58 | 2026-08-18 | implementation | tags: ui, prompts, questionary | Issue #445
+
+wade's interactive pickers all funnel through src/wade/ui/prompts.py (select/menu/multi_select -> questionary select/checkbox). questionary builds the choice list as a prompt_toolkit Window(InquirerControl) with the default wrap_lines=False, which CROPS long choices to the terminal width (silently losing text). To make choices wrap, call _enable_choice_wrapping(question) BETWEEN constructing the Question and calling .ask(): it walks question.application.layout.find_all_windows() and sets win.wrap_lines = to_filter(True) on each Window whose .content is a questionary.prompts.common.InquirerControl. Done best-effort (try/except -> logger.debug("prompts.choice_wrap_failed")) because it depends on undocumented questionary internals; pyproject pins questionary>=2.0 with no ceiling. wrap_lines reflows on resize and the highlight/[SetCursorPosition] carry across wrapped rows, so selection + arrow-key scroll still work. Known limit: no hanging indent (continuation rows start at col 0). confirm()/input_prompt() left unchanged (Yes/No + free text, no long choices). Pinned by tests/unit/test_utils/test_prompts.py::TestEnableChoiceWrapping.
+
+---
+
+## b7e8aa32094f | 2026-08-18 | implementation | tags: ui, prompts, rich | Issue #445
+
+The plain non-interactive Rich-printed list rows (task_service.list_tasks, implementation_service/cleanup.list_sessions/worktree list) already WRAP via Console.print defaults (no_wrap=False, overflow "fold") — verified rendering a long row at width 40. The long-choice CROP bug (#445) was ONLY ever in the questionary arrow-key picker (src/wade/ui/prompts.py), never in the Rich output path. So a "text is cropped" report about a list should be triaged to whether it came from the interactive picker vs the plain printed list.
+
+---
+
+## 3cc3b0e546d4 | 2026-08-18 | implementation | tags: review, review-polling, coderabbit, codex | Issue #448
+
+Review completion must be *expectation-verified*, not presence-inferred: absence of a blocking bot signal does NOT mean the review is done. WADE must verify each *enabled* bot (bot_review.bots) has posted a review covering HEAD before reporting all-clear. Generalizes cc91cd11 (a CodeRabbit COMPLETED marker != done posting). Implementation (#448): models/review.py holds a pure compute_bot_arrivals() returning per-bot ARRIVED/AWAITING/ACKNOWLEDGED/MISSING, plus bot_login_matches() (verified logins: coderabbitai[bot], chatgpt-codex-connector; bugbot is a best-effort cursor/bugbot guess). review_covers_latest_commit gates on blocking_bots when expected_bots is set. The service (annotate_bot_expectations) is the ONLY layer with config+clock, so it populates expected_bots and computes the arrival map (window measured from later of commit push and a .wade/bot-triggered marker; falls back to commit push). The arrival window (arrival_timeout=300s / ack_timeout=900s config) bounds the wait so an enabled-but-not-installed bot cannot hang the session — past its window it stops blocking but is reported distinctly (never swallowed into done). A bot's PR-level reaction (Codex posts THUMBS_UP/+1, NOT a comment-level eyes as first assumed) is an ack that extends its window to ack_timeout. Every completion surface must read the map: poll_for_reviews, start(), fetch_reviews(), format_review_status_summary.
+
+---

{wade_cli-0.48.1 → wade_cli-0.50.2}/KNOWLEDGE.ratings.jsonl

@@ -117,3 +117,10 @@
 {"dir": "up", "id": "cc91cd11", "ts": "2026-08-16T19:08:31.215429+00:00"}
 {"dir": "up", "id": "69f9e60c7c9f", "ts": "2026-08-17T19:57:30.614903+00:00"}
 {"dir": "up", "id": "69f9e60c7c9f", "ts": "2026-08-17T20:16:14.514374+00:00"}
+{"dir": "up", "id": "94ae0a8fd4c9", "ts": "2026-08-18T10:21:14.990160+00:00"}
+{"dir": "up", "id": "69f9e60c7c9f", "ts": "2026-08-18T10:21:15.166868+00:00"}
+{"dir": "up", "id": "702b274f", "ts": "2026-08-18T12:09:19.959012+00:00"}
+{"dir": "up", "id": "cc91cd11", "ts": "2026-08-18T12:50:49.552087+00:00"}
+{"dir": "up", "id": "3f6332a4b61d", "ts": "2026-08-18T13:33:29.057439+00:00"}
+{"id": "cc91cd11", "superseded_by": "3cc3b0e546d4", "ts": "2026-08-18T14:36:50.310773+00:00"}
+{"dir": "up", "id": "b6ca74e5", "ts": "2026-08-18T17:31:53.785409+00:00"}