archive-ai 1.0.0__tar.gz → 1.0.2__tar.gz

{archive_ai-1.0.0 → archive_ai-1.0.2}/LICENSE

@@ -1,21 +1,21 @@
-MIT License
-
-Copyright (c) 2026
-
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
+MIT License
+
+Copyright (c) 2026
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

{archive_ai-1.0.0 → archive_ai-1.0.2}/MANIFEST.in

@@ -1,4 +1,4 @@
-include LICENSE
-include README.md
-recursive-include archive_ai/templates *
-recursive-include archive_ai/static *
+include LICENSE
+include README.md
+recursive-include archive_ai/templates *
+recursive-include archive_ai/static *

{archive_ai-1.0.0/archive_ai.egg-info → archive_ai-1.0.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: archive-ai
-Version: 1.0.0
+Version: 1.0.2
 Summary: A Django app for AI and OCR logic.
 Home-page: https://github.com/debeski1/archive-ai-app
 Author: Debeski

{archive_ai-1.0.0 → archive_ai-1.0.2}/README.md

@@ -1,25 +1,25 @@
-# Archive AI
-
-A Django app that provides AI and OCR capabilities.
-
-## Installation
-
-1. Install via pip:
-   ```bash
-   pip install archive-ai
-   ```
-
-2. Add to `INSTALLED_APPS`:
-   ```python
-   INSTALLED_APPS = [
-       ...
-       'archive_ai',
-   ]
-   ```
-
-3. Configure settings:
-   ```python
-   OCR_API_URL = "http://ocr-service:8000/ocr"
-   LLM_API_URL = "http://llm-service:8000/v1/chat/completions"
-   LLM_MODEL = "your-model-name"
-   ```
+# Archive AI
+
+A Django app that provides AI and OCR capabilities.
+
+## Installation
+
+1. Install via pip:
+   ```bash
+   pip install archive-ai
+   ```
+
+2. Add to `INSTALLED_APPS`:
+   ```python
+   INSTALLED_APPS = [
+       ...
+       'archive_ai',
+   ]
+   ```
+
+3. Configure settings:
+   ```python
+   OCR_API_URL = "http://ocr-service:8000/ocr"
+   LLM_API_URL = "http://llm-service:8000/v1/chat/completions"
+   LLM_MODEL = "your-model-name"
+   ```

{archive_ai-1.0.0 → archive_ai-1.0.2}/archive_ai/ai.py

@@ -1,213 +1,251 @@
-# Fundemental imports
-#####################################################################
-from django.utils import timezone
-from django.conf import settings
-from django.apps import apps
-from django.core.exceptions import FieldDoesNotExist
-import difflib
-import os
-import re
-import json
-import requests
-import logging
-
-logger = logging.getLogger(__name__)
-
-def extract_metadata_with_llm(model_name, ocr_text):
-    url = settings.LLM_API_URL
-
-    system_prompt = """
-You are a strict JSON generator.
-
-Analyze Arabic OCR text and extract structured metadata for one document.
-
-Respond with EXACTLY one JSON object.
-No explanations, no comments, no markdown, no text outside the JSON.
-
-Current document type: {model_name}
-
-Output format:
-{
-  "title": "",
-  "body": "",
-
-  "document_number": "",
-  "og_document_number": "",
-
-  "document_date": "",
-  "og_document_date": "",
-
-  "source": "",
-  "destination": ""
-}
-
-Rules:
-
-1. Document types:
-- decree: formal decree, numbered articles, legal references, decree number, subject, footer date.
-- incoming: original letter + received stamp (usually 2 dates, 2 numbers), source and destination, letter body.
-- outgoing / internal / other: usually one date and one number. Source is a ministry department, destination is another entity.
-
-2. Dates:
-- One date → document_date
-- Two dates → older (letter header) = og_document_date, newer (stamp) = document_date
-
-3. Numbers:
-- document_number:
-  - outgoing/internal/other → official document number
-  - incoming → received stamp number
-- og_document_number:
-  - incoming only → original document number
-- Number fields MUST always contain digits only.
-- DO NOT use Eastern Arabic numerals (like ٠١٢٣) or any other numeral system.
-- DO NOT include letters. If OCR included letters with the numbers, REMOVE the letters and keep the numbers.
-
-4. Source & Destination:
-- Extract real entity names only. Clean OCR noise but do NOT invent names.
-- Incoming source: organization in the header.
-- Outgoing source: department under the signer’s name.
-- Destination often appears near the start, commonly after the words sir or sirs and a "/".
-
-5. Title:
-- Extract explicit subject/title if present (will always be in the beggining if present).
-- Otherwise infer a short, accurate phrase from the document text.
-
-6. Body:
-- Return the full cleaned document text with OCR noise removed.
-
-Always fill all fields.
-If unknown or not applicable, use an empty string "".
-
-Return ONLY the JSON object.
-"""
-
-    user_prompt = f"""
-OCR TEXT:
-\"\"\"
-{ocr_text}
-\"\"\"
-"""
-
-    payload = {
-        "model": "/models/" + settings.LLM_MODEL,
-        "messages": [
-            {"role": "system", "content": system_prompt},
-            {"role": "user", "content": user_prompt}
-        ],
-        "temperature": 0.5,
-        "max_tokens": 1024,
-        "response_format": {"type": "json_object"},
-        "top_p": 0.8,
-        "min_p": 0,
-        "top_k": 20,
-        "presence_penalty": 0.5,
-    }
-
-    # ------------------------------------------------
-    # Call the LLM API
-    # ------------------------------------------------
-    try:
-        response = requests.post(url, json=payload, timeout=(10, 1200))
-        response.raise_for_status()
-    except requests.exceptions.RequestException:
-        raise
-
-    data = response.json()
-
-    if "choices" not in data or not data["choices"]:
-        raise ValueError("LLM API returned no choices.")
-
-    raw_content = data["choices"][0]["message"]["content"]
-
-    # ------------------------------------------------
-    # Extract JSON safely
-    # ------------------------------------------------
-    metadata = extract_json_from_text(raw_content)
-
-    # Ensure required keys exist
-    required = ["source", "destination", "document_number", "document_date", "title", "body"]
-    for key in required:
-        metadata.setdefault(key, "")
-
-    return metadata
-
-def extract_json_from_text(text):
-    """
-    Safely extracts the first JSON object from LLM output.
-    Handles cases like:
-    - ```json {...} ```
-    - text before/after JSON
-    - invalid trailing content
-    """
-    match = re.search(r"\{.*\}", text, flags=re.DOTALL)
-    if not match:
-        raise ValueError("No JSON object found in LLM response.")
-
-    json_str = match.group()
-
-    try:
-        return json.loads(json_str)
-    except json.JSONDecodeError as exc:
-        raise ValueError(f"Malformed JSON: {exc}")
-
-def validate_date(date_str):
-    """Return a cleaned YYYY-MM-DD or None."""
-    if not date_str:
-        return None
-    try:
-        timezone.datetime.strptime(date_str, "%Y-%m-%d")
-        return date_str
-    except Exception:
-        return None
-
-def _find_best_match(input_text, model_name, field_name, cutoff=60):
-
-    if not input_text:
-        return None
-
-    text = str(input_text).strip()
-    
-    app_label = getattr(settings, "OCR_APP_LABEL", "documents")
-
-    try:
-        model = apps.get_model(app_label, model_name)
-    except LookupError:
-        logger.error(f"Model {model_name} not found")
-        return None
-
-    try:
-        field = model._meta.get_field(field_name)
-    except FieldDoesNotExist:
-        logger.error(f"Field {field_name} not found in model {model_name}")
-        return None
-
-    related_model = field.related_model
-    possible_fields = ['source', 'destination']
-    
-    for field in possible_fields:
-        if hasattr(related_model, field):
-            # Get all values for this field
-            choices = list(related_model.objects.values_list('id', field))
-            
-            if choices:
-                # Extract just the text values for matching
-                values = [value for _, value in choices]
-                
-                # Use difflib to find closest match
-                matches = difflib.get_close_matches(text, values, n=1, cutoff=cutoff)
-                
-                if matches:
-                    best_match = matches[0]
-                    # Find and return the corresponding ID
-                    for id, value in choices:
-                        if value == best_match:
-                            return id
-    
-    return None
-
-def find_closest_source(raw_text, model_name):
-    return _find_best_match(raw_text, model_name, "source")
-
-
-def find_closest_destination(raw_text, model_name):
-    return _find_best_match(raw_text, model_name, "destination")
+# Fundemental imports
+#####################################################################
+from django.utils import timezone
+from django.conf import settings
+from django.apps import apps
+from django.core.exceptions import FieldDoesNotExist
+import difflib
+import os
+import re
+import json
+import requests
+import logging
+
+logger = logging.getLogger(__name__)
+
+def extract_metadata_with_llm(model_name, ocr_text):
+    url = settings.LLM_API_URL
+
+    system_prompt = """
+You are a strict JSON generator.
+
+Analyze Arabic OCR text and extract structured metadata for one document.
+
+Respond with EXACTLY one JSON object.
+No explanations, no comments, no markdown, no text outside the JSON.
+
+Current document type: {model_name}
+
+Output format:
+{
+  "title": "",
+  "body": "",
+
+  "document_number": "",
+  "og_document_number": "",
+
+  "document_date": "",
+  "og_document_date": "",
+
+  "source": "",
+  "destination": ""
+}
+
+Rules:
+
+1. Document types:
+- decree: formal decree, numbered articles, legal references, decree number, subject, footer date.
+- incoming: original letter + received stamp (usually 2 dates, 2 numbers), source and destination, letter body.
+- outgoing / internal / other: usually one date and one number. Source is a ministry department, destination is another entity.
+
+2. Dates:
+- One date → document_date
+- Two dates → older (letter header) = og_document_date, newer (stamp) = document_date
+
+3. Numbers:
+- document_number:
+  - outgoing/internal/other → official document number
+  - incoming → received stamp number
+- og_document_number:
+  - incoming only → original document number
+- Number fields MUST always contain digits only.
+- DO NOT use Eastern Arabic numerals (like ٠١٢٣) or any other numeral system.
+- DO NOT include letters. If OCR included letters with the numbers, REMOVE the letters and keep the numbers.
+
+4. Source & Destination:
+- Extract real entity names only. Clean OCR noise but do NOT invent names.
+- Incoming source: organization in the header.
+- Outgoing source: department under the signer’s name.
+- Destination often appears near the start, commonly after the words sir or sirs and a "/".
+
+5. Title:
+- Extract explicit subject/title if present (will always be in the beggining if present).
+- Otherwise infer a short, accurate phrase from the document text.
+
+6. Body:
+- Return the full cleaned document text with OCR noise removed.
+
+Always fill all fields.
+If unknown or not applicable, use an empty string "".
+
+Return ONLY the JSON object.
+"""
+
+    user_prompt = f"""
+OCR TEXT:
+\"\"\"
+{ocr_text}
+\"\"\"
+"""
+
+    payload = {
+        "model": "/models/" + settings.LLM_MODEL,
+        "messages": [
+            {"role": "system", "content": system_prompt},
+            {"role": "user", "content": user_prompt}
+        ],
+        "temperature": 0.5,
+        "max_tokens": 1024,
+        "response_format": {"type": "json_object"},
+        "top_p": 0.8,
+        "min_p": 0,
+        "top_k": 20,
+        "presence_penalty": 0.5,
+    }
+
+    # ------------------------------------------------
+    # Call the LLM API
+    # ------------------------------------------------
+    try:
+        response = requests.post(url, json=payload, timeout=(10, 1200))
+        response.raise_for_status()
+    except requests.exceptions.RequestException:
+        raise
+
+    data = response.json()
+
+    if "choices" not in data or not data["choices"]:
+        raise ValueError("LLM API returned no choices.")
+
+    raw_content = data["choices"][0]["message"]["content"]
+
+    # ------------------------------------------------
+    # Extract JSON safely
+    # ------------------------------------------------
+    metadata = extract_json_from_text(raw_content)
+
+
+
+    # Ensure required keys exist
+    required = ["source", "destination", "document_number", "document_date", "title", "body"]
+    for key in required:
+        metadata.setdefault(key, "")
+
+    # Sanitize number fields (strict integer enforcement)
+    # Even if the AI hallucinates "Ref: 123", we want "123".
+    # If "No number", we want "".
+    metadata["document_number"] = clean_number_field(metadata["document_number"])
+    if "og_document_number" in metadata:
+         metadata["og_document_number"] = clean_number_field(metadata["og_document_number"])
+
+    return metadata
+
+def extract_json_from_text(text):
+    """
+    Safely extracts the first JSON object from LLM output.
+    Handles cases like:
+    - ```json {...} ```
+    - text before/after JSON
+    - invalid trailing content
+    """
+    match = re.search(r"\{.*\}", text, flags=re.DOTALL)
+    if not match:
+        raise ValueError("No JSON object found in LLM response.")
+
+    json_str = match.group()
+
+    try:
+        return json.loads(json_str)
+    except json.JSONDecodeError as exc:
+        raise ValueError(f"Malformed JSON: {exc}")
+
+def validate_date(date_str):
+    """Return a cleaned YYYY-MM-DD or None."""
+    if not date_str:
+        return None
+    try:
+        timezone.datetime.strptime(date_str, "%Y-%m-%d")
+        return date_str
+    except Exception:
+        return None
+
+def _find_best_match(input_text, model_name, field_name, cutoff=60):
+
+    if not input_text:
+        return None
+
+    text = str(input_text).strip()
+    
+    app_label = getattr(settings, "OCR_APP_LABEL", "documents")
+
+    try:
+        model = apps.get_model(app_label, model_name)
+    except LookupError:
+        logger.error(f"Model {model_name} not found")
+        return None
+
+    try:
+        field = model._meta.get_field(field_name)
+    except FieldDoesNotExist:
+        logger.error(f"Field {field_name} not found in model {model_name}")
+        return None
+
+    related_model = field.related_model
+    possible_fields = ['source', 'destination']
+    
+    for field in possible_fields:
+        if hasattr(related_model, field):
+            # Get all values for this field
+            choices = list(related_model.objects.values_list('id', field))
+            
+            if choices:
+                # Extract just the text values for matching
+                values = [value for _, value in choices]
+                
+                # Use difflib to find closest match
+                matches = difflib.get_close_matches(text, values, n=1, cutoff=cutoff)
+                
+                if matches:
+                    best_match = matches[0]
+                    # Find and return the corresponding ID
+                    for id, value in choices:
+                        if value == best_match:
+                            return id
+    
+    return None
+
+def find_closest_source(raw_text, model_name):
+    return _find_best_match(raw_text, model_name, "source")
+
+
+def find_closest_destination(raw_text, model_name):
+    return _find_best_match(raw_text, model_name, "destination")
+
+def clean_number_field(value):
+    """
+    Removes all non-digit characters from the string.
+    Also handles common OCR substitutions (letters that look like numbers).
+    Returns a string of digits, or empty string if no digits found.
+    """
+    if not value:
+        return ""
+    
+    val_str = str(value)
+    
+    # Common OCR substitutions map
+    substitutions = {
+        's': '5', 'S': '5',
+        'z': '7', 'Z': '7',
+        'b': '8', 'B': '8',
+        'l': '1', 'I': '1', 'i': '1',
+        'o': '0', 'O': '0',
+    }
+    
+    # Apply substitutions
+    for char, digit in substitutions.items():
+        val_str = val_str.replace(char, digit)
+    
+    # Filter only digits
+    cleaned = "".join(filter(str.isdigit, val_str))
+    
+    return cleaned

{archive_ai-1.0.0 → archive_ai-1.0.2}/archive_ai/apps.py

@@ -1,8 +1,8 @@
-from django.apps import AppConfig
-
-class SmartArchiveConfig(AppConfig):
-    default_auto_field = 'django.db.models.BigAutoField'
-    name = 'smart_archive'
-
-    def ready(self):
-        import smart_archive.signals
+from django.apps import AppConfig
+
+class ArchiveAiConfig(AppConfig):
+    default_auto_field = 'django.db.models.BigAutoField'
+    name = 'archive_ai'
+
+    def ready(self):
+        import archive_ai.signals

{archive_ai-1.0.0 → archive_ai-1.0.2}/archive_ai/signals.py

@@ -1,29 +1,29 @@
-from django.db.models.signals import post_save
-from django.dispatch import receiver
-from .tasks import send_pdf_to_ocr
-
-# Defines which models should trigger the OCR.
-# Ideally this should be configurable, but for now we follow the original list.
-MODELS = [
-    "Decree",
-    "Incoming",
-    "Outgoing",
-    "Internal",
-    "Report",
-    "Other",
-]
-
-@receiver(post_save)
-def trigger_ocr(sender, instance, created, **kwargs):
-    if not created:
-        return
-    if sender.__name__ not in MODELS:
-        return
-
-    # All your models have a 'pdf' field but if the name differs,
-    # adjust here.
-    send_pdf_to_ocr.delay(
-        model_name=sender.__name__,
-        obj_id=instance.id,
-        pdf_field_name="pdf_file"   # change if your field name differs
-    )
+from django.db.models.signals import post_save
+from django.dispatch import receiver
+from .tasks import send_pdf_to_ocr
+
+# Defines which models should trigger the OCR.
+# Ideally this should be configurable, but for now we follow the original list.
+MODELS = [
+    "Decree",
+    "Incoming",
+    "Outgoing",
+    "Internal",
+    "Report",
+    "Other",
+]
+
+@receiver(post_save)
+def trigger_ocr(sender, instance, created, **kwargs):
+    if not created:
+        return
+    if sender.__name__ not in MODELS:
+        return
+
+    # All your models have a 'pdf' field but if the name differs,
+    # adjust here.
+    send_pdf_to_ocr.delay(
+        model_name=sender.__name__,
+        obj_id=instance.id,
+        pdf_field_name="pdf_file"   # change if your field name differs
+    )

{archive_ai-1.0.0 → archive_ai-1.0.2}/archive_ai/tasks.py

@@ -1,111 +1,111 @@
-from celery import shared_task
-from django.apps import apps
-from django.conf import settings
-import requests
-import os
-import logging
-from .ai import extract_metadata_with_llm, validate_date, find_closest_source, find_closest_destination
-
-logger = logging.getLogger(__name__)
-
-@shared_task(bind=True, max_retries=3, default_retry_delay=120, soft_time_limit=800, time_limit=1200)
-def send_pdf_to_ocr(self, model_name, obj_id, pdf_field_name):
-    """
-    Uploads a PDF to the OCR API, processes text through LLM metadata extractor,
-    and updates the related Django model.
-
-    Retries ONLY for network-related issues.
-    """
-    
-    app_label = getattr(settings, "OCR_APP_LABEL", "documents")
-    Model = apps.get_model(app_label, model_name)
-
-    try:
-        obj = Model.objects.get(id=obj_id)
-    except Model.DoesNotExist:
-        logger.error(f"[OCR TASK] Object {obj_id} of type {model_name} does not exist.")
-        return False
-
-    pdf_field = getattr(obj, pdf_field_name)
-    pdf_path = pdf_field.path
-
-    # ------------------------------------------------
-    # Ensure PDF exists
-    # ------------------------------------------------
-    if not os.path.exists(pdf_path):
-        logger.error(f"[OCR TASK] PDF file missing: {pdf_path}")
-        return False
-
-    # ------------------------------------------------
-    # 1. Send to OCR API
-    # ------------------------------------------------
-    try:
-        with open(pdf_path, "rb") as f:
-            response = requests.post(
-                settings.OCR_API_URL,
-                files={"file": f},
-                data={"rotation": 0, "mode": "Markdown"},
-                timeout=1200,
-            )
-        response.raise_for_status()
-    except requests.exceptions.RequestException as exc:
-        # Retry only real network failures
-        logger.warning(f"[OCR TASK] Network error, retrying: {exc}")
-        raise self.retry(exc=exc)
-
-    try:
-        ocr_text = response.json().get("text", "")
-    except Exception as exc:
-        logger.error(f"[OCR TASK] Invalid OCR response JSON: {exc}")
-        return False
-
-    # ------------------------------------------------
-    # 2. Extract metadata using LLM
-    # ------------------------------------------------
-    try:
-        metadata = extract_metadata_with_llm(model_name, ocr_text)
-    except requests.exceptions.RequestException as exc:
-        logger.warning(f"[OCR TASK] LLM network failure, retrying: {exc}")
-        raise self.retry(exc=exc)
-    except Exception as exc:
-        logger.error(f"[OCR TASK] Metadata parsing failed: {exc}")
-        return False
-
-    # ------------------------------------------------
-    # 3. Apply metadata safely
-    # ------------------------------------------------
-    try:
-        # Universal fields
-        obj.title = metadata.get("title") or obj.title
-        obj.text = metadata.get("body", obj.text)
-
-        # Set number depending on doc type
-        if metadata.get("doc_type") == "incoming":
-            obj.number = metadata.get("og_document_number") or obj.number
-            obj.received_number = metadata.get("document_number") or obj.received_number
-        else:
-            obj.number = metadata.get("document_number") or obj.number
-
-        # Dates
-        doc_date = validate_date(metadata.get("document_date", ""))
-        og_date = validate_date(metadata.get("og_document_date", ""))
-
-        if metadata.get("doc_type") == "incoming":
-            if og_date:
-                obj.og_date = og_date
-        else:
-            if doc_date:
-                obj.date = doc_date
-
-        # Source / Destination logic if needed
-        # (This part was commented out in original, keeping it that way or enabling based on user needs. 
-        # I'll keep it commented out to match original behavior, but ensure imports are available if they uncomment)
-        
-        obj.save()
-
-    except Exception as exc:
-        logger.error(f"[OCR TASK] Failed to update object {obj_id}: {exc}")
-        return False
-
-    logger.info(f"[OCR TASK] Completed successfully for obj {obj_id}")
-    return True
+from celery import shared_task
+from django.apps import apps
+from django.conf import settings
+import requests
+import os
+import logging
+from .ai import extract_metadata_with_llm, validate_date, find_closest_source, find_closest_destination
+
+logger = logging.getLogger(__name__)
+
+@shared_task(bind=True, max_retries=3, default_retry_delay=120, soft_time_limit=800, time_limit=1200)
+def send_pdf_to_ocr(self, model_name, obj_id, pdf_field_name):
+    """
+    Uploads a PDF to the OCR API, processes text through LLM metadata extractor,
+    and updates the related Django model.
+
+    Retries ONLY for network-related issues.
+    """
+    
+    app_label = getattr(settings, "OCR_APP_LABEL", "documents")
+    Model = apps.get_model(app_label, model_name)
+
+    try:
+        obj = Model.objects.get(id=obj_id)
+    except Model.DoesNotExist:
+        logger.error(f"[OCR TASK] Object {obj_id} of type {model_name} does not exist.")
+        return False
+
+    pdf_field = getattr(obj, pdf_field_name)
+    pdf_path = pdf_field.path
+
+    # ------------------------------------------------
+    # Ensure PDF exists
+    # ------------------------------------------------
+    if not os.path.exists(pdf_path):
+        logger.error(f"[OCR TASK] PDF file missing: {pdf_path}")
+        return False
+
+    # ------------------------------------------------
+    # 1. Send to OCR API
+    # ------------------------------------------------
+    try:
+        with open(pdf_path, "rb") as f:
+            response = requests.post(
+                settings.OCR_API_URL,
+                files={"file": f},
+                data={"rotation": 0, "mode": "Markdown"},
+                timeout=1200,
+            )
+        response.raise_for_status()
+    except requests.exceptions.RequestException as exc:
+        # Retry only real network failures
+        logger.warning(f"[OCR TASK] Network error, retrying: {exc}")
+        raise self.retry(exc=exc)
+
+    try:
+        ocr_text = response.json().get("text", "")
+    except Exception as exc:
+        logger.error(f"[OCR TASK] Invalid OCR response JSON: {exc}")
+        return False
+
+    # ------------------------------------------------
+    # 2. Extract metadata using LLM
+    # ------------------------------------------------
+    try:
+        metadata = extract_metadata_with_llm(model_name, ocr_text)
+    except requests.exceptions.RequestException as exc:
+        logger.warning(f"[OCR TASK] LLM network failure, retrying: {exc}")
+        raise self.retry(exc=exc)
+    except Exception as exc:
+        logger.error(f"[OCR TASK] Metadata parsing failed: {exc}")
+        return False
+
+    # ------------------------------------------------
+    # 3. Apply metadata safely
+    # ------------------------------------------------
+    try:
+        # Universal fields
+        obj.title = metadata.get("title") or obj.title
+        obj.text = metadata.get("body", obj.text)
+
+        # Set number depending on doc type
+        if metadata.get("doc_type") == "incoming":
+            obj.number = metadata.get("og_document_number") or obj.number
+            obj.received_number = metadata.get("document_number") or obj.received_number
+        else:
+            obj.number = metadata.get("document_number") or obj.number
+
+        # Dates
+        doc_date = validate_date(metadata.get("document_date", ""))
+        og_date = validate_date(metadata.get("og_document_date", ""))
+
+        if metadata.get("doc_type") == "incoming":
+            if og_date:
+                obj.og_date = og_date
+        else:
+            if doc_date:
+                obj.date = doc_date
+
+        # Source / Destination logic if needed
+        # (This part was commented out in original, keeping it that way or enabling based on user needs. 
+        # I'll keep it commented out to match original behavior, but ensure imports are available if they uncomment)
+        
+        obj.save()
+
+    except Exception as exc:
+        logger.error(f"[OCR TASK] Failed to update object {obj_id}: {exc}")
+        return False
+
+    logger.info(f"[OCR TASK] Completed successfully for obj {obj_id}")
+    return True

{archive_ai-1.0.0 → archive_ai-1.0.2/archive_ai.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: archive-ai
-Version: 1.0.0
+Version: 1.0.2
 Summary: A Django app for AI and OCR logic.
 Home-page: https://github.com/debeski1/archive-ai-app
 Author: Debeski

{archive_ai-1.0.0 → archive_ai-1.0.2}/setup.py

@@ -1,29 +1,29 @@
-from setuptools import setup, find_packages
-
-setup(
-    name="archive-ai",
-    version="1.0.0",
-    packages=find_packages(),
-    include_package_data=True,
-    license="MIT",
-    description="A Django app for AI and OCR logic.",
-    long_description=open("README.md").read(),
-    long_description_content_type="text/markdown",
-    url="https://github.com/debeski1/archive-ai-app",
-    author="Debeski",
-    author_email="debeski1@gmail.com",
-    classifiers=[
-        "Environment :: Web Environment",
-        "Framework :: Django",
-        "Intended Audience :: Developers",
-        "License :: OSI Approved :: MIT License",
-        "Operating System :: OS Independent",
-        "Programming Language :: Python",
-        "Programming Language :: Python :: 3",
-    ],
-    install_requires=[
-        "Django>=3.2",
-        "requests",
-        "celery",
-    ],
-)
+from setuptools import setup, find_packages
+
+setup(
+    name="archive-ai",
+    version="1.0.2",
+    packages=find_packages(),
+    include_package_data=True,
+    license="MIT",
+    description="A Django app for AI and OCR logic.",
+    long_description=open("README.md").read(),
+    long_description_content_type="text/markdown",
+    url="https://github.com/debeski1/archive-ai-app",
+    author="Debeski",
+    author_email="debeski1@gmail.com",
+    classifiers=[
+        "Environment :: Web Environment",
+        "Framework :: Django",
+        "Intended Audience :: Developers",
+        "License :: OSI Approved :: MIT License",
+        "Operating System :: OS Independent",
+        "Programming Language :: Python",
+        "Programming Language :: Python :: 3",
+    ],
+    install_requires=[
+        "Django>=3.2",
+        "requests",
+        "celery",
+    ],
+)