- Improvements on document uploads (accept other files than html-files when entering a URL)

- Introduction of API-functionality (to be continued). Deduplication of document and url uploads between views and api.
- Improvements on document processing - introduction of processor classes to streamline document inputs
- Removed pure Youtube functionality, as Youtube retrieval of documents continuously changes. But added upload of srt, mp3, ogg and mp4
This commit is contained in:
Josako
2024-09-02 12:37:44 +02:00
parent a158655247
commit 914c265afe
21 changed files with 1425 additions and 852 deletions

View File

@@ -18,6 +18,10 @@ from minio.error import S3Error
from common.models.document import Document, DocumentVersion
from common.extensions import db, minio_client
from common.utils.document_utils import validate_file_type, create_document_stack, start_embedding_task, process_url, \
process_multiple_urls, prepare_youtube_document, create_version_for_document, upload_file_for_version
from common.utils.eveai_exceptions import EveAIInvalidLanguageException, EveAIUnsupportedFileType, \
EveAIDoubleURLException, EveAIYoutubeError
from .document_forms import AddDocumentForm, AddURLForm, EditDocumentForm, EditDocumentVersionForm, AddYoutubeForm, \
AddURLsForm
from common.utils.middleware import mw_before_request
@@ -56,30 +60,37 @@ def before_request():
@roles_accepted('Super User', 'Tenant Admin')
def add_document():
form = AddDocumentForm()
current_app.logger.debug('Adding document')
# If the form is submitted
if form.validate_on_submit():
current_app.logger.info(f'Adding document for tenant {session["tenant"]["id"]}')
file = form.file.data
filename = secure_filename(file.filename)
extension = filename.rsplit('.', 1)[1].lower()
form_dict = form_to_dict(form)
try:
current_app.logger.debug('Validating file type')
tenant_id = session['tenant']['id']
file = form.file.data
filename = secure_filename(file.filename)
extension = filename.rsplit('.', 1)[1].lower()
new_doc, new_doc_vers = create_document_stack(form_dict, file, filename, extension)
validate_file_type(extension)
task = current_celery.send_task('create_embeddings', queue='embeddings', args=[
session['tenant']['id'],
new_doc_vers.id,
])
current_app.logger.info(f'Embedding creation started for tenant {session["tenant"]["id"]}, '
f'Document Version {new_doc_vers.id}. '
f'Embedding creation task: {task.id}')
flash(f'Processing on document {new_doc.name}, version {new_doc_vers.id} started. Task ID: {task.id}.',
'success')
api_input = {
'name': form.name.data,
'language': form.language.data,
'user_context': form.user_context.data,
'valid_from': form.valid_from.data
}
return redirect(prefixed_url_for('document_bp.documents'))
else:
form_validation_failed(request, form)
new_doc, new_doc_vers = create_document_stack(api_input, file, filename, extension, tenant_id)
task_id = start_embedding_task(tenant_id, new_doc_vers.id)
flash(f'Processing on document {new_doc.name}, version {new_doc_vers.id} started. Task ID: {task_id}.',
'success')
return redirect(prefixed_url_for('document_bp.documents'))
except (EveAIInvalidLanguageException, EveAIUnsupportedFileType) as e:
flash(str(e), 'error')
except Exception as e:
current_app.logger.error(f'Error adding document: {str(e)}')
flash('An error occurred while adding the document.', 'error')
return render_template('document/add_document.html', form=form)
@@ -90,189 +101,107 @@ def add_url():
form = AddURLForm()
if form.validate_on_submit():
current_app.logger.info(f'Adding url for tenant {session["tenant"]["id"]}')
url = form.url.data
try:
response = requests.head(url, allow_redirects=True)
content_type = response.headers.get('Content-Type', '').split(';')[0]
tenant_id = session['tenant']['id']
url = form.url.data
# Determine file extension based on Content-Type
extension = get_extension_from_content_type(content_type)
file_content, filename, extension = process_url(url, tenant_id)
# Generate filename
parsed_url = urlparse(url)
path = unquote(parsed_url.path)
filename = os.path.basename(path)
api_input = {
'name': form.name.data or filename,
'url': url,
'language': form.language.data,
'user_context': form.user_context.data,
'valid_from': form.valid_from.data
}
if not filename or '.' not in filename:
# Use the last part of the path or a default name
filename = path.strip('/').split('/')[-1] or 'document'
filename = secure_filename(f"{filename}.{extension}")
else:
filename = secure_filename(filename)
new_doc, new_doc_vers = create_document_stack(api_input, file_content, filename, extension, tenant_id)
task_id = start_embedding_task(tenant_id, new_doc_vers.id)
# Check if a document with this URL already exists
existing_doc = DocumentVersion.query.filter_by(url=url).first()
if existing_doc:
flash(f'A document with URL {url} already exists. No new document created.', 'info')
return redirect(prefixed_url_for('document_bp.documents'))
# Download the content
response = requests.get(url)
response.raise_for_status()
file_content = response.content
# Create document and document version
form_dict = form_to_dict(form)
new_doc, new_doc_vers = create_document_stack(form_dict, file_content, filename, extension)
# Upload file to storage
minio_client.upload_document_file(
session['tenant']['id'],
new_doc_vers.doc_id,
new_doc_vers.language,
new_doc_vers.id,
filename,
file_content
)
# Start embedding task
task = current_celery.send_task('create_embeddings', queue='embeddings', args=[
session['tenant']['id'],
new_doc_vers.id,
])
current_app.logger.info(f'Embedding creation started for tenant {session["tenant"]["id"]}, '
f'Document Version {new_doc_vers.id}. '
f'Embedding creation task: {task.id}')
flash(f'Processing on document {new_doc.name}, version {new_doc_vers.id} started. Task ID: {task.id}.',
flash(f'Processing on document {new_doc.name}, version {new_doc_vers.id} started. Task ID: {task_id}.',
'success')
return redirect(prefixed_url_for('document_bp.documents'))
except requests.RequestException as e:
current_app.logger.error(f'Error fetching URL {url}: {str(e)}')
flash(f'Error fetching URL: {str(e)}', 'danger')
except SQLAlchemyError as e:
current_app.logger.error(f'Database error: {str(e)}')
flash('An error occurred while saving the document.', 'danger')
except EveAIDoubleURLException:
flash(f'A document with url {url} already exists. No new document created.', 'info')
except (EveAIInvalidLanguageException, EveAIUnsupportedFileType) as e:
flash(str(e), 'error')
except Exception as e:
current_app.logger.error(f'Unexpected error: {str(e)}')
flash('An unexpected error occurred.', 'danger')
current_app.logger.error(f'Error adding document: {str(e)}')
flash('An error occurred while adding the document.', 'error')
return render_template('document/add_url.html', form=form)
def get_extension_from_content_type(content_type):
content_type_map = {
'text/html': 'html',
'application/pdf': 'pdf',
'text/plain': 'txt',
'application/msword': 'doc',
'application/vnd.openxmlformats-officedocument.wordprocessingml.document': 'docx',
# Add more mappings as needed
}
return content_type_map.get(content_type, 'html') # Default to 'html' if unknown
@document_bp.route('/add_urls', methods=['GET', 'POST'])
@roles_accepted('Super User', 'Tenant Admin')
def add_urls():
form = AddURLsForm()
if form.validate_on_submit():
urls = form.urls.data.split('\n')
urls = [url.strip() for url in urls if url.strip()]
try:
tenant_id = session['tenant']['id']
urls = form.urls.data.split('\n')
urls = [url.strip() for url in urls if url.strip()]
for i, url in enumerate(urls):
try:
doc_vers = DocumentVersion.query.filter_by(url=url).all()
if doc_vers:
current_app.logger.info(f'A document with url {url} already exists. No new document created.')
flash(f'A document with url {url} already exists. No new document created.', 'info')
continue
api_input = {
'name': form.name.data,
'language': form.language.data,
'user_context': form.user_context.data,
'valid_from': form.valid_from.data
}
html = fetch_html(url)
file = io.BytesIO(html)
results = process_multiple_urls(urls, tenant_id, api_input)
parsed_url = urlparse(url)
path_parts = parsed_url.path.split('/')
filename = path_parts[-1] if path_parts[-1] else 'index'
if not filename.endswith('.html'):
filename += '.html'
for result in results:
if result['status'] == 'success':
flash(
f"Processed URL: {result['url']} - Document ID: {result['document_id']}, Version ID: {result['document_version_id']}",
'success')
else:
flash(f"Error processing URL: {result['url']} - {result['message']}", 'error')
# Use the name prefix if provided, otherwise use the filename
doc_name = f"{form.name.data}-{filename}" if form.name.data else filename
return redirect(prefixed_url_for('document_bp.documents'))
new_doc, new_doc_vers = create_document_stack({
'name': doc_name,
'url': url,
'language': form.language.data,
'user_context': form.user_context.data,
'valid_from': form.valid_from.data
}, file, filename, 'html')
task = current_celery.send_task('create_embeddings', queue='embeddings', args=[
session['tenant']['id'],
new_doc_vers.id,
])
current_app.logger.info(f'Embedding creation started for tenant {session["tenant"]["id"]}, '
f'Document Version {new_doc_vers.id}. '
f'Embedding creation task: {task.id}')
flash(f'Processing on document {new_doc.name}, version {new_doc_vers.id} started. Task ID: {task.id}.',
'success')
except Exception as e:
current_app.logger.error(f"Error processing URL {url}: {str(e)}")
flash(f'Error processing URL {url}: {str(e)}', 'danger')
return redirect(prefixed_url_for('document_bp.documents'))
else:
form_validation_failed(request, form)
except Exception as e:
current_app.logger.error(f'Error adding multiple URLs: {str(e)}')
flash('An error occurred while adding the URLs.', 'error')
return render_template('document/add_urls.html', form=form)
@document_bp.route('/add_youtube', methods=['GET', 'POST'])
@roles_accepted('Super User', 'Tenant Admin')
def add_youtube():
form = AddYoutubeForm()
if form.validate_on_submit():
current_app.logger.info(f'Adding Youtube document for tenant {session["tenant"]["id"]}')
url = form.url.data
current_app.logger.debug(f'Value of language field: {form.language.data}')
try:
tenant_id = session['tenant']['id']
url = form.url.data
doc_vers = DocumentVersion.query.filter_by(url=url).all()
if doc_vers:
current_app.logger.info(f'A document with url {url} already exists. No new document created.')
flash(f'A document with url {url} already exists. No new document created.', 'info')
api_input = {
'name': form.name.data,
'language': form.language.data,
'user_context': form.user_context.data,
'valid_from': form.valid_from.data
}
new_doc, new_doc_vers = prepare_youtube_document(url, tenant_id, api_input)
task_id = start_embedding_task(tenant_id, new_doc_vers.id)
flash(
f'Processing on YouTube document {new_doc.name}, version {new_doc_vers.id} started. Task ID: {task_id}.',
'success')
return redirect(prefixed_url_for('document_bp.documents'))
# As downloading a Youtube document can take quite some time, we offload this downloading to the worker
# We just pass a simple file to get things conform
file = "Youtube placeholder file"
filename = 'placeholder.youtube'
extension = 'youtube'
form_dict = form_to_dict(form)
current_app.logger.debug(f'Form data: {form_dict}')
new_doc, new_doc_vers = create_document_stack(form_dict, file, filename, extension)
task = current_celery.send_task('create_embeddings', queue='embeddings', args=[
session['tenant']['id'],
new_doc_vers.id,
])
current_app.logger.info(f'Processing and Embedding on Youtube document started for tenant '
f'{session["tenant"]["id"]}, '
f'Document Version {new_doc_vers.id}. '
f'Processing and Embedding Youtube task: {task.id}')
flash(f'Processing on Youtube document {new_doc.name}, version {new_doc_vers.id} started. Task ID: {task.id}.',
'success')
return redirect(prefixed_url_for('document_bp.documents'))
else:
form_validation_failed(request, form)
except EveAIYoutubeError as e:
flash(str(e), 'error')
except (EveAIInvalidLanguageException, EveAIUnsupportedFileType) as e:
flash(str(e), 'error')
except Exception as e:
current_app.logger.error(f'Error adding YouTube document: {str(e)}')
flash('An error occurred while adding the YouTube document.', 'error')
return render_template('document/add_youtube.html', form=form)
@@ -487,7 +416,7 @@ def refresh_document(doc_id):
current_app.logger.info(f'Document added successfully for tenant {session["tenant"]["id"]}, '
f'Document Version {new_doc_vers.id}')
upload_file_for_version(new_doc_vers, file, extension)
upload_file_for_version(new_doc_vers, file, extension, session["tenant"]["id"])
task = current_celery.send_task('create_embeddings', queue='embeddings', args=[
session['tenant']['id'],
@@ -535,116 +464,11 @@ def update_logging_information(obj, timestamp):
obj.updated_by = current_user.id
def create_document_stack(form, file, filename, extension):
# Create the Document
new_doc = create_document(form, filename)
# Create the DocumentVersion
new_doc_vers = create_version_for_document(new_doc,
form.get('url', ''),
form.get('language', 'en'),
form.get('user_context', '')
)
try:
db.session.add(new_doc)
db.session.add(new_doc_vers)
db.session.commit()
except SQLAlchemyError as e:
current_app.logger.error(f'Error adding document for tenant {session["tenant"]["id"]}: {e}')
flash('Error adding document.', 'alert')
db.session.rollback()
error = e.args
raise
except Exception as e:
current_app.logger.error('Unknown error')
raise
current_app.logger.info(f'Document added successfully for tenant {session["tenant"]["id"]}, '
f'Document Version {new_doc.id}')
upload_file_for_version(new_doc_vers, file, extension)
return new_doc, new_doc_vers
def log_session_state(session, msg=""):
current_app.logger.debug(f"{msg} - Session dirty: {session.dirty}")
current_app.logger.debug(f"{msg} - Session new: {session.new}")
def create_document(form, filename):
new_doc = Document()
if form['name'] == '':
new_doc.name = filename.rsplit('.', 1)[0]
else:
new_doc.name = form['name']
if form['valid_from'] and form['valid_from'] != '':
new_doc.valid_from = form['valid_from']
else:
new_doc.valid_from = dt.now(tz.utc)
new_doc.tenant_id = session['tenant']['id']
set_logging_information(new_doc, dt.now(tz.utc))
return new_doc
def create_version_for_document(document, url, language, user_context):
new_doc_vers = DocumentVersion()
if url != '':
new_doc_vers.url = url
if language == '':
new_doc_vers.language = session['default_language']
else:
new_doc_vers.language = language
if user_context != '':
new_doc_vers.user_context = user_context
new_doc_vers.document = document
set_logging_information(new_doc_vers, dt.now(tz.utc))
return new_doc_vers
def upload_file_for_version(doc_vers, file, extension):
doc_vers.file_type = extension
doc_vers.file_name = doc_vers.calc_file_name()
doc_vers.file_location = doc_vers.calc_file_location()
# Normally, the tenant bucket should exist. But let's be on the safe side if a migration took place.
tenant_id = session['tenant']['id']
minio_client.create_tenant_bucket(tenant_id)
try:
minio_client.upload_document_file(
tenant_id,
doc_vers.doc_id,
doc_vers.language,
doc_vers.id,
doc_vers.file_name,
file
)
db.session.commit()
current_app.logger.info(f'Successfully saved document to MinIO for tenant {tenant_id} for '
f'document version {doc_vers.id} while uploading file.')
except S3Error as e:
db.session.rollback()
flash('Error saving document to MinIO.', 'error')
current_app.logger.error(
f'Error saving document to MinIO for tenant {tenant_id}: {e}')
raise
except SQLAlchemyError as e:
db.session.rollback()
flash('Error saving document metadata.', 'error')
current_app.logger.error(
f'Error saving document metadata for tenant {tenant_id}: {e}')
raise
def fetch_html(url):
# Fetches HTML content from a URL
try: