keannu125/DocManagerBackend
1
0
Fork 0
mirror of https://github.com/lemeow125/DocManagerBackend.git synced 2025-01-19 01:23:02 +08:00
Code Issues Projects Releases Packages Wiki Activity Actions

Improve start_watcher OCR

Browse source
This commit is contained in:
Keannu Bernasol 2025-01-09 14:53:40 +08:00
parent 9529560fed
commit dba3f6df62
1 changed files with 59 additions and 11 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

70
docmanager_backend/config/management/commands/start_watcher.py
View file

@ -97,7 +97,7 @@ class PDFHandler(FileSystemEventHandler):
encoded_image = base64.b64encode( encoded_image = base64.b64encode(
img_buffer.getvalue()).decode() img_buffer.getvalue()).decode()
# First LLM API call to determine category # Determine category
class DocumentSchema(BaseModel): class DocumentSchema(BaseModel):
category: str = "other" category: str = "other"
explanation: Optional[str] = None explanation: Optional[str] = None
@ -109,8 +109,6 @@ class PDFHandler(FileSystemEventHandler):
Possible document types are: {possible_categories}. You are free to create a new one if none are suitable. Possible document types are: {possible_categories}. You are free to create a new one if none are suitable.
If the document_type is Special Order or Memorandum, provide the sender of the document under sent_from.
Do all of this and return your output in JSON. Do all of this and return your output in JSON.
""" """
@ -130,20 +128,14 @@ class PDFHandler(FileSystemEventHandler):
response.message.content) response.message.content)
document_type = result.category document_type = result.category
# Second LLM API call to determine other details # Determine sender
class DocumentSchema(BaseModel): class DocumentSchema(BaseModel):
sent_from: str = "N/A" sent_from: str = "N/A"
subject: str = "N/A"
document_date: Optional[date]
explanation: Optional[str] = None explanation: Optional[str] = None
prompt = f""" prompt = f"""
Determine who sent the document. Otherwise, return N/A. Determine who sent the document. Otherwise, return N/A.
Identify the subject or possible title of the document.
Return the date of the document if it exists.
Do all of this and return your output in JSON. Do all of this and return your output in JSON.
""" """
response = client.chat( response = client.chat(
@ -162,6 +154,61 @@ class PDFHandler(FileSystemEventHandler):
response.message.content) response.message.content)
sent_from = result.sent_from sent_from = result.sent_from
# Determine subject
class DocumentSchema(BaseModel):
subject: str = "N/A"
explanation: Optional[str] = None
prompt = f"""
Identify the subject of the document if it exists.
Do all of this and return your output in JSON.
"""
response = client.chat(
model=get_secret("OLLAMA_MODEL"),
messages=[
{"role": "user",
"content": prompt,
"images": [encoded_image]},
],
format=DocumentSchema.model_json_schema(),
options={
"temperature": 0
},
)
result = DocumentSchema.model_validate_json(
response.message.content)
document_subject = result.subject
# Determine date
class DocumentSchema(BaseModel):
document_date: Optional[date]
explanation: Optional[str] = None
prompt = f"""
Identify the date of the document if it exists.
If you are unable to determine the date, return nothing.
Do all of this and return your output in JSON.
"""
response = client.chat(
model=get_secret("OLLAMA_MODEL"),
messages=[
{"role": "user",
"content": prompt,
"images": [encoded_image]},
],
format=DocumentSchema.model_json_schema(),
options={
"temperature": 0
},
)
result = DocumentSchema.model_validate_json(
response.message.content)
document_date = result.document_date document_date = result.document_date
if document_date: if document_date:
@ -199,7 +246,8 @@ class PDFHandler(FileSystemEventHandler):
document_type=document_type, document_type=document_type,
sent_from=sent_from, sent_from=sent_from,
document_month=document_month, document_month=document_month,
document_year=document_year document_year=document_year,
subject=document_subject
) )
DOCUMENT.file.save( DOCUMENT.file.save(