Skip to content

Dodaios is an open-source agent OS for your desktop: it runs your coding agents, browser and apps by voice or chat, on your own machine. Formerly Personal Jarvis.

fix: clear the CI gates the local commits left red

rubenluetke10-beep19352c36716a
diff --git a/docs/authoring/public-docs-manifest.yaml b/docs/authoring/public-docs-manifest.yaml
index 507c73b9fd..e828c06876 100644
--- a/docs/authoring/public-docs-manifest.yaml
+++ b/docs/authoring/public-docs-manifest.yaml
@@ -1,7 +1,7 @@
version: 1
language: en
content_root: docs/product
-page_count: 50
+page_count: 49
navigation_principle: reader-journey
authoring_contract:
audience: People using Personal Jarvis, including readers with no software background.
@@ -202,7 +202,6 @@ pages:
- jarvis-agents
- workflows-and-commands
- safety-and-approvals
- - automations
reader_level: beginner
maturity: stable
@@ -1054,26 +1053,6 @@ pages:
reader_level: advanced
maturity: stable
- - section: Everyday use
- section_order: 2
- order: 9
- slug: automations
- file: docs/product/everyday-use/automations.md
- title: Automations
- diataxis: howto
- summary: Browse automation templates, control recurring work, and inspect schedules and run history.
- authoritative_sources:
- - jarvis/ui/web/frontend/src/views/AutomationsView.tsx
- - jarvis/ui/web/frontend/src/views/automations/TemplateAddDialog.tsx
- - jarvis/tasks/templates/__init__.py
- related:
- - tasks-and-reminders
- - costs
- - workflows-and-commands
- - safety-and-approvals
- reader_level: beginner
- maturity: stable
-
- section: Everyday use
section_order: 2
order: 10
@@ -1087,7 +1066,7 @@ pages:
- jarvis/ui/web/costs_routes.py
related:
- tasks-and-reminders
- - automations
+ - jarvis-agents
- providers-and-api-keys
reader_level: beginner
maturity: stable
diff --git a/docs/commands-reference.md b/docs/commands-reference.md
index 3f4c576839..8380c11dcd 100644
--- a/docs/commands-reference.md
+++ b/docs/commands-reference.md
@@ -87,7 +87,7 @@ The ONLY way to make an agent do something on a schedule (daily briefing, weekly
- **Endpoint:** `POST /api/society/agents/{agent_id}/routines`
- **Arguments:** `agent_id` (string; required); `title` (string; required); `prompt` (string; required); `schedule` (object; required); `announce_on_success` (string; optional)
-- **Requires confirmation:** yes
+- **Requires confirmation:** no
- **Desktop UI section:** `agents`
- **Voice example (EN):** "give me a briefing every morning at eight"
diff --git a/docs/product/everyday-use/automations.md b/docs/product/everyday-use/automations.md
deleted file mode 100644
index 2871abebb4..0000000000
--- a/docs/product/everyday-use/automations.md
+++ /dev/null
@@ -1,176 +0,0 @@
----
-title: "Automations"
-slug: automations
-summary: "Browse automation templates, control recurring work, and inspect schedules and run history."
-section: "Everyday use"
-section_order: 2
-order: 9
-diataxis: howto
-status: active
-owner: maintainers
-last_reviewed: 2026-10-01
-phase: "-"
-audience: end-user
-tags: [automations, templates, schedules, tasks]
-related: [tasks-and-reminders, costs, workflows-and-commands, safety-and-approvals]
----
-
-The **Automations** section brings together reusable recurring work, upcoming
-one-off schedules, completed runs, and a catalogue of templates. Depending on
-your navigation layout, scheduled work may appear under **Scheduled**.
-
-Use this section to answer three questions: what is enabled, what is due next,
-and what actually happened on the last run.
-
-## Choose the Right Starting Point
-
-| Your goal | Start here |
-|---|---|
-| Reuse an available pre-built routine | Catalogue |
-| Create your own recurring instruction | New automation or the custom catalogue tile |
-| Run one instruction later | Schedules |
-| Review a previous attempt | Runs |
-| Pause or inspect recurring work | Automations |
-
-A template saves form-filling. Once added, it creates a task with its own saved
-settings. Browsing a template alone does not create or run that task.
-
-For the detailed one-off task form, intervals, and task approval behavior, see
-[Tasks and Reminders](tasks-and-reminders).
-
-## Before You Add an Automation
-
-- Keep Personal Jarvis running when the automation is due.
-- Connect the services required by the intended task.
-- Check that the selected agent or model is available for that work.
-- Decide how often the result is useful before choosing its schedule.
-- Write inputs that make sense without the current chat conversation.
-
-Use the app's protected connection fields for credentials. Template inputs are
-task content, not a place to save API keys or passwords.
-
-## Browse the Catalogue
-
-Open **Catalogue** to see the templates available in your installation.
-Category filters reduce the list to the kind of work you need.
-
-Each card describes the automation and exposes its add action. Requirements and
-missing capabilities help explain why a template may not be ready to run.
-
-The custom tile opens the task editor when no template fits your purpose.
-An already-added template can offer a shortcut to its existing automation.
-
-An empty catalogue does not mean your saved schedules have disappeared. Templates
-and saved tasks are separate lists; check **Automations** and **Schedules**.
-
-## Add a Template
-
-1. Choose a template and select **Add**.
-2. Review its prefilled title and give it a recognizable name if needed.
-3. Complete the template's inputs, including all required fields.
-4. Review any missing-service or capability message.
-5. Choose the offered schedule kind, time, and weekday where applicable.
-6. Select **Add** to save the automation.
-7. Confirm that its row appears in **Automations**.
-
-The schedule dialog uses your browser's time zone. Review the saved schedule
-after adding it, especially when your device's time zone changes.
-
-An error in the dialog means the addition did not complete. Keep the dialog open,
-correct the indicated input or connection, and confirm a saved row before relying
-on the schedule.
-
-Adding an automation and completing its first run are separate events. Inspect
-the run result before relying on a new routine unattended.
-
-## Read the Automations Table
-
-Each row brings together the recurring task's identity, schedule, last run, and
-enabled state. The table gives an overview without opening every task.
-
-Open a row to see more detail, including the latest result and its step timeline
-when those records are available. A task that has never run has no result yet.
-
-The row's action menu exposes operations such as **Run now** and deletion.
-Actions that fail show an error notice; the existing row remains the reference
-for the task's actual saved state.
-
-## Arm or Pause Recurring Work
-
-The arm switch controls whether the recurring automation is enabled.
-
-1. Find the automation by name.
-2. Turn its switch off to pause future scheduled execution.
-3. Confirm the row shows the disabled state.
-4. Turn it back on when you want scheduled execution to resume.
-
-Pausing preserves the task and its settings. Deleting removes the task instead.
-Do not treat a switch change as proof that an already-running operation stopped;
-check the run's status separately.
-
-## Run Once on Demand
-
-Use **Run now** when you deliberately want an immediate attempt. This executes the
-task's configured work and may call its connected services or model.
-
-After requesting a run, open its details or the **Runs** tab. A successfully
-accepted request does not establish that every step finished successfully.
-
-If the task needs approval, handle the pending request through the app's normal
-approval surface. The catalogue does not grant blanket permission to its tasks.
-
-## Check One-Off Schedules
-
-**Schedules** lists one-off work waiting to fire. Its time and countdown help you
-confirm when an instruction is due.
-
-Use the schedule creation action for a new one-time task. After saving, verify
-the new item appears with the expected time.
-
-A schedule is a plan for future execution. Its outcome belongs in **Runs**, where
-you can check whether it completed, failed, or was cancelled.
-
-## Review Run History
-
-**Runs** shows previous attempts rather than only enabled recurring tasks.
-Use its status filters to find completed, failed, or otherwise relevant work.
-
-Open an entry to read the recorded result and inspect its step timeline.
-The timeline can show which part finished and where a problem occurred.
-
-When investigating a failure:
-
-1. Read the run's result or error.
-2. Check the required connection and model or agent availability.
-3. Correct the cause before requesting another attempt.
-4. Inspect the next run's result to confirm recovery.
-
-Avoid repeatedly selecting **Run now** while an earlier attempt is still active.
-Use the current run's state to decide what to do next.
-
-## Troubleshooting
-
-- **No catalogue:** the backend may not offer templates; existing tasks can still
- be reviewed in the other tabs.
-- **No matching rows:** clear the search or status/category filter.
-- **Missing service:** connect the named service before expecting its steps to work.
-- **No latest result:** the task may not have completed a run yet.
-- **A run did not finish:** inspect its recorded error or pending approval.
-
-## How It Fits Together
-
-A template fills in a saved task. The scheduler watches enabled tasks and starts
-their configured work when due. Run records describe the resulting attempt;
-arming a task and completing a run are distinct states.
-
-## Check That It Works
-
-1. Open an existing automation and check its saved schedule.
-2. Compare the last-run status with its recorded result in **Runs**.
-3. If it has never run, confirm that the view shows no completed result.
-4. Check **Schedules** separately for upcoming one-off tasks.
-
-## Next Steps
-
-For the usage associated with scheduled activity, open
-[Spend & Tokens](costs) and filter to the relevant activity and period.
diff --git a/docs/product/everyday-use/costs.md b/docs/product/everyday-use/costs.md
index 4eb0655a7a..08c028240b 100644
--- a/docs/product/everyday-use/costs.md
+++ b/docs/product/everyday-use/costs.md
@@ -12,7 +12,7 @@ last_reviewed: 2026-10-01
phase: "-"
audience: end-user
tags: [costs, tokens, providers, usage]
-related: [tasks-and-reminders, automations, providers-and-api-keys]
+related: [tasks-and-reminders, jarvis-agents, providers-and-api-keys]
---
Open **Spend & Tokens** to see which recorded activity used tokens and how its
@@ -167,5 +167,5 @@ Use `jarvis costs --help` to discover the supported summaries and filters.
## Next Steps
-For scheduled activity, see [Automations](automations). To create a specific
-one-off task, see [Tasks and Reminders](tasks-and-reminders).
+For recurring work, see the routines on your [Agents](jarvis-agents). To
+create a specific one-off task, see [Tasks and Reminders](tasks-and-reminders).
diff --git a/docs/product/everyday-use/tasks-and-reminders.md b/docs/product/everyday-use/tasks-and-reminders.md
index bcb70e5c97..076c37cb48 100644
--- a/docs/product/everyday-use/tasks-and-reminders.md
+++ b/docs/product/everyday-use/tasks-and-reminders.md
@@ -12,7 +12,7 @@ last_reviewed: 2026-10-01
phase: "-"
audience: end-user
tags: [tasks, reminders, scheduling, recurring, routines, agents]
-related: [jarvis-agents, workflows-and-commands, safety-and-approvals, automations]
+related: [jarvis-agents, workflows-and-commands, safety-and-approvals]
---
Scheduled work lives on your **Agents**. Each agent can carry **routines**: a
diff --git a/jarvis/agentic_ide/orchestration.py b/jarvis/agentic_ide/orchestration.py
index 588aeb9ce4..b2ebcec88b 100644
--- a/jarvis/agentic_ide/orchestration.py
+++ b/jarvis/agentic_ide/orchestration.py
@@ -38,7 +38,7 @@ def _matches(reference: str, *values: str) -> bool:
_FILLER = frozenset(
{
"the", "my", "a", "workspace", "workspaces", "project", "projects", "folder",
- "der", "die", "das", "mein", "meinem", "meinen", "projekt", "ordner",
+ "der", "die", "das", "mein", "meinem", "meinen", "projekt", "ordner", # i18n-allow
"arbeitsbereich",
}
) # fmt: skip
@@ -163,7 +163,7 @@ def _distance(a: str, b: str) -> int:
def _is_request_id(value: str) -> bool:
try:
UUID(value)
- except ValueError:
+ except ValueError: # not a UUID is the answer, not a failure
return False
return True
@@ -378,7 +378,7 @@ class WorkspaceOrchestrator:
}
try:
count = int(args.get("count") or 1)
- except (TypeError, ValueError):
+ except (TypeError, ValueError): # a malformed count falls back to one terminal
count = 1
count = max(1, min(count, MAX_TERMINALS))
name = str(args.get("name") or "").strip()
@@ -409,7 +409,7 @@ class WorkspaceOrchestrator:
created, capped = await self.registry.add_terminals(
count, agent=cli, workspace_id=workspace["id"]
)
- except SessionError as exc:
+ except SessionError as exc: # the refusal is returned to the caller with its reason
return {
"status": "not_accepted",
"reason": str(exc),
@@ -573,7 +573,7 @@ class WorkspaceOrchestrator:
"closed": [t.name for t in closed],
"failed": failed,
}
- except SessionError as exc:
+ except SessionError as exc: # the refusal is returned to the caller with its reason
return {"status": "not_accepted", "target": target, "reason": str(exc)}
raise ValueError("Unknown pane action.")
@@ -615,7 +615,7 @@ class WorkspaceOrchestrator:
}
try:
session = await self.registry.restore_workspace(matched[0][1]["id"])
- except SessionError as exc:
+ except SessionError as exc: # the refusal is returned to the caller with its reason
return {"status": "not_accepted", "reason": str(exc)}
await self._announce(
lambda: workspace_changed_event(
@@ -666,7 +666,7 @@ class WorkspaceOrchestrator:
project_id=project_id,
name=str(args.get("name") or "").strip() or None,
)
- except SessionError as exc:
+ except SessionError as exc: # the refusal is returned to the caller with its reason
return {"status": "not_accepted", "reason": str(exc), "folder": folder}
await self._announce(
lambda: workspace_changed_event(
@@ -728,7 +728,7 @@ class WorkspaceOrchestrator:
}
try:
count = int(entry.get("count") or 1)
- except (TypeError, ValueError):
+ except (TypeError, ValueError): # a malformed count falls back to one terminal
count = 1
groups.append((cli, max(1, min(count, MAX_TERMINALS))))
return groups or [(_default_cli(), 1)]
diff --git a/jarvis/live/tools.py b/jarvis/live/tools.py
index 82303edda9..73a2aadde4 100644
--- a/jarvis/live/tools.py
+++ b/jarvis/live/tools.py
@@ -256,7 +256,7 @@ class LiveTools:
# so the catalog lookup below answered "no longer available".
try:
inner = json.loads(args.get("arguments_json") or "{}")
- except ValueError:
+ except ValueError: # bad JSON is answered by the object check just below
inner = None
if not isinstance(inner, dict):
return {"success": False, "error": "Tool arguments must be an object."}
@@ -289,7 +289,7 @@ class LiveTools:
try:
offset = max(0, int(args.get("offset") or 0))
- except (TypeError, ValueError):
+ except (TypeError, ValueError): # a malformed offset starts the listing from the top
offset = 0
return discover(self.catalog(), str(args.get("query", "")), offset)
if name == "confirm_action":
@@ -325,7 +325,7 @@ class LiveTools:
canonical = str(args.get("name", ""))
try:
args = json.loads(args.get("arguments_json") or "{}")
- except ValueError:
+ except ValueError: # bad JSON is answered by the object check just below
args = None
if not isinstance(args, dict):
return {
diff --git a/jarvis/plugins/realtime/local_voice.py b/jarvis/plugins/realtime/local_voice.py
index 76c58c602e..6d9a50c429 100644
--- a/jarvis/plugins/realtime/local_voice.py
+++ b/jarvis/plugins/realtime/local_voice.py
@@ -164,7 +164,7 @@ class _Engine:
async def wait_ready(self, timeout_s: float) -> bool:
try:
await asyncio.wait_for(self._ready.wait(), timeout_s)
- except TimeoutError:
+ except TimeoutError: # not ready in time is the answer the caller asked for
return False
return self.phase == "ready"
diff --git a/jarvis/ui/web/frontend/src/components/home/useSpokenCursor.test.ts b/jarvis/ui/web/frontend/src/components/home/useSpokenCursor.test.ts
index 4460fd7b6a..706bd00a4a 100644
--- a/jarvis/ui/web/frontend/src/components/home/useSpokenCursor.test.ts
+++ b/jarvis/ui/web/frontend/src/components/home/useSpokenCursor.test.ts
@@ -27,7 +27,7 @@ describe("useSpokenCursor", () => {
it("moves while audio plays and waits while it is silent", () => {
vi.useFakeTimers({ toFake: ["requestAnimationFrame", "cancelAnimationFrame", "performance"] });
- const text = "Morgen ist Freitag, der zweite Oktober. Hast du was Besonderes vor?";
+ const text = "Tomorrow is Friday, the second of October. Anything special planned?";
voiceOutputLevelRef.current = 0.5;
const { result } = renderHook(() => useSpokenCursor(text, true));
expect(result.current).toBe(0);
@@ -47,11 +47,11 @@ describe("useSpokenCursor", () => {
it("starts a new answer from its beginning", () => {
vi.useFakeTimers({ toFake: ["requestAnimationFrame", "cancelAnimationFrame", "performance"] });
const { result, rerender } = renderHook(({ text }) => useSpokenCursor(text, true), {
- initialProps: { text: "Erste Antwort mit ein paar Worten" },
+ initialProps: { text: "A first answer with a few words" },
});
act(() => vi.advanceTimersByTime(1000));
expect(result.current).toBeGreaterThan(0);
- rerender({ text: "Ganz andere Antwort" });
- expect(result.current).toBeLessThanOrEqual("Ganz".length);
+ rerender({ text: "Quite another answer" });
+ expect(result.current).toBeLessThanOrEqual("Quite".length);
});
});
diff --git a/jarvis/ui/web/frontend/src/components/home/voiceGlowRenderer.ts b/jarvis/ui/web/frontend/src/components/home/voiceGlowRenderer.ts
index c6a1ad0864..4fdb76dc5b 100644
--- a/jarvis/ui/web/frontend/src/components/home/voiceGlowRenderer.ts
+++ b/jarvis/ui/web/frontend/src/components/home/voiceGlowRenderer.ts
@@ -168,6 +168,10 @@ function compile(gl: WebGLRenderingContext, type: number, source: string): WebGL
* Build the renderer on `canvas`, or return null where WebGL is unavailable
* (no GPU, a blocklisted driver, the page out of contexts) — the caller then
* paints the CSS fallback.
+ *
+ * `dispose()` frees the program and buffer only. The caller owns the context
+ * and hands it back with `releaseWebglContext` after unmount (VoiceGlow does,
+ * deferred so a StrictMode remount can reuse it) — AP-32.
*/
export function createGlowRenderer(canvas: HTMLCanvasElement): GlowRenderer | null {
let gl: WebGLRenderingContext | null = null;
diff --git a/jarvis/voice_engine/bench/__main__.py b/jarvis/voice_engine/bench/__main__.py
index 66506a3912..f4ac7e7668 100644
--- a/jarvis/voice_engine/bench/__main__.py
+++ b/jarvis/voice_engine/bench/__main__.py
@@ -32,7 +32,7 @@ def _options(pairs: list[str]) -> dict[str, Any]:
key, _, raw = pair.partition("=")
try:
options[key] = json.loads(raw)
- except json.JSONDecodeError:
+ except json.JSONDecodeError: # a value that is not JSON is kept as a plain string
options[key] = raw
return options
diff --git a/jarvis/voice_engine/bench/corpus.py b/jarvis/voice_engine/bench/corpus.py
index ffaf5adb5a..a7bfa198f3 100644
--- a/jarvis/voice_engine/bench/corpus.py
+++ b/jarvis/voice_engine/bench/corpus.py
@@ -93,7 +93,7 @@ def _matches(rule: dict[str, Any], actual: Any) -> bool:
expected = rule["equals"]
try:
return float(actual) == float(expected)
- except (TypeError, ValueError):
+ except (TypeError, ValueError): # non-numeric values are compared as text instead
return fold(actual).strip() == fold(expected).strip()
if "contains" in rule:
options = rule["contains"]
@@ -132,7 +132,7 @@ def _arguments(call: dict[str, Any]) -> dict[str, Any] | None:
if isinstance(args, str):
try:
args = json.loads(args)
- except json.JSONDecodeError:
+ except json.JSONDecodeError: # unparseable arguments mean no match, scored by the caller
return None
return args if isinstance(args, dict) else None
diff --git a/jarvis/voice_engine/bench/env.py b/jarvis/voice_engine/bench/env.py
index 8227719b27..945b4c1d44 100644
--- a/jarvis/voice_engine/bench/env.py
+++ b/jarvis/voice_engine/bench/env.py
@@ -72,7 +72,7 @@ def load_snapshot() -> dict[str, Any]:
"cpu_percent": psutil.cpu_percent(interval=1.0),
"ram_available_gb": round(psutil.virtual_memory().available / 1e9, 1),
}
- except ImportError:
+ except ImportError: # optional module; the report simply omits it
return {}
@@ -96,7 +96,7 @@ def describe() -> dict[str, Any]:
try:
imported = __import__(module)
info[f"version_{module}"] = getattr(imported, "__version__", "?")
- except ImportError:
+ except ImportError: # optional module; the report records it as absent
info[f"version_{module}"] = None
info["load"] = load_snapshot()
return info
diff --git a/jarvis/voice_engine/bench/worker_suite.py b/jarvis/voice_engine/bench/worker_suite.py
index 70c9f23e1d..87cbf11fda 100644
--- a/jarvis/voice_engine/bench/worker_suite.py
+++ b/jarvis/voice_engine/bench/worker_suite.py
@@ -275,7 +275,7 @@ def run_drill(trials: int = 3) -> dict[str, Any]:
if psutil.Process(worker_pid).status() == psutil.STATUS_ZOMBIE:
gone_after = round(time.monotonic() - killed_at, 2)
break
- except psutil.NoSuchProcess:
+ except psutil.NoSuchProcess: # the process being gone is exactly what is measured
gone_after = round(time.monotonic() - killed_at, 2)
break
time.sleep(0.05)
diff --git a/jarvis/voice_engine/client.py b/jarvis/voice_engine/client.py
index 15812ae01a..6a9032821c 100644
--- a/jarvis/voice_engine/client.py
+++ b/jarvis/voice_engine/client.py
@@ -121,7 +121,7 @@ class EngineClient:
process.stdin.close()
try:
code = await asyncio.wait_for(process.wait(), timeout_s)
- except TimeoutError:
+ except TimeoutError: # a worker that ignores shutdown is killed instead
process.kill()
code = await process.wait()
if self._reader_task is not None:
diff --git a/jarvis/voice_engine/engine.py b/jarvis/voice_engine/engine.py
index 4a2b798834..9322cf90ce 100644
--- a/jarvis/voice_engine/engine.py
+++ b/jarvis/voice_engine/engine.py
@@ -485,7 +485,7 @@ class ConversationSession:
"name": call["name"], "arguments": call["arguments"]})
try:
return await asyncio.wait_for(waiter, self._config.tool_timeout_s)
- except TimeoutError:
+ except TimeoutError: # the timeout is returned to the model as the tool result
self._tool_waiters.pop(call["id"], None)
return {"success": False, "error": "The tool did not answer in time."}
diff --git a/tests/unit/ui/test_pet_notices.py b/tests/unit/ui/test_pet_notices.py
index 5c4caab5ca..d4872e5995 100644
--- a/tests/unit/ui/test_pet_notices.py
+++ b/tests/unit/ui/test_pet_notices.py
@@ -48,7 +48,8 @@ def test_a_failed_chat_turn_is_a_cross_and_an_empty_one_is_nothing() -> None:
failed = JarvisChatTurnFinished(status="error", user_text="Book a table")
assert notices.for_chat_turn(failed, "en") == ("error", "Book a table", "")
no_text = JarvisChatTurnFinished(status="error")
- assert notices.for_chat_turn(no_text, "de") == ("error", "Jarvis konnte nicht antworten", "")
+ german = "Jarvis konnte nicht antworten" # i18n-allow: quotes the de locale
+ assert notices.for_chat_turn(no_text, "de") == ("error", german, "")
assert notices.for_chat_turn(JarvisChatTurnFinished(), "en") is None