@gotcos/glasses-server 6.55.0 → 6.56.1

package/work-task-runtime/task_rows.py

@@ -0,0 +1,420 @@
+#!/usr/bin/env python3
+"""Canonical tasks.md row parser (plan v19 §0.2 / §1)."""
+
+from __future__ import annotations
+from runtime_io import safe_read_text
+
+import hashlib
+import json
+import os
+import re
+import sys
+from dataclasses import asdict, dataclass, field
+from pathlib import Path
+from typing import Iterator, Optional
+
+SCRIPT_DIR = Path(__file__).parent.resolve()
+OPERATIONS_DIR = SCRIPT_DIR.parent
+
+# Domains are DISCOVERED, not baked in. This was a hardcoded four-tuple of one
+# user's business units, and because every gate below tests membership in it, the
+# Python layer refused any other domain outright — including domains the server
+# had just resolved from the same folders. `cos_api_bridge.py task-rows work`
+# answered `invalid_domain` while `quilt` returned 156 rows.
+#
+# The signal is `tasks.md`, matching discoveredTaskDomains() on the server side,
+# so the two layers cannot disagree about which domains exist.
+def full_domains() -> tuple[str, ...]:
+    root = Path(os.environ.get("COS_TASK_ROOT") or OPERATIONS_DIR)
+    try:
+        found = sorted(
+            d.name for d in root.iterdir()
+            if d.is_dir() and not d.name.startswith(".") and (d / "tasks.md").is_file()
+        )
+    except OSError:
+        found = []
+    # A tree we cannot read must not silently become "no domains"; the legacy
+    # four keep an existing install working exactly as before.
+    return tuple(found) if found else LEGACY_DOMAINS
+
+
+LEGACY_DOMAINS = ("quilt", "personal", "hermit_crabs", "sprocket_rocket")
+
+# Convenience aliases only. `SHORT_TO_FULL.get(name, name)` passes anything
+# unknown through untouched, so these never gate which domains exist.
+SHORT_TO_FULL = {
+    "sr": "sprocket_rocket",
+    "hc": "hermit_crabs",
+}
+FULL_TO_SHORT = {
+    "quilt": "quilt",
+    "personal": "personal",
+    "sprocket_rocket": "sr",
+    "hermit_crabs": "hc",
+}
+
+# ONE marker vocabulary, two uses. MARKER_RE strips a trailing marker blob off
+# the end of a line; MARKER_ANYWHERE_RE finds a marker wherever it sits. The
+# writers use the second to REFUSE user text that would be re-read as a marker.
+# They are built from the same string so a new marker type cannot be taught to
+# the reader and forgotten by the writer.
+_MARKER_BODY = (
+    r"\[(?:run \d{4}-\d{2}-\d{2} \d{2}:\d{2}|agent running|"
+    r"agent #\d+ (?:done|failed)|agent failed|"
+    r"stage (?:planning|active|review))\]"
+)
+MARKER_RE = re.compile(r"\s*(?:" + _MARKER_BODY + r"\s*)+$")
+MARKER_ANYWHERE_RE = re.compile(_MARKER_BODY)
+# Workflow stage. An UNMARKED row is planning, so the 80 open rows need no
+# migration. Lives in the line, not a sidecar: the task id is a hash of the
+# description, so any id-keyed store orphans itself on the first text edit.
+# NOTE: the trailing \s* sits INSIDE the repeat group. The writer joins markers
+# with spaces (_rebuild_line), but the old regex allowed none between groups, so
+# "[run ...] [agent running]" stripped only the last one and silently dropped the
+# run time on every read. Pre-existing; stage makes that pairing the common case.
+# STAGE_RE only reads a value out of a blob MARKER_RE has already validated, so
+# its own alternation was a second, unreachable copy: mutating it changed nothing
+# under test BECAUSE MARKER_RE's `$`-anchored match rejects "[stage bogus]" first.
+# That is true of the READ path only. The WRITE path is gated separately by
+# STAGES below (task_write.set_stage) and by cos_api_bridge's own literal tuple,
+# so there are three stage vocabularies in this system, not one. Widening the set
+# means editing MARKER_RE, STAGES, and the bridge tuple together.
+STAGE_RE = re.compile(r"\[stage (\w+)\]")
+STAGES = ("planning", "active", "review")
+RUN_AT_RE = re.compile(r"\[run (\d{4}-\d{2}-\d{2} \d{2}:\d{2})\]")
+AGENT_NUM_RE = re.compile(r"\[agent #(\d+) (done|failed)\]")
+INBOX_RE = re.compile(r"^#{1,3}\s*INBOX\s*$", re.IGNORECASE)
+TODAY_RE = re.compile(r"^#{1,3}\s*TODAY\s+(\d{4}-\d{2}-\d{2})\s*$", re.IGNORECASE)
+DELEGATED_HEADING_RE = re.compile(r"^#{1,2}\s*Delegated\b", re.IGNORECASE)
+BOLD_OWNER_RE = re.compile(r"^\*\*(?P<body>(?!Owner:)[^*]+)\*\*:\s*")
+BRACKET_OWNER_RE = re.compile(r"^\[(?P<body>[^\]]+)\]\s+")
+CHECK_RE = re.compile(r"^- \[([ xX~])\]\s*(.*)$")
+PRIORITY_SECTIONS = frozenset(
+    {"inbox", "today", "urgent", "this_week", "backlog", "completed"}
+)
+
+
+@dataclass
+class TaskRow:
+    ref: str
+    id: str
+    domain: str
+    description: str
+    priority: str
+    is_checked: bool
+    archived: bool
+    line_number: int
+    source: Optional[str]
+    owner: Optional[str]
+    delegated: bool
+    needs_review: bool
+    thread: None
+    run_at: Optional[str]
+    agent_state: Optional[str]
+    agent_no: Optional[int]
+    section: str
+    section_day: Optional[str]
+    stage: Optional[str] = None
+    done_when: Optional[str] = None
+    work_stage: str = "planned"
+    work_identity: str = ""
+    work_revision: str = ""
+    meeting_refs: list[dict] = field(default_factory=list)
+    work_metadata_error: Optional[str] = None
+
+
+def domain_path(domain: str) -> Path:
+    root = Path(os.environ.get("COS_TASK_ROOT") or OPERATIONS_DIR)
+    return root / domain / "tasks.md"
+
+
+def split_source(line: str) -> tuple[str, Optional[str]]:
+    if "**Source:**" not in line:
+        return line.strip(), None
+    left, right = line.split("**Source:**", 1)
+    return left.strip().rstrip("—").strip(), right.strip()
+
+
+DONE_WHEN_LABEL = "**Done when:**"
+
+
+def split_done_when(text):
+    """Peel a **Done when:** segment off a description or source part.
+
+    Same convention as **Source:** so tasks.md stays readable by hand, and the
+    finish line travels with the row when the text is edited.
+    """
+    if not text or DONE_WHEN_LABEL not in text:
+        return text, None
+    left, right = text.split(DONE_WHEN_LABEL, 1)
+    return left.strip().rstrip("\u2014").strip(), (right.strip() or None)
+
+
+def strip_markers(text: str) -> tuple[str, dict]:
+    """Remove end-of-line agent/run markers. Returns (text, fields)."""
+    fields: dict = {"run_at": None, "agent_state": None, "agent_no": None, "stage": None}
+    if not text:
+        return text, fields
+    match = MARKER_RE.search(text)
+    if not match:
+        return text, fields
+    blob = match.group(0)
+    run = RUN_AT_RE.search(blob)
+    if run:
+        fields["run_at"] = run.group(1)
+    stage = STAGE_RE.search(blob)
+    if stage:
+        fields["stage"] = stage.group(1)
+    numbered = AGENT_NUM_RE.search(blob)
+    if numbered:
+        fields["agent_no"] = int(numbered.group(1))
+        fields["agent_state"] = numbered.group(2)
+    elif "[agent running]" in blob:
+        fields["agent_state"] = "running"
+    elif "[agent failed]" in blob:
+        fields["agent_state"] = "failed"
+    return text[: match.start()].rstrip(), fields
+
+
+def _merge_marker_fields(primary: dict, secondary: dict) -> dict:
+    out = dict(primary)
+    for key, value in secondary.items():
+        if value is not None:
+            out[key] = value
+    return out
+
+
+def classify_heading(stripped: str) -> tuple[Optional[str], Optional[str]]:
+    """Return (section, section_day) for a heading, else (None, None)."""
+    if not stripped.startswith("#"):
+        return None, None
+    inbox = INBOX_RE.match(stripped)
+    if inbox:
+        return "inbox", None
+    today = TODAY_RE.match(stripped)
+    if today:
+        return "today", today.group(1)
+    upper = stripped.upper()
+    if "🔴 URGENT" in stripped or "URGENT" in upper:
+        return "urgent", None
+    if "🟡 THIS WEEK" in stripped or "THIS WEEK" in upper:
+        return "this_week", None
+    if "🟢 BACKLOG" in stripped or "BACKLOG" in upper:
+        return "backlog", None
+    if "✅ COMPLETED" in stripped or "COMPLETED" in upper:
+        return "completed", None
+    return "other", None
+
+
+def _heading_depth(stripped: str) -> int:
+    i = 0
+    while i < len(stripped) and stripped[i] == "#":
+        i += 1
+    return i
+
+
+def parse_owner(text: str) -> tuple[Optional[str], str]:
+    bold = BOLD_OWNER_RE.match(text)
+    if bold:
+        return bold.group("body").strip(), text[bold.end():]
+    bracket = BRACKET_OWNER_RE.match(text)
+    if bracket:
+        return bracket.group("body").strip(), text[bracket.end():]
+    return None, text
+
+
+def owner_tokens(owner: str) -> list[str]:
+    return [part.strip() for part in re.split(r"\s*\+\s*|/", owner) if part.strip()]
+
+
+def owner_includes_miles(owner: Optional[str]) -> bool:
+    if not owner:
+        return False
+    for token in owner_tokens(owner):
+        head = token.split("(")[0].strip().split()[0].casefold()
+        if head == os.environ.get("COS_TASK_OPERATOR", "").casefold():
+            return True
+    return False
+
+
+def ends_delegated_zone(stripped: str) -> bool:
+    if stripped.startswith("---"):
+        return True
+    return stripped.startswith("#") and _heading_depth(stripped) <= 2
+
+
+def parse_path(filepath: Path, domain: str) -> list[TaskRow]:
+    # No membership gate here on purpose. The caller has already chosen the FILE;
+    # re-deciding whether its domain "exists" by scanning a root that may not be
+    # the one this path came from turns a valid parse into silence. Discovery
+    # belongs where a NAME is resolved to a path, which is parse_domain and the
+    # cmd_* entry points below.
+    if not filepath.exists():
+        return []
+
+    rows: list[TaskRow] = []
+    current_priority = "backlog"
+    current_section = "other"
+    current_day: Optional[str] = None
+    in_comment = False
+    delegated_zone = False
+    counter = 0
+
+    contents = safe_read_text(filepath)
+    for line_number, line in enumerate(contents.splitlines(), 1):
+        stripped = line.strip()
+        if "<!--" in stripped:
+            in_comment = True
+        if "-->" in stripped:
+            in_comment = False
+            continue
+        if in_comment:
+            continue
+
+        if ends_delegated_zone(stripped):
+            delegated_zone = False
+        if DELEGATED_HEADING_RE.match(stripped):
+            delegated_zone = True
+
+        heading, heading_day = classify_heading(stripped)
+        if heading is not None:
+            current_section = heading
+            if heading != "other":
+                current_priority = heading
+                current_day = heading_day
+            else:
+                current_day = None
+            continue
+
+        checked = CHECK_RE.match(line)
+        if not checked:
+            continue
+
+        mark, raw = checked.group(1), checked.group(2)
+        archived = mark == "~"
+        is_checked = mark.lower() == "x"
+        description, source = split_source(raw)
+        description, desc_markers = strip_markers(description)
+        if source is not None:
+            source, src_markers = strip_markers(source)
+            markers = _merge_marker_fields(desc_markers, src_markers)
+        else:
+            markers = desc_markers
+
+        description, dw_desc = split_done_when(description)
+        if source is not None:
+            source, dw_src = split_done_when(source)
+        else:
+            dw_src = None
+        done_when = dw_desc or dw_src
+
+        needs_review = "[REVIEW]" in description
+        if needs_review:
+            description = description.replace("[REVIEW]", "").strip()
+
+        owner, description = parse_owner(description)
+        description = description.strip()
+        delegated = delegated_zone or (
+            owner is not None and not owner_includes_miles(owner)
+        )
+        priority = "completed" if is_checked else (
+            current_section if current_section in PRIORITY_SECTIONS
+            else current_priority
+        )
+        counter += 1
+        rows.append(TaskRow(
+            ref=f"{domain}-{counter}",
+            id="",
+            domain=domain,
+            description=description,
+            priority=priority,
+            is_checked=is_checked,
+            archived=archived,
+            line_number=line_number,
+            source=source,
+            owner=owner,
+            delegated=delegated,
+            needs_review=needs_review,
+            thread=None,
+            run_at=markers["run_at"],
+            agent_state=markers["agent_state"],
+            agent_no=markers["agent_no"],
+            section=current_section,
+            section_day=current_day if current_section == "today" else None,
+            stage=markers["stage"],
+            done_when=done_when,
+        ))
+
+    _assign_ids(filepath, domain, rows)
+    from task_work_metadata import decode_source, phase_for
+    identities: dict[str, list[TaskRow]] = {}
+    for row in rows:
+        row.source, metadata, row.work_metadata_error = decode_source(row.source)
+        row.work_stage = phase_for(row.stage, row.is_checked, metadata)
+        row.work_identity = metadata.get("workIdentity", row.id)
+        row.work_revision = hashlib.sha256((domain + "\0" + row.id + "\0" + contents).encode()).hexdigest()
+        row.meeting_refs = metadata.get("meetingRefs", [])
+        identities.setdefault(row.work_identity, []).append(row)
+    for matches in identities.values():
+        if len(matches) > 1:
+            for row in matches:
+                row.work_metadata_error = "Duplicate Work identity; repair the copied task metadata before linking or moving it."
+                row.work_identity = row.id
+    return rows
+
+
+def _assign_ids(_filepath: Path, domain: str, rows: list[TaskRow]) -> None:
+    from task_dedup import TaskDeduplicator
+
+    seen: dict[str, int] = {}
+    for row in rows:
+        norm = TaskDeduplicator.normalize(row.description)
+        ordinal = seen.get(norm, 0)
+        seen[norm] = ordinal + 1
+        row.id = hashlib.sha256(
+            f"{domain}/tasks.md|{norm}|{ordinal}".encode()
+        ).hexdigest()[:12]
+
+
+def parse_domain(domain: str) -> list[TaskRow]:
+    if domain not in full_domains():
+        return []
+    return parse_path(domain_path(domain), domain)
+
+
+def iter_domain_rows() -> Iterator[tuple[str, list[TaskRow]]]:
+    for domain in full_domains():
+        yield domain, parse_domain(domain)
+
+
+def cmd_task_rows(domain: str, day: str) -> list[dict]:
+    del day  # server column math; required on the wire
+    return [asdict(row) for row in parse_domain(domain)]
+
+
+def _cli(argv: list[str]) -> int:
+    if len(argv) < 2 or argv[0] in {"-h", "--help"}:
+        print("usage: task_rows.py <domain> --day YYYY-MM-DD", file=sys.stderr)
+        return 2
+    domain = argv[0]
+    day = None
+    if "--day" in argv:
+        idx = argv.index("--day")
+        if idx + 1 < len(argv):
+            day = argv[idx + 1]
+    if domain not in full_domains():
+        print(json.dumps({
+            "error": {"code": "invalid_domain", "message": f"full domain required, got {domain!r}"}
+        }))
+        return 0
+    if not day or not re.fullmatch(r"\d{4}-\d{2}-\d{2}", day):
+        print(json.dumps({
+            "error": {"code": "day_required", "message": "--day YYYY-MM-DD is required"}
+        }))
+        return 0
+    print(json.dumps(cmd_task_rows(domain, day)))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(_cli(sys.argv[1:]))

package/work-task-runtime/task_work_metadata.py

@@ -0,0 +1,107 @@
+"""Bounded Work metadata inside Source; legacy task marker grammar stays unchanged."""
+from __future__ import annotations
+
+import base64
+import json
+import re
+
+PHASES = ("mentioned", "planned", "draft", "built", "qa")
+LEGACY_STAGE = {"mentioned": "planning", "planned": "planning", "draft": "active", "built": "active", "qa": "review"}
+DEFAULT_PHASE = {"planning": "planned", "active": "draft", "review": "qa"}
+PROTOCOL = "cos-work://"
+MAX_BYTES = 16_384
+MAX_REFS = 8
+LINK = re.compile(r"\[Work details\]\(cos-work://v1/([A-Za-z0-9_-]+)\)")
+
+
+class WorkMetadataError(ValueError):
+    def __init__(self, code: str, message: str):
+        super().__init__(message)
+        self.code = code
+
+
+def reject_reserved(value: str) -> None:
+    if PROTOCOL in value.lower():
+        raise WorkMetadataError("reserved_work_metadata", "Work metadata links may only be written through Work controls.")
+
+
+def validate_meeting(raw) -> dict:
+    if not isinstance(raw, dict) or set(raw) != {"recordId", "domain", "month", "filename", "title"}:
+        raise WorkMetadataError("invalid_meeting_ref", "A complete saved meeting reference is required.")
+    limits = {"recordId": 512, "domain": 64, "month": 7, "filename": 255, "title": 300}
+    if any(not isinstance(raw.get(k), str) or not raw[k].strip() or len(raw[k]) > n or any(ord(c) < 32 or ord(c) == 127 for c in raw[k]) for k, n in limits.items()):
+        raise WorkMetadataError("invalid_meeting_ref", "Meeting reference fields are invalid or too long.")
+    domain = raw["domain"]
+    if domain != domain.strip() or domain.startswith('.') or '/' in domain or '\\' in domain or not re.fullmatch(r"\d{4}-(0[1-9]|1[0-2])", raw["month"]):
+        raise WorkMetadataError("invalid_meeting_ref", "Meeting domain or month is invalid.")
+    if "/" in raw["filename"] or "\\" in raw["filename"] or not raw["filename"].endswith(".md") or raw["filename"] in (".md", "..md"):
+        raise WorkMetadataError("invalid_meeting_ref", "Meeting filename must be a saved Markdown basename.")
+    return dict(raw)
+
+
+def validate_metadata(raw) -> dict:
+    if not isinstance(raw, dict) or set(raw) - {"workflowPhase", "workIdentity", "meetingRefs"}:
+        raise ValueError("Unknown metadata fields")
+    if "workflowPhase" in raw and raw["workflowPhase"] not in PHASES:
+        raise ValueError("Invalid workflow phase")
+    if "workIdentity" in raw and (not isinstance(raw["workIdentity"], str) or not re.fullmatch(r"[a-f0-9]{12}", raw["workIdentity"])):
+        raise ValueError("Invalid work identity")
+    refs = raw.get("meetingRefs", [])
+    if not isinstance(refs, list) or len(refs) > MAX_REFS:
+        raise ValueError("Too many meeting references")
+    cleaned = [validate_meeting(ref) for ref in refs]
+    if len({ref["recordId"] for ref in cleaned}) != len(cleaned):
+        raise ValueError("Duplicate meeting references")
+    return {**raw, **({"meetingRefs": cleaned} if "meetingRefs" in raw else {})}
+
+
+def _unique_object(pairs):
+    out = {}
+    for key, value in pairs:
+        if key in out:
+            raise ValueError("Duplicate JSON key")
+        out[key] = value
+    return out
+
+
+def decode_source(source: str | None) -> tuple[str | None, dict, str | None]:
+    """Invalid metadata stays visible/preserved and never contributes authority."""
+    if not source or PROTOCOL not in source.lower():
+        return source, {}, None
+    try:
+        matches = list(LINK.finditer(source))
+        if len(matches) != 1 or source.lower().count(PROTOCOL) != 1:
+            raise ValueError("Missing, duplicate, or unsupported metadata link")
+        token = matches[0].group(1)
+        if len(token) > (MAX_BYTES * 4 // 3 + 4):
+            raise ValueError("Metadata too large")
+        data = base64.b64decode(token + '=' * (-len(token) % 4), altchars=b'-_', validate=True)
+        if len(data) > MAX_BYTES or base64.urlsafe_b64encode(data).decode().rstrip('=') != token:
+            raise ValueError("Noncanonical encoding")
+        metadata = validate_metadata(json.loads(data, object_pairs_hook=_unique_object))
+        clean = (source[:matches[0].start()] + source[matches[0].end():]).strip()
+        return clean or None, metadata, None
+    except (ValueError, TypeError, UnicodeError):
+        return source, {}, "Invalid Work metadata; preserved without applying its phase or meeting links."
+
+
+def write_source(source: str | None, metadata: dict) -> str:
+    clean, _, error = decode_source(source)
+    if error:
+        raise WorkMetadataError("work_metadata_invalid", error)
+    try:
+        data = json.dumps(validate_metadata(metadata), ensure_ascii=False, sort_keys=True, separators=(',', ':')).encode()
+        if len(data) > MAX_BYTES:
+            raise ValueError("Metadata too large")
+    except (ValueError, TypeError) as exc:
+        raise WorkMetadataError("work_metadata_invalid", "Work metadata exceeds its schema or size limits.") from exc
+    token = base64.urlsafe_b64encode(data).decode().rstrip('=')
+    return ((clean + ' ') if clean else '') + '[Work details](cos-work://v1/' + token + ')'
+
+
+def phase_for(stage: str | None, checked: bool, metadata: dict) -> str:
+    if checked:
+        return 'complete'
+    legacy = stage or 'planning'
+    phase = metadata.get('workflowPhase')
+    return phase if LEGACY_STAGE.get(phase) == legacy else DEFAULT_PHASE.get(legacy, 'planned')