plugin.py
python
sha256:b636f72dcba9e190afb980bece906fa5b717fbde014b76ef023df8cb96e01eb9
docs: expand cache plan with all seven testing tiers and do…
Sonnet 4.6
131 days ago
| 1 | """Mist domain plugin — content-addressed, signed, agent-native artifact hosting. |
| 2 | |
| 3 | A *mist* is a single versioned artifact stored under a content-derived filename. |
| 4 | The filename **is** the identity: the first 12 characters of the base-58 encoding |
| 5 | of its SHA-256 digest, optionally suffixed with a human-readable extension. |
| 6 | |
| 7 | Design goals |
| 8 | ------------ |
| 9 | - **Content-addressed** — same bytes always produce the same mist ID; no |
| 10 | collision with any other content is possible under SHA-256. |
| 11 | - **Domain-agnostic** — MIDI files, Solidity ABIs, JSON Schemas, prose, code, |
| 12 | images, and arbitrary binary blobs are all first-class citizens. |
| 13 | - **Signed** — every mist carries an MSign Ed25519 author signature; AI-produced |
| 14 | mists also embed ``agent_id`` + ``model_id`` for provenance. |
| 15 | - **VCS-native** — because MistPlugin satisfies ``MuseDomainPlugin``, all 14 |
| 16 | ``muse`` CLI commands (status, diff, merge, log, …) work on mist repos without |
| 17 | any core engine changes. |
| 18 | |
| 19 | Phase 1 scope |
| 20 | ------------- |
| 21 | Pure domain layer only. No CLI sub-commands, no MuseHub API routes, no UI. |
| 22 | Those land in Phases 2–4. |
| 23 | |
| 24 | See ``docs/mists.md`` for the full architecture document. |
| 25 | """ |
| 26 | |
| 27 | from __future__ import annotations |
| 28 | |
| 29 | import hashlib |
| 30 | import os |
| 31 | import pathlib |
| 32 | import stat as _stat |
| 33 | from typing import TYPE_CHECKING |
| 34 | |
| 35 | from muse._version import __version__ |
| 36 | from muse.core.diff_algorithms import snapshot_diff |
| 37 | from muse.core.schema import ( |
| 38 | DimensionSpec, |
| 39 | DomainSchema, |
| 40 | SetSchema, |
| 41 | ) |
| 42 | from muse.core.stat_cache import load_cache |
| 43 | from muse.core._types import Manifest |
| 44 | |
| 45 | type _ArtifactInfo = dict[str, str] |
| 46 | from muse.domain import ( |
| 47 | DriftReport, |
| 48 | LiveState, |
| 49 | MergeResult, |
| 50 | SnapshotManifest, |
| 51 | StateDelta, |
| 52 | StateSnapshot, |
| 53 | ) |
| 54 | |
| 55 | if TYPE_CHECKING: |
| 56 | pass |
| 57 | |
| 58 | # --------------------------------------------------------------------------- |
| 59 | # Module-level constants |
| 60 | # --------------------------------------------------------------------------- |
| 61 | |
| 62 | _DOMAIN_NAME = "mist" |
| 63 | |
| 64 | # Bitcoin base-58 alphabet — omits visually ambiguous characters: 0, O, I, l. |
| 65 | # Same bytes always produce the same base-58 string, making mist IDs |
| 66 | # deterministic and URL-safe. |
| 67 | _BASE58_ALPHABET = "123456789ABCDEFGHJKLMNPQRSTUVWXYZabcdefghijkmnopqrstuvwxyz" |
| 68 | |
| 69 | # Length of the mist ID prefix taken from the full base-58 encoding. |
| 70 | # 12 characters from SHA-256 → ~10^21 distinct values; collision probability is |
| 71 | # negligible for any realistic corpus. |
| 72 | _MIST_ID_LENGTH = 12 |
| 73 | |
| 74 | # Known artifact type → (category, language/subtype) pairs. |
| 75 | # Checked after magic-byte and JSON-key inspection — extension is the last resort. |
| 76 | _EXT_MAP: dict[str, tuple[str, str]] = { |
| 77 | # Code |
| 78 | ".py": ("code", "python"), |
| 79 | ".js": ("code", "javascript"), |
| 80 | ".ts": ("code", "typescript"), |
| 81 | ".tsx": ("code", "typescript"), |
| 82 | ".jsx": ("code", "javascript"), |
| 83 | ".rs": ("code", "rust"), |
| 84 | ".go": ("code", "go"), |
| 85 | ".java": ("code", "java"), |
| 86 | ".c": ("code", "c"), |
| 87 | ".cpp": ("code", "cpp"), |
| 88 | ".h": ("code", "c"), |
| 89 | ".hpp": ("code", "cpp"), |
| 90 | ".rb": ("code", "ruby"), |
| 91 | ".swift": ("code", "swift"), |
| 92 | ".kt": ("code", "kotlin"), |
| 93 | ".sol": ("code", "solidity"), |
| 94 | ".vy": ("code", "vyper"), |
| 95 | # Prose |
| 96 | ".md": ("prose", "markdown"), |
| 97 | ".txt": ("prose", "text"), |
| 98 | ".rst": ("prose", "restructuredtext"), |
| 99 | # Data |
| 100 | ".csv": ("data", "csv"), |
| 101 | ".toml": ("data", "toml"), |
| 102 | ".yaml": ("data", "yaml"), |
| 103 | ".yml": ("data", "yaml"), |
| 104 | ".xml": ("data", "xml"), |
| 105 | # MIDI — also detected by magic bytes (MThd) |
| 106 | ".mid": ("midi", "midi"), |
| 107 | ".midi": ("midi", "midi"), |
| 108 | } |
| 109 | |
| 110 | |
| 111 | # --------------------------------------------------------------------------- |
| 112 | # Pure functions — the domain's intelligence layer |
| 113 | # --------------------------------------------------------------------------- |
| 114 | |
| 115 | |
| 116 | def compute_mist_id(content: bytes) -> str: |
| 117 | """Derive the globally unique mist ID from raw artifact bytes. |
| 118 | |
| 119 | The ID is the first :data:`_MIST_ID_LENGTH` characters of the base-58 |
| 120 | encoding of the artifact's SHA-256 digest. The same bytes always produce |
| 121 | the same ID. |
| 122 | |
| 123 | Args: |
| 124 | content: Raw bytes of the artifact. |
| 125 | |
| 126 | Returns: |
| 127 | A 12-character URL-safe base-58 string, e.g. ``"aB3xQ9fWmK2r"``. |
| 128 | |
| 129 | Examples: |
| 130 | >>> compute_mist_id(b"hello") |
| 131 | 'GJGvdqT2tQ5j' |
| 132 | >>> compute_mist_id(b"") != compute_mist_id(b"x") |
| 133 | True |
| 134 | """ |
| 135 | digest = hashlib.sha256(content).digest() |
| 136 | # Encode the 32-byte digest as a big-endian integer then convert to base-58. |
| 137 | n = int.from_bytes(digest, "big") |
| 138 | chars: list[str] = [] |
| 139 | while n: |
| 140 | n, remainder = divmod(n, 58) |
| 141 | chars.append(_BASE58_ALPHABET[remainder]) |
| 142 | # Preserve leading zero bytes as base-58 '1' characters. |
| 143 | for byte in digest: |
| 144 | if byte == 0: |
| 145 | chars.append(_BASE58_ALPHABET[0]) |
| 146 | else: |
| 147 | break |
| 148 | encoded = "".join(reversed(chars)) |
| 149 | return encoded[:_MIST_ID_LENGTH] |
| 150 | |
| 151 | |
| 152 | def detect_artifact_type(filename: str, content: bytes) -> _ArtifactInfo: |
| 153 | """Infer the artifact type and language from filename and raw content. |
| 154 | |
| 155 | Detection order (most-to-least reliable): |
| 156 | 1. Magic bytes — MIDI ``MThd`` header; catches mis-named files. |
| 157 | 2. JSON key inspection — ABI arrays, JSON Schema ``$schema`` key. |
| 158 | 3. Extension map — falls back gracefully to ``("unknown", "binary")``. |
| 159 | |
| 160 | Args: |
| 161 | filename: Bare filename (no path separators), e.g. ``"contract.abi.json"``. |
| 162 | content: Raw bytes of the artifact. |
| 163 | |
| 164 | Returns: |
| 165 | A dict with keys ``"artifact_type"`` and ``"language"``, e.g. |
| 166 | ``{"artifact_type": "midi", "language": "midi"}`` or |
| 167 | ``{"artifact_type": "code", "language": "python"}``. |
| 168 | |
| 169 | Examples: |
| 170 | >>> detect_artifact_type("track.mid", b"MThd\\x00\\x00\\x00\\x06") |
| 171 | {'artifact_type': 'midi', 'language': 'midi'} |
| 172 | >>> detect_artifact_type("schema.json", b'{"$schema":"http://..."}') |
| 173 | {'artifact_type': 'json_schema', 'language': 'json'} |
| 174 | """ |
| 175 | # 1. Magic bytes — MIDI |
| 176 | if content[:4] == b"MThd": |
| 177 | return {"artifact_type": "midi", "language": "midi"} |
| 178 | |
| 179 | # 2. JSON key inspection |
| 180 | if filename.endswith(".json") or filename.endswith(".abi"): |
| 181 | try: |
| 182 | import json as _json |
| 183 | |
| 184 | parsed = _json.loads(content.decode("utf-8", errors="replace")) |
| 185 | if isinstance(parsed, list) and parsed and isinstance(parsed[0], dict): |
| 186 | # Ethereum ABI is an array of objects each with "type" and "name" |
| 187 | if "type" in parsed[0] and "name" in parsed[0]: |
| 188 | return {"artifact_type": "abi", "language": "json"} |
| 189 | if isinstance(parsed, dict): |
| 190 | if "$schema" in parsed: |
| 191 | return {"artifact_type": "json_schema", "language": "json"} |
| 192 | except Exception: |
| 193 | pass |
| 194 | |
| 195 | # 3. Extension map |
| 196 | ext = pathlib.PurePosixPath(filename).suffix.lower() |
| 197 | if ext in _EXT_MAP: |
| 198 | artifact_type, language = _EXT_MAP[ext] |
| 199 | return {"artifact_type": artifact_type, "language": language} |
| 200 | |
| 201 | return {"artifact_type": "unknown", "language": "binary"} |
| 202 | |
| 203 | |
| 204 | def _validate_mist_filename(filename: str) -> None: |
| 205 | """Validate a proposed mist filename for safety and correctness. |
| 206 | |
| 207 | This is the security gate for all user-supplied filenames. It rejects |
| 208 | any name that could be used to escape the mist store or inject control |
| 209 | sequences into terminals. |
| 210 | |
| 211 | Enforced rules |
| 212 | -------------- |
| 213 | - No null bytes (``\\x00``). |
| 214 | - No path separators (``/`` or ``\\``). |
| 215 | - No directory traversal sequences (``..``). |
| 216 | - No control characters (``\\x01``–``\\x1f``, ``\\x7f``). |
| 217 | - No ANSI escape sequences (``\\x1b[``). |
| 218 | - Length ≤ 255 characters. |
| 219 | |
| 220 | Args: |
| 221 | filename: The filename to validate (must be a bare name, no path). |
| 222 | |
| 223 | Raises: |
| 224 | ValueError: Describing exactly which rule was violated. |
| 225 | |
| 226 | Examples: |
| 227 | >>> _validate_mist_filename("aB3xQ9fWmK2r.py") # valid — no error |
| 228 | >>> _validate_mist_filename("../evil") |
| 229 | Traceback (most recent call last): |
| 230 | ... |
| 231 | ValueError: Mist filename must not contain path traversal sequences: '../evil' |
| 232 | """ |
| 233 | if len(filename) > 255: |
| 234 | raise ValueError( |
| 235 | f"Mist filename exceeds 255-character limit: {len(filename)} chars" |
| 236 | ) |
| 237 | if "\x00" in filename: |
| 238 | raise ValueError(f"Mist filename must not contain null bytes: {filename!r}") |
| 239 | if ".." in filename: |
| 240 | raise ValueError( |
| 241 | f"Mist filename must not contain path traversal sequences: {filename!r}" |
| 242 | ) |
| 243 | if "/" in filename or "\\" in filename: |
| 244 | raise ValueError( |
| 245 | f"Mist filename must not contain path separators: {filename!r}" |
| 246 | ) |
| 247 | if "\x1b[" in filename: |
| 248 | raise ValueError( |
| 249 | f"Mist filename must not contain ANSI escape sequences: {filename!r}" |
| 250 | ) |
| 251 | for ch in filename: |
| 252 | cp = ord(ch) |
| 253 | if 0x01 <= cp <= 0x1F or cp == 0x7F: |
| 254 | raise ValueError( |
| 255 | f"Mist filename must not contain control characters: {filename!r}" |
| 256 | ) |
| 257 | |
| 258 | |
| 259 | def extract_mist_symbol_anchors(filename: str, content: bytes) -> list[str]: |
| 260 | """Extract symbol anchors for code and structured mist artifacts. |
| 261 | |
| 262 | Delegates to the ``muse.plugins.code.ast_parser`` layer, which supports |
| 263 | Python, TypeScript, Solidity, Markdown headings, TOML sections, and more. |
| 264 | Returns an empty list for binary or unrecognised file types — these are |
| 265 | still valid mists; they just have no intra-file anchor points. |
| 266 | |
| 267 | Args: |
| 268 | filename: Bare filename (no path separators), e.g. ``"utils.py"``. |
| 269 | content: Raw bytes of the artifact. |
| 270 | |
| 271 | Returns: |
| 272 | A list of symbol address strings in ``"filename::SymbolName"`` format, |
| 273 | e.g. ``["utils.py::compute_checksum", "utils.py::BaseHandler"]``. |
| 274 | |
| 275 | Examples: |
| 276 | >>> anchors = extract_mist_symbol_anchors("add.py", b"def add(a, b): return a + b") |
| 277 | >>> "add.py::add" in anchors |
| 278 | True |
| 279 | """ |
| 280 | try: |
| 281 | from muse.plugins.code.ast_parser import parse_symbols |
| 282 | |
| 283 | tree = parse_symbols(content, filename) |
| 284 | # SymbolTree keys are full addresses: "filename.py::SymbolName" |
| 285 | # Filter out import pseudo-symbols (kind == "import") by checking address. |
| 286 | return [addr for addr in tree if "::import::" not in addr] |
| 287 | except Exception: |
| 288 | return [] |
| 289 | |
| 290 | |
| 291 | # --------------------------------------------------------------------------- |
| 292 | # MistPlugin — MuseDomainPlugin implementation |
| 293 | # --------------------------------------------------------------------------- |
| 294 | |
| 295 | |
| 296 | class MistPlugin: |
| 297 | """Domain plugin for mist repositories. |
| 298 | |
| 299 | Satisfies the full :class:`~muse.domain.MuseDomainPlugin` protocol plus the |
| 300 | optional OT merge extension (:class:`~muse.domain.StructuredMergePlugin`). |
| 301 | No explicit inheritance needed — structural duck-typing applies. |
| 302 | |
| 303 | All 14 ``muse`` CLI commands work immediately on any mist repo once this |
| 304 | plugin is registered. The mist-specific behaviour is: |
| 305 | |
| 306 | - Every tracked file is a *mist*: a single content-addressed artifact. |
| 307 | - The snapshot manifest maps mist IDs (filenames) to their SHA-256 hashes. |
| 308 | - Merges are set-algebraic at file granularity — a mist either exists or not. |
| 309 | - Symbol anchors are extracted for code/structured mists; binary mists have none. |
| 310 | """ |
| 311 | |
| 312 | # ------------------------------------------------------------------ |
| 313 | # MuseDomainPlugin — required core protocol |
| 314 | # ------------------------------------------------------------------ |
| 315 | |
| 316 | def snapshot(self, live_state: LiveState) -> StateSnapshot: |
| 317 | """Capture the current mist store as a content-addressed manifest. |
| 318 | |
| 319 | Walks every file under ``live_state`` (respecting ``.museignore``), |
| 320 | hashing raw bytes with SHA-256. Returns a ``SnapshotManifest`` whose |
| 321 | ``files`` dict maps workspace-relative POSIX paths to their digests. |
| 322 | |
| 323 | Args: |
| 324 | live_state: Either a ``pathlib.Path`` pointing to the mist store |
| 325 | directory, or a ``SnapshotManifest`` dict for in-memory use. |
| 326 | |
| 327 | Returns: |
| 328 | A ``SnapshotManifest`` mapping mist filenames to SHA-256 digests. |
| 329 | """ |
| 330 | if isinstance(live_state, pathlib.Path): |
| 331 | from muse.core.ignore import is_ignored, load_ignore_config, resolve_patterns |
| 332 | |
| 333 | workdir = live_state |
| 334 | patterns = resolve_patterns(load_ignore_config(workdir), _DOMAIN_NAME) |
| 335 | cache = load_cache(workdir) |
| 336 | files: Manifest = {} |
| 337 | root_str = str(workdir) |
| 338 | prefix_len = len(root_str) + 1 |
| 339 | |
| 340 | for dirpath, dirnames, filenames in os.walk(root_str, followlinks=False): |
| 341 | dirnames[:] = sorted(d for d in dirnames if not d.startswith(".")) |
| 342 | for fname in sorted(filenames): |
| 343 | if fname.startswith("."): |
| 344 | continue |
| 345 | abs_str = os.path.join(dirpath, fname) |
| 346 | try: |
| 347 | st = os.lstat(abs_str) |
| 348 | except OSError: |
| 349 | continue |
| 350 | if not _stat.S_ISREG(st.st_mode): |
| 351 | continue |
| 352 | rel = abs_str[prefix_len:] |
| 353 | if os.sep != "/": |
| 354 | rel = rel.replace(os.sep, "/") |
| 355 | if is_ignored(rel, patterns): |
| 356 | continue |
| 357 | files[rel] = cache.get_cached( |
| 358 | rel, abs_str, st.st_mtime, st.st_size, st.st_ino |
| 359 | ) |
| 360 | |
| 361 | cache.prune(set(files)) |
| 362 | cache.save() |
| 363 | return SnapshotManifest(files=files, domain=_DOMAIN_NAME, directories=[]) |
| 364 | |
| 365 | # SnapshotManifest dict path — used by merge / diff in memory |
| 366 | return live_state |
| 367 | |
| 368 | def diff( |
| 369 | self, |
| 370 | base: StateSnapshot, |
| 371 | target: StateSnapshot, |
| 372 | *, |
| 373 | repo_root: pathlib.Path | None = None, |
| 374 | ) -> StateDelta: |
| 375 | """Compute the typed operation list between two mist snapshots. |
| 376 | |
| 377 | Delegates to ``snapshot_diff`` which performs set algebra on the |
| 378 | ``files`` dicts: new mists → InsertOp, removed mists → DeleteOp, |
| 379 | replaced mists → ReplaceOp. |
| 380 | |
| 381 | Args: |
| 382 | base: Snapshot of the earlier state (e.g. HEAD). |
| 383 | target: Snapshot of the later state (e.g. working tree). |
| 384 | |
| 385 | Returns: |
| 386 | A ``StructuredDelta`` whose ``ops`` list describes every change. |
| 387 | """ |
| 388 | return snapshot_diff(self.schema(), base, target) |
| 389 | |
| 390 | def merge( |
| 391 | self, |
| 392 | base: StateSnapshot, |
| 393 | left: StateSnapshot, |
| 394 | right: StateSnapshot, |
| 395 | *, |
| 396 | repo_root: pathlib.Path | None = None, |
| 397 | ) -> MergeResult: |
| 398 | """Three-way merge of two mist snapshots against a common ancestor. |
| 399 | |
| 400 | Mists are content-addressed, so set-algebraic merge is correct by |
| 401 | construction: if both branches added the same bytes, they added the same |
| 402 | mist and there is no conflict. |
| 403 | |
| 404 | Conflict rules |
| 405 | ~~~~~~~~~~~~~~ |
| 406 | - Both sides agree → consensus wins. |
| 407 | - Only one side changed → take that side. |
| 408 | - Both sides changed differently → conflict (same path, different content). |
| 409 | |
| 410 | Args: |
| 411 | base: Common ancestor snapshot. |
| 412 | left: Snapshot from the current branch (ours). |
| 413 | right: Snapshot from the incoming branch (theirs). |
| 414 | |
| 415 | Returns: |
| 416 | A ``MergeResult`` with ``merged`` snapshot and ``conflicts`` list. |
| 417 | """ |
| 418 | base_files = base["files"] |
| 419 | left_files = left["files"] |
| 420 | right_files = right["files"] |
| 421 | |
| 422 | merged: Manifest = dict(base_files) |
| 423 | conflicts: list[str] = [] |
| 424 | |
| 425 | all_paths = set(base_files) | set(left_files) | set(right_files) |
| 426 | for path in sorted(all_paths): |
| 427 | b_val = base_files.get(path) |
| 428 | l_val = left_files.get(path) |
| 429 | r_val = right_files.get(path) |
| 430 | |
| 431 | if l_val == r_val: |
| 432 | # Both sides agree — consensus wins (including both deleted) |
| 433 | if l_val is None: |
| 434 | merged.pop(path, None) |
| 435 | else: |
| 436 | merged[path] = l_val |
| 437 | elif b_val == l_val: |
| 438 | # Only right changed |
| 439 | if r_val is None: |
| 440 | merged.pop(path, None) |
| 441 | else: |
| 442 | merged[path] = r_val |
| 443 | elif b_val == r_val: |
| 444 | # Only left changed |
| 445 | if l_val is None: |
| 446 | merged.pop(path, None) |
| 447 | else: |
| 448 | merged[path] = l_val |
| 449 | else: |
| 450 | # Both changed differently — conflict; keep left as placeholder |
| 451 | conflicts.append(path) |
| 452 | merged[path] = l_val or r_val or b_val or "" |
| 453 | |
| 454 | return MergeResult( |
| 455 | merged=SnapshotManifest(files=merged, domain=_DOMAIN_NAME, directories=[]), |
| 456 | conflicts=conflicts, |
| 457 | ) |
| 458 | |
| 459 | def drift(self, committed: StateSnapshot, live: LiveState) -> DriftReport: |
| 460 | """Report how much the mist store has drifted from the last commit. |
| 461 | |
| 462 | Called by ``muse status``. Snapshots the current working tree, diffs |
| 463 | it against the committed state, and returns a ``DriftReport``. |
| 464 | |
| 465 | Args: |
| 466 | committed: The last committed snapshot. |
| 467 | live: Current live state (path or snapshot manifest). |
| 468 | |
| 469 | Returns: |
| 470 | A ``DriftReport`` with ``has_drift``, ``summary``, and ``delta``. |
| 471 | """ |
| 472 | current = self.snapshot(live) |
| 473 | delta = self.diff(committed, current) |
| 474 | has_drift = len(delta["ops"]) > 0 |
| 475 | return DriftReport( |
| 476 | has_drift=has_drift, |
| 477 | summary=delta["summary"], |
| 478 | delta=delta, |
| 479 | ) |
| 480 | |
| 481 | def apply(self, delta: StateDelta, live_state: LiveState) -> LiveState: |
| 482 | """Apply a delta to the mist store. |
| 483 | |
| 484 | Mists are atomic blobs — the core engine already handles file-level |
| 485 | object restoration during ``muse checkout``. No domain-level |
| 486 | post-processing is needed. |
| 487 | |
| 488 | Args: |
| 489 | delta: The typed operation list to apply. |
| 490 | live_state: Current live state. |
| 491 | |
| 492 | Returns: |
| 493 | The unchanged live state. |
| 494 | """ |
| 495 | return live_state |
| 496 | |
| 497 | # ------------------------------------------------------------------ |
| 498 | # Domain schema — required |
| 499 | # ------------------------------------------------------------------ |
| 500 | |
| 501 | def schema(self) -> DomainSchema: |
| 502 | """Declare the structural shape of the mist domain. |
| 503 | |
| 504 | Mists are a **set** of content-addressed artifacts identified by |
| 505 | content (same bytes = same mist). The schema drives diff algorithm |
| 506 | selection and merge routing. |
| 507 | |
| 508 | Dimensions |
| 509 | ---------- |
| 510 | ``artifacts`` |
| 511 | The primary dimension: the set of mist files. Identity is |
| 512 | ``"by_content"`` — the mist ID *is* the content hash prefix. |
| 513 | |
| 514 | ``metadata`` |
| 515 | A set of key-value annotation pairs (tags, descriptions, provenance |
| 516 | fields). Added in Phase 3 when MuseHub-side metadata is versioned. |
| 517 | |
| 518 | Returns: |
| 519 | A ``DomainSchema`` describing the mist domain's structure. |
| 520 | """ |
| 521 | return DomainSchema( |
| 522 | domain=_DOMAIN_NAME, |
| 523 | description=( |
| 524 | "Mist domain — content-addressed, signed, agent-native artifact hosting. " |
| 525 | "A mist is a single versioned artifact (code, MIDI, ABI, prose, or any " |
| 526 | "binary blob) identified by the first 12 characters of its SHA-256 " |
| 527 | "base-58 digest. Same bytes = same mist ID, always." |
| 528 | ), |
| 529 | top_level=SetSchema( |
| 530 | kind="set", |
| 531 | element_type="artifact", |
| 532 | identity="by_content", |
| 533 | ), |
| 534 | dimensions=[ |
| 535 | DimensionSpec( |
| 536 | name="artifacts", |
| 537 | description=( |
| 538 | "The set of mist artifacts in this store. " |
| 539 | "Identity is by content — the mist ID is the hash prefix." |
| 540 | ), |
| 541 | schema=SetSchema( |
| 542 | kind="set", |
| 543 | element_type="artifact", |
| 544 | identity="by_content", |
| 545 | ), |
| 546 | independent_merge=True, |
| 547 | ), |
| 548 | DimensionSpec( |
| 549 | name="metadata", |
| 550 | description=( |
| 551 | "Annotation metadata for mists: tags, descriptions, " |
| 552 | "provenance fields (agent_id, model_id, signature)." |
| 553 | ), |
| 554 | schema=SetSchema( |
| 555 | kind="set", |
| 556 | element_type="annotation", |
| 557 | identity="by_content", |
| 558 | ), |
| 559 | independent_merge=True, |
| 560 | ), |
| 561 | ], |
| 562 | merge_mode="three_way", |
| 563 | schema_version=__version__, |
| 564 | ) |
File History
3 commits
sha256:b636f72dcba9e190afb980bece906fa5b717fbde014b76ef023df8cb96e01eb9
docs: expand cache plan with all seven testing tiers and do…
Sonnet 4.6
131 days ago
sha256:7f9e2ef5286aedad9c1e6011b4c46ca27f39dbdad6e3409357e36b26e46b3b7c
docs: docstring sprint for-each-ref→hotspots — idiomatic ru…
Sonnet 4.6
patch
137 days ago
sha256:a09b1b4f6838754495547f200aa0ce88e2f56ffc5b20b900f6f0cff2c3cdede9
fix(cursorignore): remove git-ism (.git/worktrees)
Human
minor
⚠
140 days ago