classifyre-cli 0.6.17__tar.gz → 0.7.0__tar.gz

{classifyre_cli-0.6.17 → classifyre_cli-0.7.0}/.turbo/turbo-build.log

@@ -1,3 +1,3 @@
 $ uv sync
-Resolved 325 packages in 109ms
+Resolved 325 packages in 165ms
 Checked 52 packages in 2ms

{classifyre_cli-0.6.17 → classifyre_cli-0.7.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: classifyre-cli
-Version: 0.6.17
+Version: 0.7.0
 Summary: Classifyre CLI — scan and classify unstructured data sources
 License: MIT
 Keywords: data,ingestion,metadata,pii,secrets,unstructured

{classifyre_cli-0.6.17 → classifyre_cli-0.7.0}/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@classifyre/cli",
-  "version": "0.6.17",
+  "version": "0.7.0",
   "private": true,
   "scripts": {
     "build": "uv sync",

{classifyre_cli-0.6.17 → classifyre_cli-0.7.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "classifyre-cli"
-version = "0.6.17"
+version = "0.7.0"
 description = "Classifyre CLI — scan and classify unstructured data sources"
 readme = "README.md"
 # Upper bound is load-bearing: onnxruntime publishes no cp315 wheels (notably

{classifyre_cli-0.6.17 → classifyre_cli-0.7.0}/src/detectors/custom/extractor.py

@@ -13,10 +13,11 @@ from ...models.generated_detectors import (
     CustomExtractorField,
 )
 from ..dependencies import MissingDependencyError, require_module
+from .runners._gliner2 import _from_pretrained
 
 logger = logging.getLogger(__name__)
 
-_DEFAULT_GLINER2_MODEL = "fastino/gliner2-base-v1"
+_DEFAULT_GLINER2_MODEL = "fastino/gliner2.5-base-v1"
 
 # Extraction method tags sent to the API via DetectionResult.extraction_method
 EXTRACTION_METHOD_REGEX = "REGEX"
@@ -219,7 +220,7 @@ class CustomExtractor:
         try:
             gliner2_module = require_module("gliner2", "custom", ["classification", "detectors"])
             model_name = self._config.gliner_model or _DEFAULT_GLINER2_MODEL
-            self._gliner_model = gliner2_module.GLiNER2.from_pretrained(model_name)
+            self._gliner_model = _from_pretrained(gliner2_module, model_name)
             return self._gliner_model
         except MissingDependencyError:
             raise

{classifyre_cli-0.6.17 → classifyre_cli-0.7.0}/src/detectors/custom/runners/_base.py

@@ -20,7 +20,7 @@ from ....models.generated_single_asset_scan_results import (
     Location,
 )
 
-_DEFAULT_GLINER2_MODEL = "fastino/gliner2-base-v1"
+_DEFAULT_GLINER2_MODEL = "fastino/gliner2.5-base-v1"
 _DEFAULT_IMAGE_CLASSIFICATION_MODEL = "google/vit-base-patch16-224"
 
 _TEXT_CONTENT_TYPES = [
@@ -165,6 +165,44 @@ class BaseRunner(ABC):
         runner_type = (result.metadata or {}).get("runner", "GLINER2")
         pipeline_result_dump = _slim_pipeline_result(result)
 
+        # Document-level answers go first: max_findings truncates this list, and
+        # a long document's entity spans must never crowd out its classification.
+        for task, outcome in (result.classification or {}).items():
+            labels = outcome.get("labels")
+            choices = (
+                [item for item in labels if isinstance(item, dict)]
+                if isinstance(labels, list)
+                else [outcome]
+            )
+            for choice in choices:
+                label = str(choice.get("label", ""))
+                confidence = float(choice.get("confidence", 0.0))
+                if not label:
+                    continue
+
+                choice_severity = choice.get("severity")
+                if isinstance(choice_severity, str) and choice_severity in Severity.__members__:
+                    sev = Severity(choice_severity)
+                else:
+                    sev = Severity.medium if confidence < 0.9 else Severity.high
+
+                findings.append(
+                    self._make_result(
+                        finding_type=f"classification:{task}:{label}",
+                        category="CLASSIFICATION",
+                        severity=sev,
+                        confidence=min(0.99, confidence),
+                        matched_content=text[:320],
+                        location=None,
+                        metadata={
+                            "runner": runner_type,
+                            "task": task,
+                            "label": label,
+                        },
+                        extracted_data=pipeline_result_dump,
+                    )
+                )
+
         for label, spans in (result.entities or {}).items():
             for span in spans:
                 confidence = float(span.get("confidence", 0.0))
@@ -190,6 +228,8 @@ class BaseRunner(ABC):
                 }
                 if "groups" in span:
                     meta["capture_groups"] = span["groups"]
+                if isinstance(span.get("occurrences"), int):
+                    meta["occurrences"] = span["occurrences"]
 
                 findings.append(
                     self._make_result(
@@ -204,27 +244,4 @@ class BaseRunner(ABC):
                     )
                 )
 
-        for task, outcome in (result.classification or {}).items():
-            label = str(outcome.get("label", ""))
-            confidence = float(outcome.get("confidence", 0.0))
-            if not label:
-                continue
-
-            findings.append(
-                self._make_result(
-                    finding_type=f"classification:{task}:{label}",
-                    category="CLASSIFICATION",
-                    severity=Severity.medium if confidence < 0.9 else Severity.high,
-                    confidence=min(0.99, confidence),
-                    matched_content=text[:320],
-                    location=None,
-                    metadata={
-                        "runner": runner_type,
-                        "task": task,
-                        "label": label,
-                    },
-                    extracted_data=pipeline_result_dump,
-                )
-            )
-
         return findings