taskuary 0.3.7.0__tar.gz → 0.3.7.2__tar.gz

{taskuary-0.3.7.0/taskuary.egg-info → taskuary-0.3.7.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: taskuary
-Version: 0.3.7.0
+Version: 0.3.7.2
 Summary: Automate your job - the local-first operating system for your work: AI triage in, coding agents out.
 License: MIT
 Project-URL: Homepage, https://github.com/ldbumble/taskuary
@@ -69,7 +69,7 @@ Dynamic: license-file
 
 [![CI](https://github.com/ldbumble/taskuary/actions/workflows/ci.yml/badge.svg)](https://github.com/ldbumble/taskuary/actions/workflows/ci.yml)
 [![OpenSSF Scorecard](https://api.scorecard.dev/projects/github.com/ldbumble/taskuary/badge)](https://scorecard.dev/viewer/?uri=github.com/ldbumble/taskuary)
-[![PyPI](https://img.shields.io/pypi/v/taskuary.svg?cacheSeconds=300&release=0.3.7.0&asof=2026-09-28T1810)](https://pypi.org/project/taskuary/)
+[![PyPI](https://img.shields.io/pypi/v/taskuary.svg?cacheSeconds=300&release=0.3.7.2&asof=2026-09-28T1810)](https://pypi.org/project/taskuary/)
 [![Python 3.10+](https://img.shields.io/badge/python-3.10%2B-3776ab.svg)](https://github.com/ldbumble/taskuary)
 [![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)
 [![Stars](https://img.shields.io/github/stars/ldbumble/taskuary?style=flat&color=d4a72c&label=%E2%98%85%20stars)](https://github.com/ldbumble/taskuary/stargazers)
@@ -83,7 +83,7 @@ your agents, and brings decisions back to you. Nothing sends or ships without yo
 
 ![The Taskuary Studio assembling as work arrives and AI agents take their seats.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/hero.gif?v=workspace)
 
-Taskuary is early—currently **v0.3.7.0**—so breaking changes are still possible before 1.0.
+Taskuary is early—currently **v0.3.7.2**—so breaking changes are still possible before 1.0.
 
 <p align="center">
   <a href="https://taskuary.com/demo/"><img
@@ -100,47 +100,50 @@ Taskuary is early—currently **v0.3.7.0**—so breaking changes are still possi
 One request, from arrival to your approval. Follow Ruth's request for the latest vendor spend
 numbers through the real app, using fictional demo data.
 
-### 1. Connect every system. Keep control.
+### 1. Everything lands in one place
 
-Mail, chats, issue trackers, alerts, and reports land on one Timeline. See what arrived,
-what became a task, and what needs you without opening every system in turn.
+Mail, chats, issue trackers, alerts, and reports arrive on one rail, sorted by what they need:
+**Urgent**, **On you**, **For later**, the Advisor's ideas, and **FYI**. Beside it, the Assistant
+says what came in and who is waiting, so you never open each system in turn.
 
-![An annotated Timeline: each item shows its original time below its source—Email, Teams, WhatsApp, GitHub, SQL report, Assistant, Daily digest, or Calendar.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/01-timeline-sources-and-times.png)
+![The work rail sorting today's items into Urgent, On you, Advisor ideas, and FYI, beside the Assistant's morning summary of who is waiting.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/01-timeline-sources-and-times.png?v=95295dd)
 
-### 2. Turn incoming work into tasks
+### 2. Let the Assistant walk you through it
 
-Ruth asks for the August total, the change from July, and a breakdown by category.
-Taskuary creates a task with the original request and assigns it to an agent.
+Choose **Walk me through my tasks**. The Assistant puts one item on the table at a time, as the
+same task card you would open yourself, with the next actions under it: **Next**, **Write reply**,
+**Start an agent**. The buttons always do the same thing; the chat line is for everything else,
+and the Assistant already has the message, the thread, and your past decisions to answer from.
 
-![Ruth's request becomes TQ-0018, with its instructions, owner, and agent work together on the task page.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/02-task.png)
+![The Assistant bringing Ruth's request into the chat as task TQ-0018, with Next, Write reply, and Start an agent under it.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/05-assistant.png?v=95295dd)
 
-### 3. Watch the agent work
+### 3. Every request becomes a task with its whole story
 
-Open the task to follow the analysis. Here, the general agent prepares the numbers,
-checks that the categories add up, and drafts a reply.
+Open any item to see how it got here: the message, what triage decided and why, which agent
+has it, and where the reply stands.
 
-![The general agent's completed vendor spend analysis, with category totals, comparison, and source.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/03-agent.png)
+![Ruth's request on its task: the message, the triage decision, the agent, and the reply still to write.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/02-task.png?v=95295dd)
 
-### 4. Approve the outcome
+### 4. The agent works inside the chat
 
-The reply waits in **Review**, beside the request that started it. Read it, edit it,
-and choose **Approve & send** when it is ready.
+**Start an agent** and it works in the same card. Here the analyst prepares the numbers, checks
+that the categories add up, names its source, and drafts the reply. Nothing is sent.
 
-![Review shows Ruth's original request and the prepared reply, with Approve & send waiting for the owner.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/04-review.png)
+![The analyst's finished vendor spend analysis inside the chat, with category totals, the change from July, and its source.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/03-agent.png?v=95295dd)
 
-### 5. Let the Assistant walk you through it
+### 5. The last word is yours
 
-Choose **Walk me through my tasks**. The Assistant brings one item into the conversation,
-explains what needs your attention, and puts the next action within reach.
+The prepared reply waits on the task, beside the request that started it. Read it, edit it,
+and choose **Approve & send** when it is ready.
 
-![A close-up of the Assistant bringing Ruth's request into the conversation, with a link to its task.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/05-assistant.png)
+![The drafted reply to Ruth waiting on its task, with Approve & send, Reject, and Regenerate with AI.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/04-review.png?v=95295dd)
 
-### 6. Start your day with the daily digest
+### 6. Start your day before it gets busy
 
-Open your morning brief to see what people need, what is in flight, and what is on your calendar.
-The day's meetings sit above the digest, so Ruth's request has a clear deadline: the operations review.
+Each morning the Assistant opens with today's meetings on a timeline and who needs a word from you.
+Ruth's request has a clear deadline: the 11:30 operations review.
 
-![An animated close-up of the morning digest and calendar, showing the operations review and vendor planning meeting.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/06-morning.gif)
+![An animated close-up of the morning: today's meetings on a timeline, then the people and agents waiting.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/06-morning.gif?v=95295dd)
 
 ## Key features
 

{taskuary-0.3.7.0 → taskuary-0.3.7.2}/README.md

@@ -4,7 +4,7 @@
 
 [![CI](https://github.com/ldbumble/taskuary/actions/workflows/ci.yml/badge.svg)](https://github.com/ldbumble/taskuary/actions/workflows/ci.yml)
 [![OpenSSF Scorecard](https://api.scorecard.dev/projects/github.com/ldbumble/taskuary/badge)](https://scorecard.dev/viewer/?uri=github.com/ldbumble/taskuary)
-[![PyPI](https://img.shields.io/pypi/v/taskuary.svg?cacheSeconds=300&release=0.3.7.0&asof=2026-09-28T1810)](https://pypi.org/project/taskuary/)
+[![PyPI](https://img.shields.io/pypi/v/taskuary.svg?cacheSeconds=300&release=0.3.7.2&asof=2026-09-28T1810)](https://pypi.org/project/taskuary/)
 [![Python 3.10+](https://img.shields.io/badge/python-3.10%2B-3776ab.svg)](https://github.com/ldbumble/taskuary)
 [![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)
 [![Stars](https://img.shields.io/github/stars/ldbumble/taskuary?style=flat&color=d4a72c&label=%E2%98%85%20stars)](https://github.com/ldbumble/taskuary/stargazers)
@@ -18,7 +18,7 @@ your agents, and brings decisions back to you. Nothing sends or ships without yo
 
 ![The Taskuary Studio assembling as work arrives and AI agents take their seats.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/hero.gif?v=workspace)
 
-Taskuary is early—currently **v0.3.7.0**—so breaking changes are still possible before 1.0.
+Taskuary is early—currently **v0.3.7.2**—so breaking changes are still possible before 1.0.
 
 <p align="center">
   <a href="https://taskuary.com/demo/"><img
@@ -35,47 +35,50 @@ Taskuary is early—currently **v0.3.7.0**—so breaking changes are still possi
 One request, from arrival to your approval. Follow Ruth's request for the latest vendor spend
 numbers through the real app, using fictional demo data.
 
-### 1. Connect every system. Keep control.
+### 1. Everything lands in one place
 
-Mail, chats, issue trackers, alerts, and reports land on one Timeline. See what arrived,
-what became a task, and what needs you without opening every system in turn.
+Mail, chats, issue trackers, alerts, and reports arrive on one rail, sorted by what they need:
+**Urgent**, **On you**, **For later**, the Advisor's ideas, and **FYI**. Beside it, the Assistant
+says what came in and who is waiting, so you never open each system in turn.
 
-![An annotated Timeline: each item shows its original time below its source—Email, Teams, WhatsApp, GitHub, SQL report, Assistant, Daily digest, or Calendar.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/01-timeline-sources-and-times.png)
+![The work rail sorting today's items into Urgent, On you, Advisor ideas, and FYI, beside the Assistant's morning summary of who is waiting.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/01-timeline-sources-and-times.png?v=95295dd)
 
-### 2. Turn incoming work into tasks
+### 2. Let the Assistant walk you through it
 
-Ruth asks for the August total, the change from July, and a breakdown by category.
-Taskuary creates a task with the original request and assigns it to an agent.
+Choose **Walk me through my tasks**. The Assistant puts one item on the table at a time, as the
+same task card you would open yourself, with the next actions under it: **Next**, **Write reply**,
+**Start an agent**. The buttons always do the same thing; the chat line is for everything else,
+and the Assistant already has the message, the thread, and your past decisions to answer from.
 
-![Ruth's request becomes TQ-0018, with its instructions, owner, and agent work together on the task page.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/02-task.png)
+![The Assistant bringing Ruth's request into the chat as task TQ-0018, with Next, Write reply, and Start an agent under it.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/05-assistant.png?v=95295dd)
 
-### 3. Watch the agent work
+### 3. Every request becomes a task with its whole story
 
-Open the task to follow the analysis. Here, the general agent prepares the numbers,
-checks that the categories add up, and drafts a reply.
+Open any item to see how it got here: the message, what triage decided and why, which agent
+has it, and where the reply stands.
 
-![The general agent's completed vendor spend analysis, with category totals, comparison, and source.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/03-agent.png)
+![Ruth's request on its task: the message, the triage decision, the agent, and the reply still to write.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/02-task.png?v=95295dd)
 
-### 4. Approve the outcome
+### 4. The agent works inside the chat
 
-The reply waits in **Review**, beside the request that started it. Read it, edit it,
-and choose **Approve & send** when it is ready.
+**Start an agent** and it works in the same card. Here the analyst prepares the numbers, checks
+that the categories add up, names its source, and drafts the reply. Nothing is sent.
 
-![Review shows Ruth's original request and the prepared reply, with Approve & send waiting for the owner.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/04-review.png)
+![The analyst's finished vendor spend analysis inside the chat, with category totals, the change from July, and its source.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/03-agent.png?v=95295dd)
 
-### 5. Let the Assistant walk you through it
+### 5. The last word is yours
 
-Choose **Walk me through my tasks**. The Assistant brings one item into the conversation,
-explains what needs your attention, and puts the next action within reach.
+The prepared reply waits on the task, beside the request that started it. Read it, edit it,
+and choose **Approve & send** when it is ready.
 
-![A close-up of the Assistant bringing Ruth's request into the conversation, with a link to its task.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/05-assistant.png)
+![The drafted reply to Ruth waiting on its task, with Approve & send, Reject, and Regenerate with AI.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/04-review.png?v=95295dd)
 
-### 6. Start your day with the daily digest
+### 6. Start your day before it gets busy
 
-Open your morning brief to see what people need, what is in flight, and what is on your calendar.
-The day's meetings sit above the digest, so Ruth's request has a clear deadline: the operations review.
+Each morning the Assistant opens with today's meetings on a timeline and who needs a word from you.
+Ruth's request has a clear deadline: the 11:30 operations review.
 
-![An animated close-up of the morning digest and calendar, showing the operations review and vendor planning meeting.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/06-morning.gif)
+![An animated close-up of the morning: today's meetings on a timeline, then the people and agents waiting.](https://raw.githubusercontent.com/ldbumble/taskuary/master/docs/readme/06-morning.gif?v=95295dd)
 
 ## Key features
 

{taskuary-0.3.7.0 → taskuary-0.3.7.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "taskuary"
-version = "0.3.7.0"
+version = "0.3.7.2"
 description = "Automate your job - the local-first operating system for your work: AI triage in, coding agents out."
 readme = "README.md"
 license = { text = "MIT" }

{taskuary-0.3.7.0 → taskuary-0.3.7.2}/taskuary/aidefaults.py

@@ -18,7 +18,7 @@ model for coding sessions and a LIGHT model for the one-message jobs - triage, d
 summaries, the digest (llm.make_cli_llm). Naming the same CLI as both the coder and the triage
 brain is normal and cheap; it is the light gear that keeps it cheap.
 """
-import json
+import json, re
 
 from . import climodels
 
@@ -106,7 +106,7 @@ def resolve(store, cfg, slot_key: str) -> dict:
         cli = _cli_of(cfg, store, name) if name else ''
         cat = climodels.catalog(cli)
         model, effort = climodels.split_pick(_gears(cfg, store, name).get('model') or '')
-        out.update(model=model, effort=effort, choices=cat['choices'], efforts=_efforts(cat, model),
+        out.update(model=model, effort=effort, choices=cat['choices'], efforts=_efforts(cat, model), resolved=_resolved(cat, model),
                    owner=f'the {cli} coding profile' if cli else '', owner_link='agents', cli=cli,
                    display=cli,
                    ready=bool(name))
@@ -159,7 +159,9 @@ def resolve(store, cfg, slot_key: str) -> dict:
             from .llm import ASSISTANT_EFFORT
             effort = ASSISTANT_EFFORT.get(cli, '')          # what concierge.brain runs when the pick names no effort
         said = (f"{dflt[7:]} effort" if dflt.startswith('effort:') else dflt) if dflt else f'the {cli} default'
-        out.update(model=model, effort=effort, choices=cat['choices'], efforts=_efforts(cat, model),
+        # a blank box runs the gear's default model, so its efforts and version are that model's
+        seen = model or ('' if dflt.startswith('effort:') else dflt)
+        out.update(model=model, effort=effort, choices=cat['choices'], efforts=_efforts(cat, seen), resolved=_resolved(cat, seen),
                    default_hint=f"{said} - the {'Assistant' if slot_key == 'concierge_ai' else 'light'} default",
                    owner=owner, owner_link='' if slot_key == 'concierge_ai' else 'agents', cli=cli)
         if not light and not dflt: out['note'] = f'no light model set - this runs on the {cli} coding model, which is the expensive gear'
@@ -184,10 +186,25 @@ def resolve(store, cfg, slot_key: str) -> dict:
     return out
 
 
+def _entry(cat: dict, model: str) -> dict:
+    """The catalogue row a model names: the exact id, else an alias (`sonnet`) read as the NEWEST id of that family, which is
+    what the CLI itself runs - so an alias pick shows its efforts and the version it lands on."""
+    ms = cat.get('models') or []
+    if (hit := next((m for m in ms if m.get('id') == model), None)): return hit
+    fam = [m for m in ms if model and re.fullmatch(rf'claude-{re.escape(model)}-\d[\w-]*', str(m.get('id')))]
+    return max(fam, key=lambda m: [int(n) for n in re.findall(r'\d+', m['id'])], default={})
+
+
 def _efforts(cat: dict, model: str) -> list:
     """The reasoning levels THIS model offers, per the CLI's own catalogue. Empty for every CLI
     whose effort flag we cannot spell - climodels documents why muse is deliberately empty."""
-    return next((m.get('efforts') or [] for m in cat.get('models') or [] if m.get('id') == model), [])
+    return _entry(cat, model).get('efforts') or []
+
+
+def _resolved(cat: dict, model: str) -> str:
+    """The id an alias lands on ('' when the pick already is an id, or nothing is known)."""
+    i = _entry(cat, model).get('id') or ''
+    return i if i != model else ''
 
 
 # What a blank model box actually means per provider, copied from llm.make_llm so the UI can

{taskuary-0.3.7.0 → taskuary-0.3.7.2}/taskuary/chatformat.py

@@ -188,6 +188,22 @@ def split(text: str, limit: int = HARD) -> list:
     return out
 
 
+def json_rows(text: str) -> str:
+    """Rows that arrived as one JSON object per line - a report whose summary never came, the demo's - as a markdown table,
+    so the door lays them out like any other table ({"site": "Lakeview", "headcount": 112} went to the phone as that, braces
+    and quotes, 2026-09-30 phone test). Anything that is not ALL flat objects is left exactly as it was."""
+    import json
+    lines = [l.strip() for l in str(text or '').splitlines() if l.strip()]
+    if len(lines) < 2: return text
+    try: rows = [json.loads(l) for l in lines]
+    except ValueError: return text
+    if not all(isinstance(r, dict) and r and all(not isinstance(v, (dict, list)) for v in r.values()) for r in rows): return text
+    cols = list(dict.fromkeys(k for r in rows for k in r))
+    cell = lambda v: '' if v is None else str(v).replace('|', '/')
+    return '\n'.join(['| ' + ' | '.join(cols) + ' |', '|' + '---|' * len(cols)]
+                     + ['| ' + ' | '.join(cell(r.get(c)) for c in cols) + ' |' for r in rows])
+
+
 def blocks(body, limit: int = HARD) -> list:
     """A report as the messages a chat should receive: one per section, in order.
 
@@ -200,7 +216,7 @@ def blocks(body, limit: int = HARD) -> list:
     out = []
     for title, text in sections(strip_raw(body)):
         head = f'## {title}' if title else ''
-        text = text.strip()
+        text = json_rows(text.strip()).strip()
         if not text: continue
         for piece in split(text, max(200, limit - len(head) - 2)):
             out.append(f'{head}\n{piece}'.strip() if head else piece)

{taskuary-0.3.7.0 → taskuary-0.3.7.2}/taskuary/cli.py

@@ -80,7 +80,7 @@ def main():
                          'email. Prints what it WOULD change and writes nothing unless --apply.')
     ap.add_argument('--apply', action='store_true', help='with --backfill-asks: actually write the changes')
     ap.add_argument('--include-closed', action='store_true', help='with --backfill-asks: closed tasks too')
-    ap.add_argument('--evalset', choices=['build', 'share', 'evaluate', 'ablate'], metavar='ACTION',
+    ap.add_argument('--evalset', choices=['build', 'share', 'evaluate', 'ablate', 'jev'], metavar='ACTION',
                     help='triage dataset: build (labelled cases from your verdicts -> ~/.taskuary/eval), '
                          'evaluate (score the configured AI over them), ablate (score with and without memory), '
                          'share (anonymised copy for tests/data)')

{taskuary-0.3.7.0 → taskuary-0.3.7.2}/taskuary/concierge.py

@@ -76,7 +76,11 @@ CHIP_HINTS = {'not_ours': 'File it - the card asks whether just this once, from
               'regular_agent': 'An agent takes it - triage picks a coding or a non-coding one, and you can change it on the card',
               'mine': "A task on your own list - no agent starts", 'next': 'Read it and move on',
          'defer': 'Put the task away until a day - it is Upcoming in Tasks and back on your rail that morning',
-         'continue': 'Pick the agent up where it left off - say what to tell it, or continue as is'}
+         'continue': 'Pick the agent up where it left off - say what to tell it, or continue as is',
+         # the words that said nothing on hover: each line is what the handler does (TaskPage's own titles are the same sentences)
+         'stop_agent': "Writes up what the session did, ends it, and drafts the reply to whoever asked (with nothing to write up it just stops it) - the task stays open",
+         'close': 'Closes the task - it stops coming back to Work',
+         'rerun': 'Runs the report again in the background - it comes back here when it is done'}
 # per kind, in the order they are offered. `next` is last on every one of them: moving on is always available,
 # and it is the one word that is never a decision about the thing itself.
 # THE SHORT LIST (the owner, 2026-09-25, word by word): eight buttons. Tomorrow and Later are gone - Next on
@@ -848,7 +852,12 @@ def _ask(store, llm, tid: int, item: dict | None, instruction: str, pile_items:
     return say, options, (decision or {}).get('verb') or ''
 
 
-def record(store, tid: int, role: str, text: str, card: dict = None):
+# A line said to the PHONE only (the day's opener): kept in the conversation so the model and the desk know it was said,
+# but never drawn by the desktop chat, whose own welcome already shows the day (the owner, 2026-09-30: "that's only for the phone")
+PHONE_ACTOR = 'assistant-phone'
+
+
+def record(store, tid: int, role: str, text: str, card: dict = None, phone: bool = False):
     # A presentation revision is transport freshness, not conversation meaning.  Persisting it
     # made the same adjacent card into a second durable turn when surfacing changed only its read
     # state.  Keep every semantic field, including nested FYI cards, and let the current HTTP
@@ -863,7 +872,7 @@ def record(store, tid: int, role: str, text: str, card: dict = None):
         return out
     saved_card = durable(card) if card else None
     body = redact(text).strip() + (f"\n\n{MARK}{json.dumps(saved_card, default=str)} -->" if saved_card else '')
-    actor = 'owner' if role == 'user' else 'assistant'
+    actor = 'owner' if role == 'user' else PHONE_ACTOR if phone else 'assistant'
     actor_type = general.USER_TYPE if role == 'user' else general.ASSISTANT_TYPE
     # A double click must not duplicate the owner's words: those two calls may reach the backend
     # before React disables the button. The same guard also makes empty-pipe auto-advance
@@ -911,6 +920,7 @@ def record_related(store, dock_tid: int, item: dict | None, role: str, text: str
 def history(store, tid: int) -> list:
     out = []
     for c in general.chat_rows(store, tid):
+        if c.get('Actor') == PHONE_ACTOR: continue      # said to the phone; the model still reads it through chat_rows
         body = c.get('Body') or ''
         m = _MARK.search(body)
         card = None
@@ -1518,7 +1528,11 @@ def call_turn(store, tid: int, call: dict, item: dict | None, text: str, actor:
         # ...but an AGENT is never put on an old task by default: a new ask typed while something sits on the table
         # became a run on that task (the owner, 2026-09-29: "if i ask for task it should not merge with old one or agent
         # run unless i ask"). These two must NAME the task; a new job is task.create_from_text, which the miss says.
-        if kind in toolcatalog.ATTACH_AGENT and not ref:
+        # ...but CONTINUING is not attaching: it reopens the task's OWN saved session (the owner, 2026-09-30: "continue session
+        # and tell the coding agent to merge it in" on a close-out card was refused for want of a ref, and the retry
+        # fell to the close-out itself). On the table, with a session to reopen, the table's task is what they mean.
+        resumes = kind == 'agent.continue' and not ref and it.get('tid') and store.resumable_session(it['tid'])
+        if kind in toolcatalog.ATTACH_AGENT and not ref and not resumes:
             raise CallMiss(f"{kind} puts an agent on a task that already exists - name it (its TQ ref, the one on the table "
                            "included). A new job is task.create_from_text. Nothing was started.")
         t = int(ref.group(1)) if ref else it.get('tid')
@@ -1865,7 +1879,9 @@ def surface(store, key: str = None, llm=None, actor: str = 'owner', only: str =
                    f"I'll bring {'it' if n == 1 else 'them'} round again in a while - or open {'it' if n == 1 else 'one'} now.")
             # ...and the way to them NOW, by name: a Next under this line only said it again (2026-09-23, six days running).
             # What waits on the owner first - a close-out, an agent's question - then the rest.
-            opens = [{'verb': 'open', 'key': i['key'], 'label': f"Open {i.get('ref') or _title_cut(i.get('title') or 'it', 40)}"}
+            # ...each saying WHAT it is, not a bare number (the owner, 2026-09-30: "have to write what they are about a little")
+            opens = [{'verb': 'open', 'key': i['key'], 'label': ' · '.join(x for x in (
+                         f"Open {i.get('ref') or ''}".strip(), _title_cut(' '.join(str(i.get('title') or 'it').split()), 50)) if x)}
                      for i in sorted(waiting, key=lambda i: not funnel.on_you(i))[:3]]
         elif only and left:
             # the filtered set is done; what remains is the rest of the pipe - offer it rather than call the day over
@@ -1896,7 +1912,7 @@ def surface(store, key: str = None, llm=None, actor: str = 'owner', only: str =
         return surface(store, None, llm, actor, only, trace, cancel, include_surfaced, exclude) if not key else {'item': None, 'say': say, 'options': [], 'chips': walk_chips(len(p['items'])), 'left': len(p['items'])}
     # FYIs have no action to take, so the normal walk brings four together. A row explicitly
     # clicked on the Timeline still opens by itself (`key` is set); only Next/Walk batches them.
-    if not key and item['lane'] == 'fyi':
+    if not key and funnel.batchable(item):
         # a batch already assembled (funnel_selection._batch, which next_item returns on a processing store)
         # IS the batch: passed to fyi_batch as `first` it became its own member, "someone - 4 fyi" (2026-09-18)
         batch = item.get('items') if item.get('kind') == 'fyis' else funnel.fyi_batch(store, item)
@@ -2721,7 +2737,7 @@ def _chat_proposed(store, dock_tid: int, oid: str) -> bool:
 
 
 def say(store, text: str, key: str = None, llm=None, actor: str = 'owner', trace=None, cancel=None, item: dict | None = None,
-        open_card: str | None = None) -> dict:
+        open_card: str | None = None, images: list | None = None) -> dict:
     """The owner's words, answered briefly - about the item on the table when there is one. The MODEL interprets
     them (PW-121): a question is answered, a subject named is pulled in, and a decision becomes a PROPOSAL the
     owner confirms (PW-123/124) - except Next, which moves the walk and marks nothing, and a reply request, which
@@ -2765,10 +2781,13 @@ def say(store, text: str, key: str = None, llm=None, actor: str = 'owner', trace
                       # THE CARD BROWSED OPEN in the canvas (the canvas redesign, 2026-09-29): "this", "it", "set it up"
                       # mean that card - a connector, a settings group, a report - when no item is on the table
                       + (f"ON SCREEN NOW: {_cut(open_card, 300)} - 'this' means that card; its own operations set it up\n\n" if open_card and not item else '')
-                      + f"The owner says: {text}\nAnswer them, briefly. If a look-up would answer it, CALL it now instead of saying you will. "
+                      + f"The owner says: {text}\n"
+                      # A PICTURE WITH THE LINE (the owner, 2026-09-30): an API brain sees it, a CLI brain reads the file it names
+                      + ("ATTACHED IMAGES (read these files - they are what the owner is showing you)\n" + "\n".join(images) + "\n" if images else '')
+                      + f"Answer them, briefly. If a look-up would answer it, CALL it now instead of saying you will. "
                       + ('If this is a decision about the item on the table, CALL it (bucket table).' if item else
                          'If they ask for something to be done, CALL it now - the card is their confirmation.'),
-                      max_tokens=MAX_TOKENS) or '').strip()
+                      max_tokens=MAX_TOKENS, **({'images': general._images(images)} if images else {})) or '').strip()
         raw, call = parse_call(raw)
         # A LOOK-UP runs at once and comes straight back, because it changes nothing and waits for
         # nobody. The model then answers with what it read - one round only, so a question can never

{taskuary-0.3.7.0 → taskuary-0.3.7.2}/taskuary/demo.py

@@ -47,6 +47,7 @@ ALLOWED_WRITES = (
     r'^/api/tasks$',                                   # make a task
     r'^/api/tasks/\d+$',                               # ...and change it
     r'^/api/tasks/\d+/(comments|waitroom|not-a-task|assistant/(session|messages|stream|cancel))$',
+    r'^/api/prompt-image$',                            # a picture for a prompt: saved in the demo's own home
     r'^/api/messages/\d+/(file|promote|read)$',        # the triage verdicts: the whole funnel
     r'^/api/reviews?/\d+/(hold|drop|edit)$',           # ...but never /send
     r'^/api/board/notes',                              # the wall

{taskuary-0.3.7.0 → taskuary-0.3.7.2}/taskuary/doorway_browse.py

@@ -62,14 +62,31 @@ def _connected(store) -> dict:
     return out
 
 
-def browse(store, area: str, section: str = None, open_: str = None) -> tuple[str, list]:
+POLL_MAX, PAGE = 12, 10          # WhatsApp's poll holds twelve (whatsapp/poll.mjs cuts the rest SILENTLY); a page leaves room
+
+
+def paged(rows: list, up: list, act: dict, page: int = 0) -> list:
+    """The picks for one page: ten rows, More when there are more, then the way back - never past twelve. The Connections
+    menu offered seventeen and a list of cards twelve plus Back, so the last five sections and Back could not be tapped
+    (2026-09-30 phone test)."""
+    if len(rows) + len(up) <= POLL_MAX: return rows + up
+    page = max(0, int(page or 0))
+    here = rows[page * PAGE:(page + 1) * PAGE]
+    more = [(f'More ({len(rows) - (page + 1) * PAGE} left)', {**act, 'page': page + 1})] if (page + 1) * PAGE < len(rows) else \
+        [('From the top', {**act, 'page': 0})]
+    return here + more + up[:POLL_MAX - len(here) - 1]
+
+
+def browse(store, area: str, section: str = None, open_: str = None, page: int = 0) -> tuple[str, list]:
     """(what to say, the picks under it) for one browse step. Back is always a pick; so is the way up."""
-    if area == 'connections': return _connections(store, section, open_)
-    if area == 'settings': return _settings(store, section, open_)
+    if area == 'connections': return _connections(store, section, open_, page)
+    if area == 'settings': return _settings(store, section, open_, page)
+    if area == 'reports': return _reports(store, open_, page)
+    if area == 'hub': return _hub(store, open_, page)
     return f'I cannot browse {area} here yet - it is on the desktop.', []
 
 
-def _connections(store, section, open_):
+def _connections(store, section, open_, page=0):
     from . import connectorcatalog
     cards = [c for c in connectorcatalog.cards() if c.get('group')]
     state = _connected(store)
@@ -87,18 +104,62 @@ def _connections(store, section, open_):
         rows = [c for c in cards if c['group'] == section]
         lines = [f"· {c['title']} - {state.get(c['type']) or ('planned' if c.get('planned') else 'not connected')}" for c in rows]
         picks = [(c['title'], {'t': 'browse', 'area': 'connections', 'section': section, 'open': c['type']}) for c in rows if not c.get('planned')]
-        return f"CONNECTIONS · {section.upper()} · {len(rows)}\n" + '\n'.join(lines), picks[:12] + up
+        return (f"CONNECTIONS · {section.upper()} · {len(rows)}\n" + '\n'.join(lines),
+                paged(picks, up, {'t': 'browse', 'area': 'connections', 'section': section}, page))
     on = sum(1 for v in state.values() if v == 'connected')
     lines = [f"{g} · {sum(1 for c in cards if c['group'] == g)}" for g in groups]
     return (f'Connections - {on} connected. Pick a section:\n' + '\n'.join(f'· {x}' for x in lines),
-            [(g, {'t': 'browse', 'area': 'connections', 'section': g}) for g in groups])
+            paged([(g, {'t': 'browse', 'area': 'connections', 'section': g}) for g in groups], [], {'t': 'browse', 'area': 'connections'}, page))
+
+
+def _reports(store, open_, page=0):
+    """REPORTS, the desktop sidebar's list as picks (the owner, 2026-09-30): each one's clock and how its last run went, then
+    one opened - what it is, where it goes, what the last run said. Running or changing one is words: the model proposes it."""
+    from . import appfacts
+    rows = appfacts.reports(store)
+    up = [('Back to Reports', {'t': 'browse', 'area': 'reports'})]
+    def last(r): return 'never ran' if r['last_ok'] is None else 'last run FAILED' if r['last_ok'] is False else f"ran {r['last_at'][:16]}"
+    if open_:
+        r = next((x for x in rows if str(x['source_id']) == str(open_)), None)
+        if not r: return 'That report is not there any more.', up
+        said = str(r['last_said'] or '').strip()
+        text = (f"{'WORKFLOW' if r['workflow'] else 'REPORT'} · {r['title']}{'' if r['active'] else ' (off)'}\n"
+                f"When: {r['schedule'] or 'no clock - it runs when asked'}\nReaches you: {r['reach']}\nGoes: {r['goes']}\n{last(r).capitalize()}"
+                + (f"\n> {said}" if said else '')
+                + '\n\nSay "run it now", or what to change, and I propose it for your yes.')
+        return text, up
+    if not rows: return 'No reports or workflows are set up yet. Say "set up a report" and I walk you through one.', []
+    lines = [f"· {r['title']}{'' if r['active'] else ' (off)'} - {r['schedule'] or 'no clock'}; {last(r)}" for r in rows]
+    return (f'REPORTS · {len(rows)}\n' + '\n'.join(lines),
+            paged([(r['title'], {'t': 'browse', 'area': 'reports', 'open': r['source_id']}) for r in rows], [], {'t': 'browse', 'area': 'reports'}, page))
+
+
+def _hub(store, open_, page=0):
+    """THE HUB, newest first as the desktop lists it (the owner, 2026-09-30): what the agents worked out, each one openable."""
+    up = [('Back to the Hub', {'t': 'browse', 'area': 'hub'})]
+    if open_:
+        e = store.lore_get(int(open_))
+        if not e: return 'That entry is not on the Hub any more.', up
+        body = str(e.get('Body') or '').strip()
+        text = (f"HUB · {e.get('Topic') or 'general'}\n*{e.get('Title') or ''}*\n"
+                f"by {e.get('Author') or 'an agent'} · score {e.get('Score') or 0} · {len(store.lore_comments(int(open_)) or [])} comments\n\n"
+                + (body[:1500] + ('…' if len(body) > 1500 else '')))
+        return text, up
+    posts = store.lore_posts(None, None, 40, 'new', 'live', None)
+    if not posts: return 'Nothing on the Hub yet - agents post what they work out as they go.', []
+    lines = [f"· {p.get('Title') or ''} ({p.get('Topic') or 'general'}, {p.get('Score') or 0})" for p in posts[:20]]
+    n = store.lore_count()
+    n = n.get('posts', len(posts)) if isinstance(n, dict) else n
+    return (f"HUB · {n} entries, newest first\n" + '\n'.join(lines),
+            paged([((p.get('Title') or '')[:90], {'t': 'browse', 'area': 'hub', 'open': p['LoreId']}) for p in posts[:20]], [],
+                  {'t': 'browse', 'area': 'hub'}, page))
 
 
 def _schema() -> dict:
     return json.loads((Path(__file__).with_name('settings_schema.json')).read_text(encoding='utf-8'))
 
 
-def _settings(store, section, open_):
+def _settings(store, section, open_, page=0):
     schema, st = _schema(), store.get_settings()
     knobs = schema.get('knobs') or {}
     up = [('Back to Settings', {'t': 'browse', 'area': 'settings'})]
@@ -106,6 +167,7 @@ def _settings(store, section, open_):
         v = st.get(k)
         kind = (knobs.get(k) or {}).get('type')
         if kind == 'switch': return 'on' if str(v if v is not None else '1').strip() not in ('0', 'false', 'off', '') else 'off'
+        if str(v).strip() == '*': return 'every other one set up, in order'      # backup_brains' own wildcard, said in words
         return str(v) if v not in (None, '') else '(not set)'
     if open_:
         m = knobs.get(open_) or {'label': open_}
@@ -116,7 +178,8 @@ def _settings(store, section, open_):
         rows = [(k, m) for k, m in knobs.items() if m.get('group') == section]
         lines = [f"· {m.get('label') or k}: {value(k)}" for k, m in rows]
         return (f"SETTINGS · {section.upper()}\n" + ('\n'.join(lines) or 'Its controls are on the desktop.'),
-                [(m.get('label') or k, {'t': 'browse', 'area': 'settings', 'section': section, 'open': k}) for k, m in rows][:12] + up)
+                paged([(m.get('label') or k, {'t': 'browse', 'area': 'settings', 'section': section, 'open': k}) for k, m in rows], up,
+                      {'t': 'browse', 'area': 'settings', 'section': section}, page))
     groups = [g for g in schema.get('groups') or [] if any(m.get('group') == g for m in knobs.values())]
     return ('Settings - pick a section:\n' + '\n'.join(f'· {g}' for g in groups),
-            [(g, {'t': 'browse', 'area': 'settings', 'section': g}) for g in groups])
+            paged([(g, {'t': 'browse', 'area': 'settings', 'section': g}) for g in groups], [], {'t': 'browse', 'area': 'settings'}, page))