classifyre-cli 0.4.74__tar.gz → 0.4.76__tar.gz

classifyre_cli-0.4.76/.turbo/turbo-build.log

@@ -0,0 +1,3 @@
+$ uv sync
+Resolved 303 packages in 258ms
+Checked 51 packages in 4ms

{classifyre_cli-0.4.74 → classifyre_cli-0.4.76}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: classifyre-cli
-Version: 0.4.74
+Version: 0.4.76
 Summary: Classifyre CLI — scan and classify unstructured data sources
 License: MIT
 Keywords: data,ingestion,metadata,pii,secrets,unstructured

{classifyre_cli-0.4.74 → classifyre_cli-0.4.76}/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@classifyre/cli",
-  "version": "0.4.74",
+  "version": "0.4.76",
   "private": true,
   "scripts": {
     "build": "uv sync",

{classifyre_cli-0.4.74 → classifyre_cli-0.4.76}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "classifyre-cli"
-version = "0.4.74"
+version = "0.4.76"
 description = "Classifyre CLI — scan and classify unstructured data sources"
 readme = "README.md"
 requires-python = ">=3.12"

classifyre_cli-0.4.76/src/detectors/engine_version.py

@@ -0,0 +1,53 @@
+"""Version stamps that invalidate the scan cache when *our* behaviour changes.
+
+The scan cache skips an asset when its content and each detector's configuration
+fingerprint still match the previous successful run.  Neither of those changes
+when the change is ours — a new pattern in the PII detector, a new file format in
+``utils/file_parser.py`` — so the fingerprint carries an explicit version that a
+developer bumps by hand.
+
+- Bump ``DETECTOR_ENGINE_VERSION[<type>]`` when a detector's rules, models,
+  thresholds or finding shape change.  Every asset is then re-scanned by that
+  detector alone; the others stay cached.
+- Bump ``PARSER_ENGINE_VERSION`` when text extraction changes what it produces —
+  a newly supported format, a different OCR path, a changed truncation rule.
+  That changes the *input* to every detector, so it invalidates everything.
+
+Forgetting a bump is not a correctness bug, but it is an invisible one: the
+improvement ships and silently does not apply to already-scanned assets.  It
+belongs on the release checklist.
+"""
+
+from __future__ import annotations
+
+from typing import Final
+
+PARSER_ENGINE_VERSION: Final[int] = 1
+
+DETECTOR_ENGINE_VERSION: Final[dict[str, int]] = {
+    "SECRETS": 1,
+    "PII": 1,
+    "YARA": 1,
+    "BROKEN_LINKS": 1,
+    "CODE_SECURITY": 1,
+    "CUSTOM": 1,
+}
+
+# Detectors whose verdict depends on state outside the asset, so identical
+# content plus identical config does not imply an identical result.
+#
+# BROKEN_LINKS resolves URLs over the network: a link that answered 200 last week
+# can 404 today with the document untouched.  Caching it would freeze link health
+# at whatever it was on first scan, which is the opposite of what the detector is
+# for.  These always run, even on an otherwise fully cached asset.
+NON_CACHEABLE_DETECTOR_TYPES: Final[frozenset[str]] = frozenset({"BROKEN_LINKS"})
+
+
+def detector_engine_version(detector_type: str) -> int:
+    """Engine version for a detector type; 0 for types added without a stamp."""
+    return DETECTOR_ENGINE_VERSION.get(detector_type.strip().upper(), 0)
+
+
+def is_cacheable_detector_type(detector_type: str) -> bool:
+    """False when the detector's result can change without the asset changing."""
+    return detector_type.strip().upper() not in NON_CACHEABLE_DETECTOR_TYPES

{classifyre_cli-0.4.74 → classifyre_cli-0.4.76}/src/main.py

@@ -265,6 +265,16 @@ async def run_command_async(args: argparse.Namespace, recipe: dict[str, Any]) ->
                             worker_pool=worker_pool,
                         )
 
+                    from .pipeline.scan_cache import ScanCache
+
+                    scan_cache = ScanCache(recipe, source)
+                    if scan_cache.enabled:
+                        logger.info(
+                            "Scan cache active (verify=%s): unchanged assets and"
+                            " unchanged detectors will be skipped",
+                            scan_cache.verify,
+                        )
+
                     # --- Phase 1: Discovery ---
                     source.set_discovery_only(True)
                     all_stubs: list[Any] = []
@@ -286,9 +296,17 @@ async def run_command_async(args: argparse.Namespace, recipe: dict[str, Any]) ->
                         # change_type onto runner_assets rows during ingest, so
                         # the row must exist or the CREATED stamp is lost and
                         # first-run counters report every asset as unchanged.
+                        #
+                        # This is also the only point where the stored checksum
+                        # still describes the *previous* scan — the ingest below
+                        # overwrites it — so the scan cache reads prior state here
+                        # or not at all.
                         hashes = [s["hash"] for s in stub_batch if s.get("hash")]
                         if hasattr(sink, "register_discovered_assets") and hashes:
-                            await sink.register_discovered_assets(hashes)
+                            prior = await sink.register_discovered_assets(
+                                hashes, include_scan_cache=scan_cache.enabled
+                            )
+                            scan_cache.record_prior(prior)
 
                         await sink.emit_batch(stub_batch, skip_findings=True)
                         output_batch_count += 1
@@ -317,6 +335,8 @@ async def run_command_async(args: argparse.Namespace, recipe: dict[str, Any]) ->
                             max_concurrent,
                         )
                         error_count = 0
+                        skipped_assets = 0
+                        skipped_detector_runs = 0
                         chunk_errors: list[str] = []
 
                         async def _process_one(asset: Any) -> None:
@@ -325,8 +345,45 @@ async def run_command_async(args: argparse.Namespace, recipe: dict[str, Any]) ->
 
                         async def _process_one_inner(asset: Any) -> None:
                             nonlocal processed_count, error_count
+                            nonlocal skipped_assets, skipped_detector_runs
                             asset_hash = getattr(asset, "hash", None) or ""
                             try:
+                                plan = await scan_cache.plan(asset)
+                                skipped_detector_runs += scan_cache.skipped_detector_runs(plan)
+
+                                if plan.mode == "skip":
+                                    # Report no detector outcomes at all. The API
+                                    # may only resolve a finding whose detector
+                                    # reported OK on this asset, so an empty
+                                    # outcome set is what preserves the findings
+                                    # this run deliberately did not re-check.
+                                    skipped_assets += 1
+                                    processed_count += 1
+                                    logger.info("Cached %s: %s", asset.name, plan.reason)
+                                    prior = plan.prior
+                                    if hasattr(sink, "update_asset_status"):
+                                        await sink.update_asset_status(
+                                            asset_hash,
+                                            "PROCESSED",
+                                            findings_total=prior.findings_total if prior else None,
+                                            findings_by_severity=(
+                                                prior.findings_by_severity if prior else None
+                                            ),
+                                            findings_by_detector=(
+                                                prior.findings_by_detector if prior else None
+                                            ),
+                                            cache_hit=True,
+                                        )
+                                    source.evict_asset_cache(asset_hash)
+                                    return
+
+                                if plan.mode == "partial":
+                                    logger.info(
+                                        "Partially cached %s: running %s",
+                                        asset.name,
+                                        ", ".join(sorted(plan.run_detector_keys)),
+                                    )
+
                                 if hasattr(sink, "update_asset_status"):
                                     await sink.update_asset_status(asset_hash, "PROCESSING")
 
@@ -355,9 +412,17 @@ async def run_command_async(args: argparse.Namespace, recipe: dict[str, Any]) ->
                                     asset,
                                     on_findings_flushed=_on_findings_flushed,
                                     findings_flush_size=args.detector_flush_batch_size,
+                                    only=(
+                                        plan.run_detector_keys if plan.mode == "partial" else None
+                                    ),
                                 )
-                                payload = _asset_to_payload(result)
-                                await sink.emit_batch([payload], skip_findings=False)
+
+                                # Chunks first, then the payload that carries the
+                                # scan-cache state. A later run reads that state's
+                                # presence as "this asset finished", so it must not
+                                # be written while the asset's embeddings are still
+                                # missing — that would strand them permanently.
+                                chunks_ok = True
                                 if hasattr(sink, "emit_text_chunks"):
                                     artifact = pipeline.take_text_artifact(asset_hash)
                                     if artifact is not None:
@@ -371,6 +436,7 @@ async def run_command_async(args: argparse.Namespace, recipe: dict[str, Any]) ->
                                             # The run fails after phase 2 so operators
                                             # cannot mistake missing embeddings for a
                                             # healthy completed scan.
+                                            chunks_ok = False
                                             detail = f"{asset_hash}: {chunk_exc}"
                                             chunk_errors.append(detail)
                                             logger.error(
@@ -380,10 +446,40 @@ async def run_command_async(args: argparse.Namespace, recipe: dict[str, Any]) ->
                                                 chunk_exc,
                                             )
 
+                                f_total, f_by_sev, f_by_det = _compute_findings_counts(
+                                    result.findings or []
+                                )
+                                payload = _asset_to_payload(result)
+
+                                content_hash = plan.content_hash
+                                if scan_cache.tracks_content_hash and content_hash is None:
+                                    # Bytes are still cached from the detector run,
+                                    # so this digest costs no extra fetch.
+                                    content_hash = await scan_cache.content_digest(asset)
+
+                                state = scan_cache.build_state(
+                                    plan,
+                                    content_hash=content_hash,
+                                    detector_outcomes=(
+                                        result.scan_stats.detector_outcomes
+                                        if result.scan_stats
+                                        else None
+                                    ),
+                                    scan_stats=result.scan_stats,
+                                    available_detector_keys=(
+                                        pipeline.available_detector_cache_keys()
+                                    ),
+                                    completed=chunks_ok,
+                                    findings_total=f_total,
+                                    findings_by_severity=f_by_sev,
+                                    findings_by_detector=f_by_det,
+                                )
+                                if state is not None:
+                                    payload["scan_cache"] = state
+
+                                await sink.emit_batch([payload], skip_findings=False)
+
                                 if hasattr(sink, "update_asset_status"):
-                                    f_total, f_by_sev, f_by_det = _compute_findings_counts(
-                                        result.findings or []
-                                    )
                                     await sink.update_asset_status(
                                         asset_hash,
                                         "PROCESSED",
@@ -409,10 +505,15 @@ async def run_command_async(args: argparse.Namespace, recipe: dict[str, Any]) ->
                         tasks = [asyncio.create_task(_process_one(a)) for a in all_stubs]
                         await asyncio.gather(*tasks, return_exceptions=True)
                         logger.info(
-                            "Phase 2 complete: %d processed, %d errors",
+                            "Phase 2 complete: %d processed (%d cached), %d errors,"
+                            " %d detector run(s) skipped",
                             processed_count,
+                            skipped_assets,
                             error_count,
+                            skipped_detector_runs,
                         )
+                        if hasattr(sink, "set_scan_cache_savings"):
+                            sink.set_scan_cache_savings(skipped_assets, skipped_detector_runs)
                         if chunk_errors:
                             preview = "; ".join(chunk_errors[:3])
                             remaining = len(chunk_errors) - 3

{classifyre_cli-0.4.74 → classifyre_cli-0.4.76}/src/models/generated_input.py

@@ -134,6 +134,34 @@ class SamplingConfig(BaseModel):
     )
 
 
+class Verify(StrEnum):
+    """
+    How strongly to prove the payload is unchanged. metadata trusts the source-reported checksum (provider ETag/MD5) and skips the download entirely. content downloads the bytes and compares a SHA-256 digest, still skipping parsing, OCR, transcription and detection. auto (default) picks per source: content where the checksum is only mtime and size (local folders), metadata where the provider supplies a content-derived digest.
+    """
+
+    auto = 'auto'
+    metadata = 'metadata'
+    content = 'content'
+
+
+class ScanCacheConfig(BaseModel):
+    """
+    Skips re-processing assets whose content and applicable detector configuration are unchanged since their last fully-successful scan. Supported by file-like sources only (local folders, object storage, cloud drives, email); ignored elsewhere, because a database row can change under the same key with no change signal to compare against. A detector whose configuration changed is always re-run — and only that detector.
+    """
+
+    model_config = ConfigDict(
+        extra='forbid',
+    )
+    enabled: bool | None = Field(
+        True,
+        description='When enabled (default), an asset whose content and per-detector configuration fingerprints all match its last successful scan is skipped, and its existing findings are carried forward untouched. Disable to re-scan everything on every run.',
+    )
+    verify: Verify | None = Field(
+        'auto',
+        description='How strongly to prove the payload is unchanged. metadata trusts the source-reported checksum (provider ETag/MD5) and skips the download entirely. content downloads the bytes and compares a SHA-256 digest, still skipping parsing, OCR, transcription and detection. auto (default) picks per source: content where the checksum is only mtime and size (local folders), metadata where the provider supplies a content-derived digest.',
+    )
+
+
 class Requests(BaseModel):
     model_config = ConfigDict(
         extra='forbid',
@@ -2110,6 +2138,7 @@ class CoreInput(BaseModel):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -2132,6 +2161,7 @@ class EmailInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -2154,6 +2184,7 @@ class S3CompatibleStorageInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -2176,6 +2207,7 @@ class LocalFolderInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -2198,6 +2230,7 @@ class SandboxInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -2220,6 +2253,7 @@ class AzureBlobStorageInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -2242,6 +2276,7 @@ class GoogleCloudStorageInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -2264,6 +2299,7 @@ class WordPressInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -2286,6 +2322,7 @@ class PostgreSQLInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -2308,6 +2345,7 @@ class MySQLInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -2330,6 +2368,7 @@ class MSSQLInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -2352,6 +2391,7 @@ class OracleInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -2374,6 +2414,7 @@ class HiveInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -2400,6 +2441,7 @@ class DatabricksInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -2432,6 +2474,7 @@ class SnowflakeInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -2458,6 +2501,7 @@ class MongoDBInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -2588,6 +2632,7 @@ class Neo4jInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -2614,6 +2659,7 @@ class PowerBIInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -2640,6 +2686,7 @@ class TableauInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -3051,6 +3098,7 @@ class ConfluenceInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -3073,6 +3121,7 @@ class JiraInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -3095,6 +3144,7 @@ class ServiceDeskInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -3155,6 +3205,7 @@ class SQLiteInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -3290,6 +3341,7 @@ class NotionInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -3537,6 +3589,7 @@ class KafkaInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -3671,6 +3724,7 @@ class ElasticsearchInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -3705,6 +3759,7 @@ class OpenSearchInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -3797,6 +3852,7 @@ class MeilisearchInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -4037,6 +4093,7 @@ class Microsoft365Input(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -4181,6 +4238,7 @@ class GoogleWorkspaceInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -4393,6 +4451,7 @@ class DropboxInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -4441,6 +4500,7 @@ class YouTubeInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -4473,6 +4533,7 @@ class SlackInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -4503,6 +4564,7 @@ class DeltaLakeInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,
@@ -4533,6 +4595,7 @@ class IcebergInput(CoreInput):
         description='Reusable custom detector IDs selected from the custom detector catalog.',
     )
     sampling: SamplingConfig
+    scan_cache: ScanCacheConfig | None = None
     resources: ResourceOverrides | None = None
     cleanup_removed_detector_findings: bool | None = Field(
         True,

{classifyre_cli-0.4.74 → classifyre_cli-0.4.76}/src/models/generated_single_asset_scan_results.py

@@ -211,6 +211,53 @@ class DetectorOutcome(BaseModel):
     )
 
 
+class ScanCacheState(BaseModel):
+    """
+    What a later run needs in order to prove this asset can be skipped, and what it should report when it does. Only a state with complete=true is eligible for reuse. A detector that raised or failed to initialize is omitted from the fingerprint map so it is retried next run rather than silently trusted.
+    """
+
+    complete: bool = Field(
+        ...,
+        description='True only after findings, extraction and text chunks all completed. False is an explicit tombstone that invalidates older reusable state.',
+        title='Complete',
+    )
+    findings_total: int | None = Field(
+        None,
+        description='Findings this scan saw on the asset, carried forward so a later skipped run reports the same total instead of zero',
+        title='Findings Total',
+    )
+    findings_by_severity: dict[str, int] | None = Field(
+        None,
+        description='Finding counts per severity, carried forward to a skipped run',
+        title='Findings By Severity',
+    )
+    findings_by_detector: dict[str, dict[str, int]] | None = Field(
+        None,
+        description='Finding counts per detector and severity, carried forward to a skipped run',
+        title='Findings By Detector',
+    )
+    empty_text: bool | None = Field(
+        None,
+        description='Whether text extraction yielded nothing, carried forward so a skipped run does not distort text-coverage reporting',
+        title='Empty Text',
+    )
+    text_extraction_status: str | None = Field(
+        None,
+        description='Text extraction outcome, carried forward to a skipped run',
+        title='Text Extraction Status',
+    )
+    content_hash: str | None = Field(
+        None,
+        description='SHA-256 of the bytes that were actually scanned. Null when the source proved sameness from a provider-supplied digest without downloading.',
+        title='Content Hash',
+    )
+    detectors: dict[str, str] | None = Field(
+        {},
+        description='Fingerprint of the configuration each detector last succeeded with, keyed by detector type or CUSTOM::<custom_detector_key>. A key whose fingerprint still matches on a later run means that detector would produce identical findings and can be skipped.',
+        title='Detectors',
+    )
+
+
 class ScanStats(BaseModel):
     """
     Statistics about detector scan for an asset
@@ -307,3 +354,8 @@ class SingleAssetScanResults(BaseModel):
         description='Source-specific asset metadata using normalized keys (size_bytes, row_count, etc.) where applicable',
         title='Metadata',
     )
+    scan_cache: ScanCacheState | None = Field(
+        None,
+        description='Scan-cache state to persist for this asset, letting a later run skip work that would produce identical results',
+        title='Scan Cache',
+    )