next_extract.py
python
sha256:0e9549ec7b463911bc08b7d586dc320b1ac9b1f5c943ee7e3865dcc6cb0f6f83
chore(governance): sync handover+roadmap to 84db8c8 (drift:…
Human
1 day ago
| 1 | """Parse PRIMARY / RELAY / PRODUCT RELAY / ARCHIVED / LANE TIP markers (§MR.6).""" |
| 2 | |
| 3 | from __future__ import annotations |
| 4 | |
| 5 | import hashlib |
| 6 | import re |
| 7 | from dataclasses import replace |
| 8 | |
| 9 | from tools.workspace.types import NextBlock, NextRole |
| 10 | |
| 11 | _MARKER_RE = re.compile( |
| 12 | r"<!--\s*overseer:next\s+" |
| 13 | r"role=(?P<role>primary|relay|product_relay|lane_tip|archived)" |
| 14 | r"(?:\s+lane=(?P<lane>[^\s>]+))?" |
| 15 | r"(?:\s+status=(?P<status>live|archived))?" |
| 16 | r"(?:\s+product_order=(?P<product_order>[^\s>]+))?" |
| 17 | r"(?:\s+tip_hash=sha256:(?P<tip_hash>[0-9a-fA-F]{64}))?" |
| 18 | r"\s*-->", |
| 19 | re.IGNORECASE, |
| 20 | ) |
| 21 | |
| 22 | _HEADING_NEXT = re.compile(r"^##\s+NEXT SESSION\s+[—\-]\s+(?P<title>.+?)\s*$") |
| 23 | _HEADING_PRODUCT_RELAY = re.compile(r"^##\s+PRODUCT RELAY\s+[—\-]\s+(?P<title>.+?)\s*$") |
| 24 | _HEADING_LANE_TIP = re.compile(r"^##\s+LANE TIP\s+[—\-]\s+(?P<title>.+?)\s*$") |
| 25 | _HEADING_ARCHIVED = re.compile(r"^##\s+ARCHIVED SESSION\s+[—\-]\s+(?P<title>.+?)\s*$") |
| 26 | |
| 27 | _FENCE_RE = re.compile(r"```(?:[^\n]*)\n(.*?)```", re.DOTALL) |
| 28 | _STEP_RE = re.compile(r"(?im)^\s*Step:\s*(?P<v>.+?)\s*$") |
| 29 | _MODEL_RE = re.compile(r"(?im)^\s*Model:\s*(?P<v>.+?)\s*$") |
| 30 | _AUTHORITY_RE = re.compile(r"(?im)^\s*Authority:\s*(?P<v>.+?)\s*$") |
| 31 | _ID_ROW_RE = re.compile( |
| 32 | r"(?im)^\|\s*\*?\*?ID\*?\*?\s*\|\s*\*?\*?(?P<id>[^*|\s][^*|]*)\*?\*?\s*\|" |
| 33 | ) |
| 34 | |
| 35 | |
| 36 | def lf_normalize(text: str) -> str: |
| 37 | """Normalize newlines to LF for stable hashing.""" |
| 38 | return text.replace("\r\n", "\n").replace("\r", "\n") |
| 39 | |
| 40 | |
| 41 | def tip_hash_hex(fence_bytes_text: str) -> str: |
| 42 | """SHA-256 hex of LF-normalized UTF-8 paste-fence body (§MR.5.1).""" |
| 43 | normalized = lf_normalize(fence_bytes_text).encode("utf-8") |
| 44 | return hashlib.sha256(normalized).hexdigest() |
| 45 | |
| 46 | |
| 47 | def normalize_model_display(label: str) -> str | None: |
| 48 | """Map fence Model line to canonical display label from policy.""" |
| 49 | cleaned = label.strip().strip("*").strip() |
| 50 | if not cleaned: |
| 51 | return None |
| 52 | mapping = { |
| 53 | "thinking": "Thinking", |
| 54 | "auto": "Auto", |
| 55 | "thinking → auto": "Thinking → Auto", |
| 56 | "thinking->auto": "Thinking → Auto", |
| 57 | "thinking_to_auto": "Thinking → Auto", |
| 58 | "operator + auto": "Operator + Auto", |
| 59 | "operator_plus_auto": "Operator + Auto", |
| 60 | "operator+auto": "Operator + Auto", |
| 61 | } |
| 62 | lowered = cleaned.lower().replace("—", "→").replace("–", "→") |
| 63 | lowered = re.sub(r"\s+", " ", lowered) |
| 64 | if cleaned in {"Thinking", "Auto", "Thinking → Auto", "Operator + Auto"}: |
| 65 | return cleaned |
| 66 | return mapping.get(lowered) |
| 67 | |
| 68 | |
| 69 | def _extract_fence_fields(fence: str | None) -> tuple[str | None, str | None, str | None]: |
| 70 | if not fence: |
| 71 | return None, None, None |
| 72 | step_m = _STEP_RE.search(fence) |
| 73 | model_m = _MODEL_RE.search(fence) |
| 74 | auth_m = _AUTHORITY_RE.search(fence) |
| 75 | step = step_m.group("v").strip() if step_m else None |
| 76 | model_raw = model_m.group("v").strip() if model_m else None |
| 77 | model = normalize_model_display(model_raw) if model_raw else None |
| 78 | authority = auth_m.group("v").strip().lower() if auth_m else None |
| 79 | return step, model, authority |
| 80 | |
| 81 | |
| 82 | def _step_from_body(body: str, fence_step: str | None) -> str | None: |
| 83 | if fence_step: |
| 84 | return fence_step |
| 85 | id_m = _ID_ROW_RE.search(body) |
| 86 | if id_m: |
| 87 | return id_m.group("id").strip() |
| 88 | return None |
| 89 | |
| 90 | |
| 91 | def _forbidden_archived_next_title(title: str) -> bool: |
| 92 | return "archived" in title.lower() |
| 93 | |
| 94 | |
| 95 | def _ambiguous_primary_phrase(title: str) -> bool: |
| 96 | lowered = title.lower() |
| 97 | return "primary relay" in lowered or "primary (relay)" in lowered |
| 98 | |
| 99 | |
| 100 | def extract_next_blocks(text: str) -> list[NextBlock]: |
| 101 | """Extract all marked (and legacy unmarked NEXT) blocks from handover text.""" |
| 102 | lines = lf_normalize(text).split("\n") |
| 103 | blocks: list[NextBlock] = [] |
| 104 | i = 0 |
| 105 | while i < len(lines): |
| 106 | line = lines[i] |
| 107 | marker_m = _MARKER_RE.search(line.strip()) |
| 108 | if marker_m and i + 1 < len(lines): |
| 109 | heading_line = i + 2 # 1-indexed heading |
| 110 | heading = lines[i + 1] |
| 111 | role = NextRole(marker_m.group("role").lower()) |
| 112 | lane = marker_m.group("lane") |
| 113 | status = (marker_m.group("status") or ("archived" if role is NextRole.ARCHIVED else "live")).lower() |
| 114 | product_order = marker_m.group("product_order") |
| 115 | tip_hash = marker_m.group("tip_hash") |
| 116 | if tip_hash: |
| 117 | tip_hash = tip_hash.lower() |
| 118 | |
| 119 | # Collect body until next marker or next major heading at ## level that starts a session block |
| 120 | j = i + 2 |
| 121 | body_lines: list[str] = [] |
| 122 | while j < len(lines): |
| 123 | peek = lines[j].strip() |
| 124 | if _MARKER_RE.search(peek): |
| 125 | break |
| 126 | if ( |
| 127 | _HEADING_NEXT.match(lines[j]) |
| 128 | or _HEADING_PRODUCT_RELAY.match(lines[j]) |
| 129 | or _HEADING_LANE_TIP.match(lines[j]) |
| 130 | or _HEADING_ARCHIVED.match(lines[j]) |
| 131 | ): |
| 132 | # Unmarked heading — stop before it so legacy scanner can see it |
| 133 | break |
| 134 | body_lines.append(lines[j]) |
| 135 | j += 1 |
| 136 | body = "\n".join(body_lines) |
| 137 | fence_m = _FENCE_RE.search(body) |
| 138 | fence = fence_m.group(1) if fence_m else None |
| 139 | step, model, authority = _extract_fence_fields(fence) |
| 140 | step = _step_from_body(body, step) |
| 141 | blocks.append( |
| 142 | NextBlock( |
| 143 | role=role, |
| 144 | lane=lane, |
| 145 | status=status, |
| 146 | product_order=product_order, |
| 147 | tip_hash=tip_hash, |
| 148 | heading=heading.strip(), |
| 149 | heading_line=heading_line, |
| 150 | body=body, |
| 151 | fence=fence, |
| 152 | step_id=step, |
| 153 | model=model, |
| 154 | authority=authority, |
| 155 | unmarked=False, |
| 156 | ) |
| 157 | ) |
| 158 | i = j |
| 159 | continue |
| 160 | |
| 161 | # Legacy / unmarked NEXT SESSION headings |
| 162 | next_m = _HEADING_NEXT.match(line) |
| 163 | if next_m: |
| 164 | # Skip if previous non-empty line was a marker (already consumed) |
| 165 | title = next_m.group("title") |
| 166 | j = i + 1 |
| 167 | body_lines = [] |
| 168 | while j < len(lines): |
| 169 | peek = lines[j].strip() |
| 170 | if _MARKER_RE.search(peek): |
| 171 | break |
| 172 | if ( |
| 173 | _HEADING_NEXT.match(lines[j]) |
| 174 | or _HEADING_PRODUCT_RELAY.match(lines[j]) |
| 175 | or _HEADING_LANE_TIP.match(lines[j]) |
| 176 | or _HEADING_ARCHIVED.match(lines[j]) |
| 177 | ): |
| 178 | break |
| 179 | body_lines.append(lines[j]) |
| 180 | j += 1 |
| 181 | body = "\n".join(body_lines) |
| 182 | fence_m = _FENCE_RE.search(body) |
| 183 | fence = fence_m.group(1) if fence_m else None |
| 184 | step, model, authority = _extract_fence_fields(fence) |
| 185 | step = _step_from_body(body, step) |
| 186 | blocks.append( |
| 187 | NextBlock( |
| 188 | role=NextRole.PRIMARY, |
| 189 | lane=None, |
| 190 | status="live", |
| 191 | product_order=None, |
| 192 | tip_hash=None, |
| 193 | heading=line.strip(), |
| 194 | heading_line=i + 1, |
| 195 | body=body, |
| 196 | fence=fence, |
| 197 | step_id=step, |
| 198 | model=model, |
| 199 | authority=authority, |
| 200 | unmarked=True, |
| 201 | ) |
| 202 | ) |
| 203 | i = j |
| 204 | continue |
| 205 | i += 1 |
| 206 | return blocks |
| 207 | |
| 208 | |
| 209 | def legacy_forbidden_archived_headings(text: str) -> list[tuple[int, str]]: |
| 210 | """Return (line, heading) for ``## NEXT SESSION — … archived …`` (forbidden).""" |
| 211 | out: list[tuple[int, str]] = [] |
| 212 | for idx, line in enumerate(lf_normalize(text).split("\n"), start=1): |
| 213 | m = _HEADING_NEXT.match(line) |
| 214 | if m and _forbidden_archived_next_title(m.group("title")): |
| 215 | out.append((idx, line.strip())) |
| 216 | return out |
| 217 | |
| 218 | |
| 219 | def select_live_primary(blocks: list[NextBlock], *, lane: str = "product") -> NextBlock | None: |
| 220 | """Select LIVE PRIMARY for a lane (ignores archived).""" |
| 221 | live = [ |
| 222 | b |
| 223 | for b in blocks |
| 224 | if b.role is NextRole.PRIMARY |
| 225 | and b.status == "live" |
| 226 | and not b.unmarked |
| 227 | and (b.lane is None or b.lane == lane) |
| 228 | ] |
| 229 | if len(live) == 1: |
| 230 | return live[0] |
| 231 | if len(live) > 1: |
| 232 | return None # ambiguous — caller treats as ambiguous_primary |
| 233 | return None |
| 234 | |
| 235 | |
| 236 | def select_product_tip(blocks: list[NextBlock], *, lane: str = "product") -> tuple[NextBlock | None, str | None]: |
| 237 | """Select relay XOR product_relay tip for product lane. |
| 238 | |
| 239 | Returns ``(block, error_code)`` where error_code is ``ambiguous_primary`` when both present. |
| 240 | """ |
| 241 | relays = [ |
| 242 | b |
| 243 | for b in blocks |
| 244 | if b.role is NextRole.RELAY |
| 245 | and b.status == "live" |
| 246 | and (b.lane is None or b.lane == lane) |
| 247 | ] |
| 248 | product_relays = [ |
| 249 | b |
| 250 | for b in blocks |
| 251 | if b.role is NextRole.PRODUCT_RELAY |
| 252 | and b.status == "live" |
| 253 | and (b.lane is None or b.lane == lane or b.lane == "product") |
| 254 | ] |
| 255 | if relays and product_relays: |
| 256 | return None, "ambiguous_primary" |
| 257 | if len(relays) > 1 or len(product_relays) > 1: |
| 258 | return None, "ambiguous_primary" |
| 259 | if len(relays) == 1: |
| 260 | return relays[0], None |
| 261 | if len(product_relays) == 1: |
| 262 | return product_relays[0], None |
| 263 | return None, None |
| 264 | |
| 265 | |
| 266 | def primary_paste_hash(block: NextBlock) -> str | None: |
| 267 | """Hash of PRIMARY paste fence bytes (required for tip freshness).""" |
| 268 | if not block.fence: |
| 269 | return None |
| 270 | return tip_hash_hex(block.fence) |
| 271 | |
| 272 | |
| 273 | def heading_role_mismatch(block: NextBlock) -> str | None: |
| 274 | """Return a short reason when heading pattern disagrees with marker role.""" |
| 275 | heading = block.heading |
| 276 | if block.role is NextRole.PRIMARY: |
| 277 | if not heading.startswith("## NEXT SESSION"): |
| 278 | return "primary marker requires ## NEXT SESSION heading" |
| 279 | if "(PRIMARY)" not in heading: |
| 280 | return "PRIMARY heading must end with (PRIMARY)" |
| 281 | title = heading.split("—", 1)[-1] if "—" in heading else heading.split("-", 1)[-1] |
| 282 | if _forbidden_archived_next_title(title) or _ambiguous_primary_phrase(title): |
| 283 | return "forbidden PRIMARY title phrasing" |
| 284 | elif block.role is NextRole.RELAY: |
| 285 | if not heading.startswith("## NEXT SESSION"): |
| 286 | return "relay marker requires ## NEXT SESSION heading" |
| 287 | if "(RELAY →" not in heading and "(RELAY->" not in heading: |
| 288 | return "RELAY heading must contain (RELAY → …)" |
| 289 | title = heading.split("—", 1)[-1] if "—" in heading else heading |
| 290 | if _forbidden_archived_next_title(title): |
| 291 | return "forbidden RELAY title with archived" |
| 292 | elif block.role is NextRole.PRODUCT_RELAY: |
| 293 | if not heading.startswith("## PRODUCT RELAY"): |
| 294 | return "product_relay marker requires ## PRODUCT RELAY heading" |
| 295 | elif block.role is NextRole.LANE_TIP: |
| 296 | if not heading.startswith("## LANE TIP"): |
| 297 | return "lane_tip marker requires ## LANE TIP heading" |
| 298 | elif block.role is NextRole.ARCHIVED: |
| 299 | if heading.startswith("## NEXT SESSION"): |
| 300 | return "archived must not use ## NEXT SESSION heading" |
| 301 | if not heading.startswith("## ARCHIVED SESSION"): |
| 302 | return "archived marker requires ## ARCHIVED SESSION heading" |
| 303 | return None |
| 304 | |
| 305 | |
| 306 | def count_next_session_headings(text: str) -> int: |
| 307 | """Count ``## NEXT SESSION —`` headings (KH1 H2 / H13).""" |
| 308 | count = 0 |
| 309 | for line in lf_normalize(text).split("\n"): |
| 310 | if _HEADING_NEXT.match(line): |
| 311 | count += 1 |
| 312 | return count |
| 313 | |
| 314 | |
| 315 | def with_computed_hash(block: NextBlock) -> NextBlock: |
| 316 | """Return block with tip_hash filled from fence when role is primary.""" |
| 317 | if block.role is NextRole.PRIMARY and block.fence and not block.tip_hash: |
| 318 | return replace(block, tip_hash=tip_hash_hex(block.fence)) |
| 319 | return block |
File History
1 commit
sha256:6abcf1fa82a7a621ccbc945f19acdba5bc0db54569599404a1452fb4a096a199
fix(ISR): default require_independent_second_reviewer to require
Human
minor
⚠
1 day ago