threadkeeper 0.12.0__tar.gz → 0.13.0__tar.gz

{threadkeeper-0.12.0 → threadkeeper-0.13.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: threadkeeper
-Version: 0.12.0
+Version: 0.13.0
 Summary: Multi-agent shared brain across Claude Code/Desktop, Codex, Antigravity CLI, Gemini, Copilot, VS Code. Cross-session memory, self-improving skill loops, inter-agent signaling — one local MCP server.
 Author: thread-keeper contributors
 License: MIT
@@ -221,7 +221,7 @@ tk-agent-status --cleanup-memory
 ```
 
 `apps/macos-agent-status/` contains a small macOS menu-bar app that polls this
-command every 5 seconds and shows every autonomous learning loop: enabled/off,
+command every 15 seconds and shows every autonomous learning loop: enabled/off,
 running/idle/ready, last pass, backlog, and active child RSS when that loop has
 spawned a worker. PyPI wheels and sdists also bundle the same Swift source under
 `threadkeeper/assets/macos-agent-status/`, so a normal `pipx`/`uv tool` install
@@ -239,12 +239,16 @@ memory button, self-restarts when its own RSS crosses
 notification permission, and sends a notification when a newly completed
 autonomous child task produces a useful result in `recent_results`; the first
 poll only marks existing results as seen, so old completions do not spam
-notifications. The header gear opens a separate Settings window for
+notifications. Status polling and cleanup commands run off the main actor, so
+opening the popover does not wait for `tk-agent-status --json`. The header gear
+opens a separate Settings window for
 `~/.threadkeeper/.env`: common knobs are grouped into guided controls, the raw
 `.env` remains editable for advanced values, three local presets can be saved
 and loaded, and Save & Restart writes the file then asks existing
 `threadkeeper.server` processes to exit so MCP hosts reconnect with the new
-configuration. Probe backlog is due objective
+configuration. Spawn CLI selectors collapse `agy` into canonical `antigravity`
+while keeping `gemini` as legacy, and model selectors use dropdowns with exact
+CLI model ids/labels instead of free-text fields. Probe backlog is due objective
 probes only, not every registered probe, so a healthy cooldown shows `0 due
 probes` instead of looking stuck. On macOS, `python -m threadkeeper.server`
 automatically installs and launches it on MCP startup, and restarts the app when
@@ -633,11 +637,15 @@ keys are lowercased:
 # default agent for roles with no explicit pin ("" / unset = use the active CLI)
 THREADKEEPER_SPAWN__DEFAULT=claude
 # per-role CLI:  THREADKEEPER_SPAWN__LOOP__<ROLE>=<cli>
+# supported CLI keys: claude, codex, antigravity (agy executable), gemini (legacy), copilot
 THREADKEEPER_SPAWN__LOOP__SHADOW_OBSERVER=claude   # heaviest reasoning → keep on Claude
 THREADKEEPER_SPAWN__LOOP__CURATOR=codex            # weekly audit → Codex is fine
 THREADKEEPER_SPAWN__LOOP__CANDIDATE_REVIEWER=auto  # "auto" = follow active CLI
 # model pin per CLI or per role:  THREADKEEPER_SPAWN__MODEL__<KEY>=<model>
 THREADKEEPER_SPAWN__MODEL__CLAUDE=opus
+THREADKEEPER_SPAWN__MODEL__CODEX=gpt-5.5
+THREADKEEPER_SPAWN__MODEL__AGY="Gemini 3.1 Pro (High)"
+THREADKEEPER_SPAWN__MODEL__GEMINI=gemini-3.1-pro-preview
 THREADKEEPER_SPAWN__MODEL__DIALECTIC_VALIDATOR=opus
 ```
 
@@ -645,7 +653,9 @@ Resolution per role: `SPAWN__LOOP__<role>` → `SPAWN__DEFAULT` → active CLI
 `claude`; `"auto"` (or unset) defers to the active CLI. Real environment
 variables override the `.env`. Force host detection with
 `THREADKEEPER_ACTIVE_CLI=claude` (or `codex`, `antigravity`/`agy`,
-`gemini`, `copilot`). See `.env.example` for the full knob list.
+`gemini`, `copilot`). `agy` is normalized to `antigravity`; `gemini` remains a
+legacy Gemini CLI adapter for old installs/enterprise paths. See `.env.example`
+for the full knob list.
 
 Adapters without headless support (Claude Desktop, VS Code) can't be
 spawn targets — `spawn_status()` reports them as "no adapter" and any
@@ -745,12 +755,34 @@ unchanged.
 ## Verifying ingest across CLIs
 
 ```bash
-python scripts/tk_verify_ingest.py
+python scripts/tk_verify_ingest.py            # both checks below
+python scripts/tk_verify_ingest.py --contract # parse/ingest contract only
+python scripts/tk_verify_ingest.py --live      # production verdict only
+python scripts/tk_verify_ingest.py --live --json   # machine-readable
 ```
 
-Walks every installed CLI adapter, parses recent transcripts in an
-isolated tempdir DB, reports per-source message counts and any silent
-parse failures. Read-only with respect to live state.
+Two read-only checks:
+
+- **Contract test** (`--contract`) — walks every installed CLI adapter,
+  parses recent transcripts into an isolated tempdir DB, reports
+  per-source message counts and flags any adapter that parsed messages
+  but silently failed to persist them. Answers *"does the pipeline
+  work?"*
+- **Production verification** (`--live`) — reads the **live**
+  `dialog_messages` table read-only and scores the three acceptance
+  criteria from [roadmap issue #1](https://github.com/po4erk91/thread-keeper/issues/1):
+  (1) every targeted CLI *slot* has production rows, (2) shadow-review
+  sees more than one adapter in the same recent window, (3) the learning
+  loop has fired on non-Claude sessions. Emits a `PASS` / `PARTIAL` /
+  `FAIL` verdict. The four slots are `claude-code`, `codex`, `copilot`,
+  and `google` — where the Google slot is satisfied by *either* the
+  legacy `gemini` adapter or its successor Antigravity (`agy`), since
+  both live under `~/.gemini`.
+
+`--strict` makes the process exit non-zero unless the live verdict is
+`PASS`, so it can gate CI; `PARTIAL` (e.g. a box that doesn't run all
+four CLIs) is a valid real-world state and exits 0 by default. The
+reusable verdict logic lives in `threadkeeper/verify_ingest.py`.
 
 ---
 
@@ -776,6 +808,7 @@ threadkeeper/
 ├── db.py                 # SQLite schema + sqlite-vec loader
 ├── identity.py           # session, self-cid, daemon launchers
 ├── ingest.py             # adapter-driven transcript ingest
+├── verify_ingest.py      # cross-CLI production verification verdict
 ├── brief.py              # render_brief / render_context
 ├── shadow_review.py      # autonomous learning observer
 ├── i18n.py               # 10 locales of regex + prompt bundles
@@ -814,3 +847,5 @@ locale. Look for the `good-first-issue` label.
 ## License
 
 MIT — see [LICENSE](LICENSE).
+
+<!-- mcp-name: io.github.po4erk91/thread-keeper -->

{threadkeeper-0.12.0 → threadkeeper-0.13.0}/README.md

@@ -180,7 +180,7 @@ tk-agent-status --cleanup-memory
 ```
 
 `apps/macos-agent-status/` contains a small macOS menu-bar app that polls this
-command every 5 seconds and shows every autonomous learning loop: enabled/off,
+command every 15 seconds and shows every autonomous learning loop: enabled/off,
 running/idle/ready, last pass, backlog, and active child RSS when that loop has
 spawned a worker. PyPI wheels and sdists also bundle the same Swift source under
 `threadkeeper/assets/macos-agent-status/`, so a normal `pipx`/`uv tool` install
@@ -198,12 +198,16 @@ memory button, self-restarts when its own RSS crosses
 notification permission, and sends a notification when a newly completed
 autonomous child task produces a useful result in `recent_results`; the first
 poll only marks existing results as seen, so old completions do not spam
-notifications. The header gear opens a separate Settings window for
+notifications. Status polling and cleanup commands run off the main actor, so
+opening the popover does not wait for `tk-agent-status --json`. The header gear
+opens a separate Settings window for
 `~/.threadkeeper/.env`: common knobs are grouped into guided controls, the raw
 `.env` remains editable for advanced values, three local presets can be saved
 and loaded, and Save & Restart writes the file then asks existing
 `threadkeeper.server` processes to exit so MCP hosts reconnect with the new
-configuration. Probe backlog is due objective
+configuration. Spawn CLI selectors collapse `agy` into canonical `antigravity`
+while keeping `gemini` as legacy, and model selectors use dropdowns with exact
+CLI model ids/labels instead of free-text fields. Probe backlog is due objective
 probes only, not every registered probe, so a healthy cooldown shows `0 due
 probes` instead of looking stuck. On macOS, `python -m threadkeeper.server`
 automatically installs and launches it on MCP startup, and restarts the app when
@@ -592,11 +596,15 @@ keys are lowercased:
 # default agent for roles with no explicit pin ("" / unset = use the active CLI)
 THREADKEEPER_SPAWN__DEFAULT=claude
 # per-role CLI:  THREADKEEPER_SPAWN__LOOP__<ROLE>=<cli>
+# supported CLI keys: claude, codex, antigravity (agy executable), gemini (legacy), copilot
 THREADKEEPER_SPAWN__LOOP__SHADOW_OBSERVER=claude   # heaviest reasoning → keep on Claude
 THREADKEEPER_SPAWN__LOOP__CURATOR=codex            # weekly audit → Codex is fine
 THREADKEEPER_SPAWN__LOOP__CANDIDATE_REVIEWER=auto  # "auto" = follow active CLI
 # model pin per CLI or per role:  THREADKEEPER_SPAWN__MODEL__<KEY>=<model>
 THREADKEEPER_SPAWN__MODEL__CLAUDE=opus
+THREADKEEPER_SPAWN__MODEL__CODEX=gpt-5.5
+THREADKEEPER_SPAWN__MODEL__AGY="Gemini 3.1 Pro (High)"
+THREADKEEPER_SPAWN__MODEL__GEMINI=gemini-3.1-pro-preview
 THREADKEEPER_SPAWN__MODEL__DIALECTIC_VALIDATOR=opus
 ```
 
@@ -604,7 +612,9 @@ Resolution per role: `SPAWN__LOOP__<role>` → `SPAWN__DEFAULT` → active CLI
 `claude`; `"auto"` (or unset) defers to the active CLI. Real environment
 variables override the `.env`. Force host detection with
 `THREADKEEPER_ACTIVE_CLI=claude` (or `codex`, `antigravity`/`agy`,
-`gemini`, `copilot`). See `.env.example` for the full knob list.
+`gemini`, `copilot`). `agy` is normalized to `antigravity`; `gemini` remains a
+legacy Gemini CLI adapter for old installs/enterprise paths. See `.env.example`
+for the full knob list.
 
 Adapters without headless support (Claude Desktop, VS Code) can't be
 spawn targets — `spawn_status()` reports them as "no adapter" and any
@@ -704,12 +714,34 @@ unchanged.
 ## Verifying ingest across CLIs
 
 ```bash
-python scripts/tk_verify_ingest.py
+python scripts/tk_verify_ingest.py            # both checks below
+python scripts/tk_verify_ingest.py --contract # parse/ingest contract only
+python scripts/tk_verify_ingest.py --live      # production verdict only
+python scripts/tk_verify_ingest.py --live --json   # machine-readable
 ```
 
-Walks every installed CLI adapter, parses recent transcripts in an
-isolated tempdir DB, reports per-source message counts and any silent
-parse failures. Read-only with respect to live state.
+Two read-only checks:
+
+- **Contract test** (`--contract`) — walks every installed CLI adapter,
+  parses recent transcripts into an isolated tempdir DB, reports
+  per-source message counts and flags any adapter that parsed messages
+  but silently failed to persist them. Answers *"does the pipeline
+  work?"*
+- **Production verification** (`--live`) — reads the **live**
+  `dialog_messages` table read-only and scores the three acceptance
+  criteria from [roadmap issue #1](https://github.com/po4erk91/thread-keeper/issues/1):
+  (1) every targeted CLI *slot* has production rows, (2) shadow-review
+  sees more than one adapter in the same recent window, (3) the learning
+  loop has fired on non-Claude sessions. Emits a `PASS` / `PARTIAL` /
+  `FAIL` verdict. The four slots are `claude-code`, `codex`, `copilot`,
+  and `google` — where the Google slot is satisfied by *either* the
+  legacy `gemini` adapter or its successor Antigravity (`agy`), since
+  both live under `~/.gemini`.
+
+`--strict` makes the process exit non-zero unless the live verdict is
+`PASS`, so it can gate CI; `PARTIAL` (e.g. a box that doesn't run all
+four CLIs) is a valid real-world state and exits 0 by default. The
+reusable verdict logic lives in `threadkeeper/verify_ingest.py`.
 
 ---
 
@@ -735,6 +767,7 @@ threadkeeper/
 ├── db.py                 # SQLite schema + sqlite-vec loader
 ├── identity.py           # session, self-cid, daemon launchers
 ├── ingest.py             # adapter-driven transcript ingest
+├── verify_ingest.py      # cross-CLI production verification verdict
 ├── brief.py              # render_brief / render_context
 ├── shadow_review.py      # autonomous learning observer
 ├── i18n.py               # 10 locales of regex + prompt bundles
@@ -773,3 +806,5 @@ locale. Look for the `good-first-issue` label.
 ## License
 
 MIT — see [LICENSE](LICENSE).
+
+<!-- mcp-name: io.github.po4erk91/thread-keeper -->

{threadkeeper-0.12.0 → threadkeeper-0.13.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "threadkeeper"
-version = "0.12.0"
+version = "0.13.0"
 description = "Multi-agent shared brain across Claude Code/Desktop, Codex, Antigravity CLI, Gemini, Copilot, VS Code. Cross-session memory, self-improving skill loops, inter-agent signaling — one local MCP server."
 requires-python = ">=3.11"
 authors = [{ name = "thread-keeper contributors" }]

{threadkeeper-0.12.0 → threadkeeper-0.13.0}/tests/test_menubar_app.py

@@ -46,6 +46,8 @@ def test_menubar_status_item_uses_idle_chip_and_running_gears():
     assert 'button.title = ""' in swift
     assert 'button.title = " TK' not in swift
     assert 'return "TK ' not in swift
+    assert "statusPollInterval: TimeInterval = 15.0" in swift
+    assert "Timer.scheduledTimer(withTimeInterval: statusPollInterval" in swift
     assert "Timer(timeInterval: gearSpinInterval" in swift
     assert "gearFrameStepDegrees = 17.0" in swift
     assert "largeGearDiameter: CGFloat = 12.0" in swift
@@ -59,6 +61,9 @@ def test_menubar_status_item_uses_idle_chip_and_running_gears():
     assert "store.snapshot.runningCount > 0" not in swift
     assert "button.image = gearFrames" in swift
     assert "TimelineView" not in swift
+    assert "refreshInFlight" in swift
+    assert "Task.detached(priority: .utility)" in swift
+    assert "nonisolated private static func runStatusCommand" in swift
     assert "store.openEnvSettings()" in swift
     assert '.help("Settings")' in swift
     assert '.help("Refresh")' not in swift
@@ -67,6 +72,19 @@ def test_menubar_status_item_uses_idle_chip_and_running_gears():
     assert '.help("Clean memory")' in swift
 
 
+def test_menubar_popover_shows_before_status_refresh():
+    repo = Path(__file__).resolve().parents[1]
+    swift = (
+        repo / "apps" / "macos-agent-status" / "ThreadKeeperAgentStatus.swift"
+    ).read_text(encoding="utf-8")
+
+    start = swift.index("@objc private func togglePopover")
+    end = swift.index("    private func updateStatusButton", start)
+    body = swift[start:end]
+
+    assert body.index("popover.show(") < body.index("store.refresh()")
+
+
 def test_menubar_env_settings_window_edits_env_and_presets():
     repo = Path(__file__).resolve().parents[1]
     swift = (
@@ -81,6 +99,22 @@ def test_menubar_env_settings_window_edits_env_and_presets():
     assert "(1...3).map" in swift
     assert "EnvPresetCard" in swift
     assert "mergeEnvText(raw:" in swift
+    assert "EnvSettingsTab" in swift
+    assert "case .raw:" in swift
+    assert "saveRaw(restart:" in swift
+    assert ".onChange(of: envStore.rawEnvText)" not in swift
+    assert "syncRawEditsIntoForm" not in swift
+    assert 'ChoiceOption("antigravity", label: "antigravity (agy)")' in swift
+    assert 'ChoiceOption("agy")' not in swift
+    assert 'ChoiceOption("gemini", label: "gemini (legacy)")' in swift
+    assert "antigravityModelChoices" in swift
+    assert "geminiLegacyModelChoices" in swift
+    assert '"Gemini 3.1 Pro (High)"' in swift
+    assert '"Gemini 3.5 Flash (Medium)"' in swift
+    assert '"gemini-3.1-pro-preview"' in swift
+    assert '"gemini-3.1-pro"' not in swift
+    assert "THREADKEEPER_SPAWN__MODEL__CODEX" in swift
+    assert "THREADKEEPER_SPAWN__MODEL__GEMINI" in swift
     assert "THREADKEEPER_DISABLE_BG_DAEMONS" in swift
     assert "THREADKEEPER_EVOLVE_APPLY_INTERVAL_S" in swift
     assert "THREADKEEPER_SPAWN__MODEL__EVOLVE_APPLIER" in swift

{threadkeeper-0.12.0 → threadkeeper-0.13.0}/tests/test_spawn_config.py

@@ -114,12 +114,12 @@ def test_resolve_model_from_dotenv(tmp_path, monkeypatch):
     envf = tmp_path / "tk.env"
     envf.write_text(
         "THREADKEEPER_SPAWN__MODEL__CODEX=gpt-5.4\n"
-        "THREADKEEPER_SPAWN__MODEL__AGY=gemini-3.1-pro\n"
+        'THREADKEEPER_SPAWN__MODEL__AGY="Gemini 3.1 Pro (High)"\n'
         "THREADKEEPER_SPAWN__MODEL__GEMINI=gemini-2.5-pro\n"
     )
     sc = _reset(monkeypatch, tmp_path, env_file=str(envf))
     assert sc.resolve_model("codex") == "gpt-5.4"
-    assert sc.resolve_model("antigravity") == "gemini-3.1-pro"
+    assert sc.resolve_model("antigravity") == "Gemini 3.1 Pro (High)"
     assert sc.resolve_model("gemini") == "gemini-2.5-pro"
     assert sc.resolve_model("claude") == ""  # no entry
 
@@ -216,12 +216,12 @@ def test_antigravity_spawn_argv_uses_p_flag(tmp_path, monkeypatch):
     for name in [m for m in list(sys.modules) if m.startswith("threadkeeper")]:
         del sys.modules[name]
     from threadkeeper.adapters.antigravity import ADAPTER
-    argv = ADAPTER.spawn_argv("hello", model="gemini-3.1-pro")
+    argv = ADAPTER.spawn_argv("hello", model="Gemini 3.1 Pro (High)")
     if argv is None:
         pytest.skip("agy binary not installed in test env")
     assert "-p" in argv
     assert "--model" in argv
-    assert "gemini-3.1-pro" in argv
+    assert "Gemini 3.1 Pro (High)" in argv
 
 
 def test_gemini_spawn_argv_uses_p_flag(tmp_path, monkeypatch):

threadkeeper-0.13.0/tests/test_verify_ingest.py

@@ -0,0 +1,162 @@
+"""Tests for the cross-CLI production verification harness (issue #1).
+
+The verdict logic is pure, so most of this exercises ``evaluate_coverage``
+and ``evaluate_verdict`` directly. One test drives the read-only SQL layer
+against an in-memory sqlite so the live-DB query path is covered without a
+real ~/.threadkeeper store.
+"""
+from __future__ import annotations
+
+import sqlite3
+
+from threadkeeper.verify_ingest import (
+    CANONICAL_SLOTS,
+    collect_live_signals,
+    evaluate_coverage,
+    evaluate_verdict,
+    format_report,
+    slot_for_source,
+)
+
+
+def test_slot_mapping_groups_gemini_and_antigravity():
+    # Gemini legacy and Antigravity (agy) both satisfy the single Google slot.
+    assert slot_for_source("gemini") == "google"
+    assert slot_for_source("antigravity") == "google"
+    assert slot_for_source("claude-code") == "claude-code"
+    assert slot_for_source("vscode") is None  # not a canonical slot
+
+
+def test_coverage_status_verified_thin_absent():
+    cov = evaluate_coverage(
+        {"claude-code": 200, "codex": 50, "copilot": 2, "gemini": 0},
+        thin_threshold=5,
+    )
+    assert cov["claude-code"]["status"] == "verified"
+    assert cov["codex"]["status"] == "verified"
+    assert cov["copilot"]["status"] == "thin"   # 2 rows, below threshold
+    assert cov["google"]["status"] == "absent"  # gemini=0, no antigravity rows
+    # every canonical slot is represented even when no source mapped to it
+    assert set(cov) == set(CANONICAL_SLOTS)
+
+
+def test_coverage_antigravity_fills_google_slot():
+    cov = evaluate_coverage({"antigravity": 42}, thin_threshold=5)
+    assert cov["google"]["status"] == "verified"
+    assert cov["google"]["sources"] == {"antigravity": 42}
+
+
+def test_verdict_pass_when_all_criteria_met():
+    rep = evaluate_verdict(
+        source_counts={
+            "claude-code": 100, "codex": 100, "copilot": 100, "antigravity": 100,
+        },
+        window_sources=["claude-code", "codex", "antigravity"],
+        shadow_passes=10,
+    )
+    assert rep["verdict"] == "PASS"
+    assert rep["criteria"]["all_sources_present"]["pass"] is True
+    assert rep["criteria"]["cross_adapter_window"]["pass"] is True
+    assert rep["criteria"]["learning_loop_non_claude"]["pass"] is True
+
+
+def test_verdict_partial_three_of_four_slots():
+    # This is the real shape on a dev box: claude/codex/copilot present,
+    # google slot empty, but cross-adapter window + non-claude loop confirmed.
+    rep = evaluate_verdict(
+        source_counts={"claude-code": 200000, "codex": 11000, "copilot": 10},
+        window_sources=["claude-code", "codex"],
+        shadow_passes=2567,
+    )
+    assert rep["verdict"] == "PARTIAL"
+    assert rep["criteria"]["all_sources_present"]["pass"] is False
+    assert rep["criteria"]["all_sources_present"]["verified_slots"] == [
+        "claude-code", "codex", "copilot",
+    ]
+    assert rep["criteria"]["cross_adapter_window"]["pass"] is True
+    assert rep["criteria"]["learning_loop_non_claude"]["pass"] is True
+    assert "codex" in rep["criteria"]["learning_loop_non_claude"]["sources"]
+
+
+def test_verdict_fail_single_adapter_only():
+    # Only Claude Code has data and the window — not a cross-CLI demonstration.
+    rep = evaluate_verdict(
+        source_counts={"claude-code": 5000},
+        window_sources=["claude-code"],
+        shadow_passes=100,
+    )
+    assert rep["verdict"] == "FAIL"
+    assert rep["criteria"]["cross_adapter_window"]["pass"] is False
+    assert rep["criteria"]["learning_loop_non_claude"]["pass"] is False
+
+
+def test_verdict_fail_when_loop_never_ran():
+    rep = evaluate_verdict(
+        source_counts={"claude-code": 100, "codex": 100},
+        window_sources=["claude-code", "codex"],
+        shadow_passes=0,  # learning loop has never fired
+    )
+    # cross-adapter window passes, but loop criterion fails and only 2 slots
+    # verified → PARTIAL (loop is one signal, window is the other).
+    assert rep["verdict"] == "PARTIAL"
+    assert rep["criteria"]["learning_loop_non_claude"]["pass"] is False
+
+
+def _seed_live_db(conn: sqlite3.Connection) -> None:
+    conn.execute(
+        "CREATE TABLE dialog_messages (source TEXT, created_at INTEGER)"
+    )
+    conn.execute("CREATE TABLE events (kind TEXT)")
+    rows = [
+        ("claude-code", 1_000_000),
+        ("claude-code", 1_000_500),
+        ("codex", 1_000_600),   # interleaved with claude in the window
+        ("copilot", 100),       # ancient — outside the recent window
+    ]
+    conn.executemany(
+        "INSERT INTO dialog_messages (source, created_at) VALUES (?, ?)", rows
+    )
+    conn.executemany(
+        "INSERT INTO events (kind) VALUES (?)",
+        [("shadow_review_pass",)] * 3 + [("ingest_pass",)],
+    )
+    conn.commit()
+
+
+def test_collect_live_signals_reads_window_and_passes():
+    conn = sqlite3.connect(":memory:")
+    conn.row_factory = sqlite3.Row
+    _seed_live_db(conn)
+
+    sig = collect_live_signals(conn, window_hours=24)
+    assert sig["source_counts"] == {
+        "claude-code": 2, "codex": 1, "copilot": 1,
+    }
+    # newest is 1_000_600; copilot@100 is far outside a 24h window of it.
+    assert set(sig["window_sources"]) == {"claude-code", "codex"}
+    assert sig["shadow_passes"] == 3
+    assert sig["newest_ts"] == 1_000_600
+
+
+def test_collect_live_signals_tolerates_missing_events_table():
+    conn = sqlite3.connect(":memory:")
+    conn.row_factory = sqlite3.Row
+    conn.execute("CREATE TABLE dialog_messages (source TEXT, created_at INTEGER)")
+    conn.execute("INSERT INTO dialog_messages VALUES ('codex', 5)")
+    conn.commit()
+    sig = collect_live_signals(conn)
+    assert sig["shadow_passes"] == 0  # no events table → graceful 0
+
+
+def test_format_report_renders_verdict_and_slots():
+    rep = evaluate_verdict(
+        source_counts={"claude-code": 200000, "codex": 11000, "copilot": 10},
+        window_sources=["claude-code", "codex"],
+        shadow_passes=2567,
+    )
+    rep["db_path"] = "/tmp/x.sqlite"
+    rep["signals"] = {"window_hours": 24}
+    text = format_report(rep)
+    assert "VERDICT: PARTIAL" in text
+    assert "claude-code" in text
+    assert "learning_loop_non_claude" in text

{threadkeeper-0.12.0 → threadkeeper-0.13.0}/threadkeeper/assets/macos-agent-status/README.md

@@ -5,7 +5,7 @@ The status-bar item itself is AppKit `NSStatusItem`; the popover content is
 SwiftUI. That lets the app update the menu-bar image directly instead of relying
 on SwiftUI `MenuBarExtra` label animation.
 
-It polls `tk-agent-status --json` every 5 seconds and shows:
+It polls `tk-agent-status --json` every 15 seconds and shows:
 
 - an icon-only menu-bar status item, with loop counts in the popover and
   tooltip,
@@ -20,10 +20,14 @@ It polls `tk-agent-status --json` every 5 seconds and shows:
 - active spawned-child RSS when a loop has a worker running,
 - a Clean memory button that runs `tk-agent-status --cleanup-memory`,
 - a Settings gear that opens a separate `~/.threadkeeper/.env` editor with
-  guided controls, raw text editing, three saved presets, and Save & Restart,
+  guided controls, exact dropdowns for spawn CLI/model choices, raw text
+  editing, three saved presets, and Save & Restart,
 - macOS notifications for newly completed autonomous child tasks that produced
   a useful result.
 
+Status polling and cleanup commands run in the background, so opening the
+popover does not wait for `tk-agent-status --json`.
+
 The first poll primes the seen-result list, so the app does not notify for old
 completed tasks that existed before it started.
 
@@ -55,7 +59,9 @@ The Settings gear edits `~/.threadkeeper/.env` by default, or the path in
 `THREADKEEPER_ENV_FILE` when the app was launched with that override. Save &
 Restart writes the file, runs the safe cleanup command, and sends TERM to
 running `threadkeeper.server` processes so MCP hosts reconnect with the new
-environment.
+environment. In the spawn routing controls, `antigravity` is the stored CLI
+value and `agy` is only the executable alias; `gemini` remains available as the
+legacy Gemini CLI adapter.
 
 ## Build