Promote dev to main: research map redesign, trial polish, issue fixes #127 #129 #132

.claude-plugin/marketplace.json

@@ -9,7 +9,7 @@
       "name": "argus",
       "source": "./plugins/argus",
       "description": "Persistent reviewed research and engineering runtime",
-      "version": "0.1.7",
+      "version": "0.1.8",
       "category": "Productivity"
     }
   ]

.github/workflows/extended.yml

@@ -1,8 +1,8 @@
 name: extended checks
 
 # On-demand only. These jobs were the second, third and fourth jobs of
-# tests.yml until 2026-09-14; the per-push signal is now the single Linux
-# full-suite job there. Run this workflow by hand before a release or when a
+# tests.yml until 2026-09-14, and the mypy type gate was its second step
+# until 2026-09-18; the per-push signal is now lint + tests in one Linux job. Run this workflow by hand before a release or when a
 # change touches the chat bridges, the pairing/QR path, the cockpits or the
 # Tauri desktop host.
 #
@@ -25,6 +25,11 @@ name: extended checks
 
 on:
   workflow_dispatch:
+    inputs:
+      typecheck_base:
+        description: Git revision the type gate compares against
+        default: origin/main
+        required: false
 
 concurrency:
   group: extended-${{ github.ref }}
@@ -128,6 +133,35 @@ env:
     tests/test_engineer_sandbox.py
 
 jobs:
+  typecheck:
+    name: linux · Python type gate
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+      - uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+          cache: pip
+      - name: Install
+        run: |
+          python -m pip install --upgrade pip
+          python -m pip install -e '.[dev,qr,trial]'
+      - name: Reject new Python type diagnostics
+        env:
+          TYPECHECK_BASE: ${{ inputs.typecheck_base || 'origin/main' }}
+        run: >-
+          python -m argus.release_tools.typecheck_gate
+          --base "$TYPECHECK_BASE" --output-dir "$RUNNER_TEMP/typecheck"
+      - name: Preserve Python type diagnostics
+        if: always()
+        uses: actions/upload-artifact@v4
+        with:
+          name: python-typecheck
+          path: ${{ runner.temp }}/typecheck
+          if-no-files-found: ignore
+
   portable:
     name: ${{ matrix.os }} · portable surface
     runs-on: ${{ matrix.os }}

.github/workflows/release.yml

@@ -3,15 +3,16 @@ name: release
 # Release version changes are operator-controlled. Do not bump package versions
 # for a bug fix unless the operator explicitly authorizes that release change.
 # v0.1.2 through v0.1.6 were operator-authorized desktop releases.
-# v0.1.7 is a Windows-only release: build it with workflow_dispatch and an
-# explicit windows-x86_64 target, verify its artifacts, then publish them.
-# Excluding its tag prevents a second all-platform/PyPI publication.
+# v0.1.7 and v0.1.8 are Windows-only releases: build with workflow_dispatch
+# and an explicit windows-x86_64 target, verify the artifacts, then publish.
+# Excluding these tags prevents a second all-platform publication.
 
 on:
   push:
     tags:
       - "v*"
       - "!v0.1.7"
+      - "!v0.1.8"
   workflow_dispatch:
     inputs:
       platform:
@@ -104,7 +105,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        include: ${{ fromJSON(inputs.platform == 'windows-x86_64' && '[{"os":"windows-latest","platform":"windows-x86_64"}]' || inputs.platform == 'darwin-aarch64' && '[{"os":"macos-15","platform":"darwin-aarch64"}]' || inputs.platform == 'darwin-x86_64' && '[{"os":"macos-15-intel","platform":"darwin-x86_64"}]' || inputs.platform == 'linux-x86_64' && '[{"os":"ubuntu-22.04","platform":"linux-x86_64"}]' || '[{"os":"windows-latest","platform":"windows-x86_64"},{"os":"macos-15","platform":"darwin-aarch64"},{"os":"macos-15-intel","platform":"darwin-x86_64"},{"os":"ubuntu-22.04","platform":"linux-x86_64"}]') }}
+        include: ${{ fromJSON(inputs.platform == 'windows-x86_64' && '[{"os":"windows-2022","platform":"windows-x86_64"}]' || inputs.platform == 'darwin-aarch64' && '[{"os":"macos-15","platform":"darwin-aarch64"}]' || inputs.platform == 'darwin-x86_64' && '[{"os":"macos-15-intel","platform":"darwin-x86_64"}]' || inputs.platform == 'linux-x86_64' && '[{"os":"ubuntu-22.04","platform":"linux-x86_64"}]' || '[{"os":"windows-2022","platform":"windows-x86_64"},{"os":"macos-15","platform":"darwin-aarch64"},{"os":"macos-15-intel","platform":"darwin-x86_64"},{"os":"ubuntu-22.04","platform":"linux-x86_64"}]') }}
     runs-on: ${{ matrix.os }}
     defaults:
       run:
@@ -113,18 +114,28 @@ jobs:
       - uses: actions/checkout@v4
       - uses: actions/setup-node@v4
         with:
-          node-version: '22.12.0'
+          node-version: ${{ runner.os == 'Windows' && '24.19.0' || '22.12.0' }}
           cache: npm
           cache-dependency-path: |
             frontend/web/package-lock.json
             frontend/tui/package-lock.json
             desktop-tauri/package-lock.json
       - uses: actions/setup-python@v5
         with:
-          python-version: '3.12'
+          python-version: ${{ runner.os == 'Windows' && '3.11.9' || '3.12' }}
           cache: pip
-          cache-dependency-path: pyproject.toml
+          cache-dependency-path: |
+            pyproject.toml
+            uv.lock
+            desktop-tauri/scripts/windows-build-requirements.txt
       - uses: dtolnay/rust-toolchain@stable
+        with:
+          toolchain: ${{ runner.os == 'Windows' && '1.98.0-x86_64-pc-windows-msvc' || 'stable' }}
+      - uses: astral-sh/setup-uv@v5
+        if: runner.os == 'Windows'
+        with:
+          version: '0.12.16'
+          enable-cache: false
       # The Tauri shell pulls in several hundred crates; compiling them cold is
       # the single longest step on every platform (10 min on the Intel Mac
       # runner). Keep the compiled dependencies between runs. Only main saves,
@@ -139,9 +150,23 @@ jobs:
         run: |
           sudo apt-get update
           sudo apt-get install -y libwebkit2gtk-4.1-dev build-essential libxdo-dev libssl-dev librsvg2-dev libayatana-appindicator3-dev patchelf xvfb xauth dbus-x11 libfuse2
-      - name: Install build dependencies
+      - name: Install frozen Windows build dependencies
+        if: runner.os == 'Windows'
+        env:
+          UV_PYTHON_DOWNLOADS: never
+        run: |
+          uv sync --frozen --extra dev --extra trial --no-install-project --python 3.11.9
+          uv pip install --python .venv/Scripts/python.exe --no-deps -r desktop-tauri/scripts/windows-build-requirements.txt
+          uv pip install --python .venv/Scripts/python.exe --no-deps --no-build-isolation --editable .
+          uv pip check --python .venv/Scripts/python.exe
+          echo "$GITHUB_WORKSPACE/.venv/Scripts" >> "$GITHUB_PATH"
+          echo 'RUSTUP_TOOLCHAIN=1.98.0-x86_64-pc-windows-msvc' >> "$GITHUB_ENV"
+          cargo fetch --locked --manifest-path desktop-tauri/src-tauri/Cargo.toml --target x86_64-pc-windows-msvc
+      - name: Install non-Windows Python build dependencies
+        if: runner.os != 'Windows'
+        run: python -m pip install -e ".[trial]" "pyinstaller>=6.11,<7" tzdata
+      - name: Install frontend build dependencies
         run: |
-          python -m pip install -e ".[trial]" "pyinstaller>=6.11,<7" tzdata
           npm --prefix frontend/web ci
           npm --prefix frontend/tui ci
           npm --prefix desktop-tauri ci
@@ -166,6 +191,16 @@ jobs:
         if: runner.os == 'Windows'
         shell: pwsh
         run: ./desktop-tauri/scripts/stage-release.ps1
+      - name: Verify reviewed Windows source and artifacts
+        if: runner.os == 'Windows'
+        run: |
+          python -m argus.release_tools.generate_manifest --check
+          python -m argus.release_tools.check_artifacts
+          git diff --exit-code HEAD
+          if [ -n "$(git ls-files --others --exclude-standard -- argus frontend)" ]; then
+            echo 'Unreviewed generated inputs remain; refusing to upload a release candidate.'
+            exit 1
+          fi
       - name: Stage Mac installer and updater
         if: runner.os == 'macOS'
         run: python desktop-tauri/scripts/stage-macos-release.py

.github/workflows/tests.yml

@@ -1,16 +1,13 @@
 name: tests
 
-# One check on every push and pull request: the Linux full suite (ruff, the
-# mypy type gate against the change's base, then the whole pytest suite).
-# Linux is the development platform and this job is the one that decides
-# whether a change is green.
+# One check on every push and pull request: lint, then the whole pytest suite
+# in parallel on Linux. Linux is the development platform and this job is the
+# one that decides whether a change is green.
 #
-# The platform-portability, frontend and Windows desktop-package checks that
-# used to run alongside it are kept verbatim in extended.yml and run only on
-# demand (workflow_dispatch) -- before a release, or when a change touches the
-# chat bridges, the pairing/QR path, the cockpits or the Tauri host. They were
-# moved out to keep the per-push signal to a single job and to stop paying
-# macOS/Windows runner minutes on every commit.
+# Everything else -- the mypy type gate, the platform-portability, frontend
+# and Windows desktop-package checks -- lives in extended.yml and runs only on
+# demand (workflow_dispatch): before a release, or when a change touches the
+# chat bridges, the pairing/QR path, the cockpits or the Tauri host.
 
 on:
   push:
@@ -24,12 +21,10 @@ concurrency:
 
 jobs:
   linux:
-    name: linux · full suite
+    name: linux · lint + tests
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
       - uses: actions/setup-python@v5
         with:
           python-version: '3.11'
@@ -44,18 +39,5 @@ jobs:
           python -m pip install -e '.[dev,qr,trial]'
       - name: Lint
         run: python -m ruff check argus argus_skill tests
-      - name: Reject new Python type diagnostics
-        env:
-          TYPECHECK_BASE: ${{ github.event.pull_request.base.sha || github.event.before }}
-        run: >-
-          python -m argus.release_tools.typecheck_gate
-          --base "$TYPECHECK_BASE" --output-dir "$RUNNER_TEMP/typecheck"
-      - name: Preserve Python type diagnostics
-        if: always()
-        uses: actions/upload-artifact@v4
-        with:
-          name: python-typecheck
-          path: ${{ runner.temp }}/typecheck
-          if-no-files-found: ignore
       - name: Test
-        run: python -m pytest -q
+        run: python -m pytest -q -n auto

README.zh-CN.md

@@ -423,6 +423,14 @@ argus --web
 
 首选地址：[http://127.0.0.1:8799](http://127.0.0.1:8799)；被占用时会自动顺延。
 
+研究地图区分任务历史与项目交付：审查器故障显示为“当次审查异常”，不代表科学否决。
+后续最终审稿通过后，地图提供审稿任务与项目成果入口，保留旧异常记录，不把旧任务改成成功。
+已通过审阅的中间任务仍计为完成，不因当时项目尚未结束而降为未完成。
+地图卡片的标题、摘要、状态和阶段信息使用固定排版，随画布一起缩放，不再按缩放比例自动折叠或替换文字；任务展开和历史切换仍保留。
+
+Reviewer 通过原生工具动作提交通过、返修、等待、操作员决策或计划调整，审稿正文自由表达，不解析 `STATUS`、JSON 或关键词，也不额外调用模型转译。
+等待后台实验不等于审查失败；最终论文仍须达到所选会议的验收标准。工具动作通道支持 Pi、Copilot、Codex 和 Claude/Qoder；不支持的后端明确报错，不回退到文本解析。
+
 ```bash
 argus --web --web-port 8800  # 使用其他端口
 ```

argus/__init__.py

@@ -61,7 +61,7 @@
     "SkillStore",
 ]
 
-__version__ = "0.1.7"
+__version__ = "0.1.8"
 
 _LAZY_LOOP_NAMES = frozenset({"SkillLoop", "SkillLoopConfig"})
 _LAZY_SKILL_NAMES = frozenset({"Skill", "SkillStore"})

argus/adapters/agent_cli_backend/_budget_monitor.py

@@ -95,8 +95,10 @@ def check(self) -> str | None:
             self._roll_day()
             observed, tokens = self.pi_cost, self.pi_tokens
             cursor = getattr(self.ctx, "copilot_usage_cursor", None)
-            if self.ctx.backend._is_copilot and cursor is not None and self.session_id:
-                usage = read_copilot_usage_since(cursor, session_id=self.session_id, timeout=0)
+            # A new session bound before spawn is readable before any event.
+            session_id = self.session_id or getattr(self.ctx, "provider_session_id", None)
+            if self.ctx.backend._is_copilot and cursor is not None and session_id:
+                usage = read_copilot_usage_since(cursor, session_id=session_id, timeout=0)
                 if usage is not None:
                     today = datetime.fromtimestamp(_local_day_start(time.time()), UTC)
                     for row in usage.rows:

argus/adapters/agent_cli_backend/_exec_context.py

@@ -67,3 +67,10 @@ class _ExecContext:
     event_permit: Any = field(default=None)
     copilot_token_billing_expected: bool = False
     copilot_usage_cursor: Any = None
+    # Provider-session identity Argus allocated for a NEW Copilot session and
+    # bound to this call before spawn; ``None`` for resumed calls, non-Copilot
+    # backends, the warm ACP path, or a CLI without ``--session-id``.
+    provider_session_id: str | None = None
+    # The CLI named a different session than the one this call was bound to.
+    # Usage lookups and the persisted record must not fall back to any identity.
+    session_identity_conflict: str | None = None

argus/adapters/agent_cli_backend/_exec_finalize.py

@@ -76,7 +76,10 @@ def finalize_result(
     result.call_id = ctx.call_id
     result.call_id_log_correlated = True
     result.stop_kind = normalize_stop_kind(result.stop_kind)
-    result.thread_id = result.thread_id or ctx.resume_thread_id
+    if ctx.session_identity_conflict:
+        result.thread_id = None
+    else:
+        result.thread_id = result.thread_id or ctx.resume_thread_id
     result.started_at = ctx.started_at
     result.completed_at = completed_at
     result.duration_ms = max(