Implement sender LLM OCR and request remarks

This commit is contained in:
Keannu Bernasol 2025-01-08 13:38:39 +08:00
parent d5477a1e1a
commit 674a7ec592
11 changed files with 174 additions and 62 deletions

View file

@ -69,8 +69,8 @@ class PDFHandler(FileSystemEventHandler):
with fitz.open(file_path) as doc:
num_pages = len(doc)
for page_num in range(num_pages):
page = doc[page_num]
# Perform OCR only on the first page
page = doc[0]
pix = page.get_pixmap(matrix=(1.2, 1.2))
# Convert pixmap to bytes
@ -89,6 +89,7 @@ class PDFHandler(FileSystemEventHandler):
try:
class DocumentCategory(BaseModel):
category: str = "other"
sent_from: str = "N/A"
explanation: Optional[str] = None
client = Client(
@ -101,11 +102,14 @@ class PDFHandler(FileSystemEventHandler):
img_buffer.getvalue()).decode()
possible_categories = set((Document.objects.all().values_list(
"document_type", flat=True), "Documented Procedures Manual", "Form", "Special Order"))
"document_type", flat=True), "Documented Procedures Manual", "Form", "Special Order", "Memorandum"))
prompt = f"""
Read the text from the image and provide a category. Return as JSON.
Possible categories are: {possible_categories}. You are free to create a new one if none are suitable.
If the document is of type Special Order or Memorandum, provide the sender of the document. Possible senders are Vice President, President, Chancellor.
provide N/A.
"""
response = client.chat(
model=get_secret("OLLAMA_MODEL"),
@ -125,6 +129,7 @@ class PDFHandler(FileSystemEventHandler):
response.message.content)
result = json.loads(response.message.content)
document_type = result.get("category")
sent_from = result.get("sent_from")
# If that fails, just use regular OCR read the title as a dirty fix/fallback
except Exception as e:
@ -158,8 +163,10 @@ class PDFHandler(FileSystemEventHandler):
name=filename, content=File(open(file_path, "rb")))
self.logger.info(
f"Document '{filename}' created successfully with type '{
document_type}'."
document_type}'. sent_from: {sent_from}"
)
DOCUMENT.sent_from = sent_from
DOCUMENT.save()
else:
self.logger.info(f"Document '{filename}' already exists.")

View file

@ -0,0 +1,18 @@
# Generated by Django 5.1.3 on 2025-01-08 04:29
from django.db import migrations, models
class Migration(migrations.Migration):
dependencies = [
("document_requests", "0002_documentrequest_questionnaire"),
]
operations = [
migrations.AddField(
model_name="documentrequest",
name="denied_remarks",
field=models.TextField(blank=True, max_length=512, null=True),
),
]

View file

@ -0,0 +1,18 @@
# Generated by Django 5.1.3 on 2025-01-08 04:51
from django.db import migrations
class Migration(migrations.Migration):
dependencies = [
("document_requests", "0003_documentrequest_denied_remarks"),
]
operations = [
migrations.RenameField(
model_name="documentrequest",
old_name="denied_remarks",
new_name="remarks",
),
]

View file

@ -27,6 +27,8 @@ class DocumentRequest(models.Model):
("denied", "Denied"),
)
remarks = models.TextField(max_length=512, blank=True, null=True)
status = models.CharField(
max_length=32, choices=STATUS_CHOICES, default="pending")

View file

@ -101,6 +101,7 @@ class DocumentRequestSerializer(serializers.ModelSerializer):
"purpose",
"date_requested",
"documents",
"remarks",
"status",
]
read_only_fields = [
@ -112,6 +113,7 @@ class DocumentRequestSerializer(serializers.ModelSerializer):
"purpose",
"date_requested",
"documents",
"remarks,"
"status",
]
@ -146,6 +148,7 @@ class FullDocumentRequestSerializer(serializers.ModelSerializer):
"purpose",
"date_requested",
"documents",
"remarks",
"status",
]
read_only_fields = [
@ -167,27 +170,41 @@ class DocumentRequestUpdateSerializer(serializers.ModelSerializer):
class Meta:
model = DocumentRequest
fields = ["id", "status"]
read_only_fields = ["id", "status"]
fields = ["id", "status", "remarks"]
read_only_fields = ["id"]
def update(self, instance, validated_data):
print(validated_data)
if instance.status == "denied" or instance.status == "approved":
raise serializers.ValidationError(
{
"error": "Already approved/denied requests cannot be updated. You should instead create a new request and approve it from there"
}
)
elif "status" not in validated_data:
raise serializers.ValidationError(
{
"error": "No status value update provided"
}
)
elif validated_data["status"] == instance.status:
raise serializers.ValidationError(
{"error": "Request form status provided is the same as current status"}
)
elif validated_data["status"] == "denied" and "remarks" not in validated_data:
raise serializers.ValidationError(
{"error": "Request denial requires remarks"}
)
representation = super().update(instance, validated_data)
# Send an email on request status update
try:
email = RequestUpdateEmail()
email.context = {"request_status": instance.status}
if instance.status == "denied":
email.context = {"remarks": instance.remarks}
else:
email.context = {"remarks": "N/A"}
email.send(to=[instance.requester.email])
except:
# Silence out errors if email sending fails

View file

@ -0,0 +1,18 @@
# Generated by Django 5.1.3 on 2025-01-08 04:39
from django.db import migrations, models
class Migration(migrations.Migration):
dependencies = [
("documents", "0002_alter_document_document_type"),
]
operations = [
migrations.AddField(
model_name="document",
name="memorandum_from",
field=models.CharField(blank=True, max_length=128, null=True),
),
]

View file

@ -0,0 +1,18 @@
# Generated by Django 5.1.3 on 2025-01-08 04:44
from django.db import migrations
class Migration(migrations.Migration):
dependencies = [
("documents", "0003_document_memorandum_from"),
]
operations = [
migrations.RenameField(
model_name="document",
old_name="memorandum_from",
new_name="sent_from",
),
]

View file

@ -9,6 +9,9 @@ class Document(models.Model):
document_type = models.CharField(
max_length=128, null=False, blank=False
)
sent_from = models.CharField(
max_length=128, null=True, blank=True
)
number_pages = models.IntegerField(null=False, blank=False)
ocr_metadata = models.TextField(null=True, blank=True)

View file

@ -53,6 +53,7 @@ class DocumentSerializer(serializers.ModelSerializer):
"document_type",
"number_pages",
"ocr_metadata",
"sent_from",
"date_uploaded",
]
read_only_fields = [
@ -61,6 +62,7 @@ class DocumentSerializer(serializers.ModelSerializer):
"document_type",
"number_pages",
"ocr_metadata",
"sent_from",
"date_uploaded",
]
@ -81,6 +83,7 @@ class DocumentFileSerializer(serializers.ModelSerializer):
"number_pages",
"ocr_metadata",
"date_uploaded",
"sent_from",
"file",
]
read_only_fields = [
@ -90,5 +93,6 @@ class DocumentFileSerializer(serializers.ModelSerializer):
"number_pages",
"ocr_metadata",
"date_uploaded",
"sent_from",
"file",
]

View file

@ -8,6 +8,7 @@ class RequestUpdateEmail(email.BaseEmailMessage):
def get_context_data(self):
context = super().get_context_data()
context["request_status"] = context.get("request_status")
context["remarks"] = context.get("remarks")
context["url"] = FRONTEND_URL
context.update(self.context)
return context

View file

@ -7,6 +7,8 @@
{% block text_body %}
{% blocktrans %}You're receiving this email because your document request has been {{ request_status }}.{% endblocktrans %}
{% blocktrans %}Remarks: {{ remarks }}{% endblocktrans %}
{% trans 'Please visit the site to check your request:' %}
{{ url|safe }}
@ -18,6 +20,10 @@
{% blocktrans %}You're receiving this email because your document request has been {{ request_status }}.{% endblocktrans %}
</p>
<p>
{% blocktrans %}Remarks {{ remarks }}{% endblocktrans %}
</p>
<p>
{% trans 'Please visit the site to check your request:' %}
</p>