Issue #45 : Add document analysis function (#74)

* Add document analysis function to process fertilizer labels with OCR and LLM

* Update version to 0.0.9 and add test_label.py to .gitignore

* Bump version to 0.0.7 and add unit test for analyze_document function

* Update .gitignore to exclude analyze_label.py and modify test_pipeline to use organization name for manufacturer assertion

.gitignore

@@ -7,6 +7,7 @@ __pycache__/
 logs/
 test_logs/
 reports/
+analyze_label.py
 
 # VS Code
 .vscode

pipeline/__init__.py

@@ -55,3 +55,31 @@ def analyze(label_storage: LabelStorage, ocr: OCR, gpt: GPT, log_dir_path: str =
     os.remove(f"{log_dir_path}/{now}.json")
 
     return inspection
+
+def analyze_document(document: bytes, ocr: OCR, gpt: GPT, log_dir_path: str = './logs') -> FertilizerInspection:
+    """
+    Analyze the raw document of the fertiliser label using an OCR and an LLM.
+    It returns the data extracted from the label in a FertiliserForm.
+    """
+    result = ocr.extract_text(document=document)
+
+    # Logs the results from document intelligence
+    now = datetime.now().strftime('%Y-%m-%d_%H-%M-%S')
+    save_text_to_file(result.content, f"{log_dir_path}/{now}.md")
+
+    # Generate inspection from extracted text
+    prediction = gpt.create_inspection(result.content)
+
+    # Check the coninspectionity of the JSON
+    inspection = prediction.inspection
+
+    # Logs the results from GPT
+    save_text_to_file(prediction.reasoning, f"{log_dir_path}/{now}.txt")
+    save_text_to_file(inspection.model_dump_json(indent=2), f"{log_dir_path}/{now}.json")
+
+    # Delete the logs if there's no error
+    os.remove(f"{log_dir_path}/{now}.md")   
+    os.remove(f"{log_dir_path}/{now}.txt")     
+    os.remove(f"{log_dir_path}/{now}.json")
+
+    return inspection

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "fertiscan_pipeline"
-version = "0.0.8"
+version = "0.0.9"
 description = "A pipeline for the FertiScan project"
 authors = [
     { name = "Albert Bryan Ndjeutcha", email = "[email protected]" }

tests/test_pipeline.py

@@ -6,7 +6,7 @@
 
 from dotenv import load_dotenv
 
-from pipeline import GPT, OCR, LabelStorage, analyze
+from pipeline import GPT, OCR, LabelStorage, analyze, analyze_document
 from pipeline.inspection import FertilizerInspection, Value
 from tests import curl_file, levenshtein_similarity
 
@@ -89,6 +89,39 @@ def test_analyze(self):
         self.assertFalse(os.path.exists(json_log_path))
         self.assertFalse(os.path.exists(txt_log_path))
 
+    def test_analyze_document(self):
+        # Run the analyze function
+        self.setUpClass()
+        inspection = analyze_document(
+            self.label_storage.get_document(), self.ocr, self.gpt, log_dir_path=self.log_dir_path
+        )
+
+        # Perform assertions
+        self.assertIsInstance(inspection, FertilizerInspection, inspection)
+        self.assertIn(Value(value="25", unit="kg"), inspection.weight, inspection)
+        manufacturer_or_company = inspection.organizations[0].name
+        self.assertIsNotNone(manufacturer_or_company, inspection)
+        self.assertGreater(
+            levenshtein_similarity(
+                manufacturer_or_company, "TerraLink Horticulture Inc."
+            ),
+            0.95,
+            inspection,
+        )
+        self.assertGreater(
+            levenshtein_similarity(inspection.npk, "10-52-0"), 0.90, inspection
+        )
+
+        # Ensure logs are created and then deleted
+        now = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+        md_log_path = f"{self.log_dir_path}/{now}.md"
+        json_log_path = f"{self.log_dir_path}/{now}.json"
+        txt_log_path = f"{self.log_dir_path}/{now}.txt"
+
+        self.assertFalse(os.path.exists(md_log_path))
+        self.assertFalse(os.path.exists(json_log_path))
+        self.assertFalse(os.path.exists(txt_log_path))
+
 
 class TestInspectionAnnotatedFields(unittest.TestCase):
     @classmethod