From 9c23c25ebb4ed9edee4fbf7fc76608ac799c0453 Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Thu, 5 Feb 2026 13:40:11 -0500 Subject: [PATCH 01/46] Add macOS Desktop Services Store file to gitignore --- .gitignore | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/.gitignore b/.gitignore index d2cdbd62..984178dd 100644 --- a/.gitignore +++ b/.gitignore @@ -1,3 +1,4 @@ config/env/* !config/env/*.example -.idea/ \ No newline at end of file +.idea/ +.DS_Store \ No newline at end of file From 59b40f0eeff3b87b207a8598dbb9a79023afd01b Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Fri, 13 Feb 2026 15:51:21 -0500 Subject: [PATCH 02/46] REFACTOR Pull apart get_closest_embeddings to make testing easier --- server/api/services/embedding_services.py | 161 +++++++++++++++------- 1 file changed, 112 insertions(+), 49 deletions(-) diff --git a/server/api/services/embedding_services.py b/server/api/services/embedding_services.py index e35f7965..0720a9c8 100644 --- a/server/api/services/embedding_services.py +++ b/server/api/services/embedding_services.py @@ -11,18 +11,17 @@ logger = logging.getLogger(__name__) -def get_closest_embeddings( - user, message_data, document_name=None, guid=None, num_results=10 -): + +def build_query(user, embedding_vector, document_name=None, guid=None, num_results=10): """ - Find the closest embeddings to a given message for a specific user. + Build an unevaluated QuerySet for the closest embeddings. Parameters ---------- user : User The user whose uploaded documents will be searched - message_data : str - The input message to find similar embeddings for + embedding_vector : array-like + Pre-computed embedding vector to compare against document_name : str, optional Filter results to a specific document name guid : str, optional @@ -32,59 +31,52 @@ def get_closest_embeddings( Returns ------- - list[dict] - List of dictionaries containing embedding results with keys: - - name: document name - - text: embedded text content - - page_number: page number in source document - - chunk_number: chunk number within the document - - distance: L2 distance from query embedding - - file_id: GUID of the source file + QuerySet + Unevaluated Django QuerySet ordered by L2 distance, sliced to num_results """ - - encoding_start = time.time() - transformerModel = TransformerModel.get_instance().model - embedding_message = transformerModel.encode(message_data) - encoding_time = time.time() - encoding_start - - db_query_start = time.time() - # Django QuerySets are lazily evaluated if user.is_authenticated: # User sees their own files + files uploaded by superusers - closest_embeddings_query = ( - Embeddings.objects.filter( - Q(upload_file__uploaded_by=user) | Q(upload_file__uploaded_by__is_superuser=True) - ) - .annotate( - distance=L2Distance("embedding_sentence_transformers", embedding_message) - ) - .order_by("distance") + queryset = Embeddings.objects.filter( + Q(upload_file__uploaded_by=user) | Q(upload_file__uploaded_by__is_superuser=True) ) else: # Unauthenticated users only see superuser-uploaded files - closest_embeddings_query = ( - Embeddings.objects.filter(upload_file__uploaded_by__is_superuser=True) - .annotate( - distance=L2Distance("embedding_sentence_transformers", embedding_message) - ) - .order_by("distance") - ) + queryset = Embeddings.objects.filter(upload_file__uploaded_by__is_superuser=True) + + queryset = ( + queryset + .annotate(distance=L2Distance("embedding_sentence_transformers", embedding_vector)) + .order_by("distance") + ) # Filtering to a document GUID takes precedence over a document name if guid: - closest_embeddings_query = closest_embeddings_query.filter( - upload_file__guid=guid - ) + queryset = queryset.filter(upload_file__guid=guid) elif document_name: - closest_embeddings_query = closest_embeddings_query.filter(name=document_name) + queryset = queryset.filter(name=document_name) # Slicing is equivalent to SQL's LIMIT clause - closest_embeddings_query = closest_embeddings_query[:num_results] + return queryset[:num_results] + + +def format_results(queryset): + """ + Evaluate a QuerySet and return a list of result dicts. + + Parameters + ---------- + queryset : iterable + Iterable of Embeddings objects (or any objects with the expected attributes) + Returns + ------- + list[dict] + List of dicts with keys: name, text, page_number, chunk_number, distance, file_id + """ # Iterating evaluates the QuerySet and hits the database # TODO: Research improving the query evaluation performance - results = [ + return [ { "name": obj.name, "text": obj.text, @@ -93,13 +85,36 @@ def get_closest_embeddings( "distance": obj.distance, "file_id": obj.upload_file.guid if obj.upload_file else None, } - for obj in closest_embeddings_query + for obj in queryset ] - db_query_time = time.time() - db_query_start +def log_search_usage( + results, message_data, user, guid, document_name, num_results, encoding_time, db_query_time +): + """ + Create a SemanticSearchUsage record. Swallows exceptions so search isn't interrupted. + + Parameters + ---------- + results : list[dict] + The search results, each containing a "distance" key + message_data : str + The original search query text + user : User + The user who performed the search + guid : str or None + Document GUID filter used in the search + document_name : str or None + Document name filter used in the search + num_results : int + Number of results requested + encoding_time : float + Time in seconds to encode the query + db_query_time : float + Time in seconds for the database query + """ try: - # Handle user having no uploaded docs or doc filtering returning no matches if results: distances = [r["distance"] for r in results] SemanticSearchUsage.objects.create( @@ -113,11 +128,10 @@ def get_closest_embeddings( num_results_returned=len(results), max_distance=max(distances), median_distance=median(distances), - min_distance=min(distances) + min_distance=min(distances), ) else: logger.warning("Semantic search returned no results") - SemanticSearchUsage.objects.create( query_text=message_data, user=user if (user and user.is_authenticated) else None, @@ -129,9 +143,58 @@ def get_closest_embeddings( num_results_returned=0, max_distance=None, median_distance=None, - min_distance=None + min_distance=None, ) except Exception as e: logger.error(f"Failed to create semantic search usage database record: {e}") + +def get_closest_embeddings( + user, message_data, document_name=None, guid=None, num_results=10 +): + """ + Find the closest embeddings to a given message for a specific user. + + Parameters + ---------- + user : User + The user whose uploaded documents will be searched + message_data : str + The input message to find similar embeddings for + document_name : str, optional + Filter results to a specific document name + guid : str, optional + Filter results to a specific document GUID (takes precedence over document_name) + num_results : int, default 10 + Maximum number of results to return + + Returns + ------- + list[dict] + List of dictionaries containing embedding results with keys: + - name: document name + - text: embedded text content + - page_number: page number in source document + - chunk_number: chunk number within the document + - distance: L2 distance from query embedding + - file_id: GUID of the source file + + Notes + ----- + Creates a SemanticSearchUsage record. Swallows exceptions so search isn't interrupted. + """ + encoding_start = time.time() + model = TransformerModel.get_instance().model + embedding_vector = model.encode(message_data) + encoding_time = time.time() - encoding_start + + db_query_start = time.time() + queryset = build_query(user, embedding_vector, document_name, guid, num_results) + results = format_results(queryset) + db_query_time = time.time() - db_query_start + + log_search_usage( + results, message_data, user, guid, document_name, num_results, encoding_time, db_query_time + ) + return results From 3ffb74af318cc927d9c11ba37e63a3093d5ecfc6 Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Fri, 13 Feb 2026 16:12:49 -0500 Subject: [PATCH 03/46] ADD Add infra required to run pytest --- .github/workflows/python-app.yml | 5 +++++ server/pytest.ini | 3 +++ server/requirements.txt | 4 +++- 3 files changed, 11 insertions(+), 1 deletion(-) create mode 100644 server/pytest.ini diff --git a/.github/workflows/python-app.yml b/.github/workflows/python-app.yml index dcb7a4bb..41bc74d9 100644 --- a/.github/workflows/python-app.yml +++ b/.github/workflows/python-app.yml @@ -27,3 +27,8 @@ jobs: run: pipx install ruff - name: Lint code with Ruff run: ruff check --output-format=github --target-version=py39 + - name: Install test dependencies + run: pip install -r server/requirements.txt + # Discover and run all files matching test_*.py or *_test.py under server/ + - name: Run tests + run: pytest server/ -v diff --git a/server/pytest.ini b/server/pytest.ini new file mode 100644 index 00000000..235b9752 --- /dev/null +++ b/server/pytest.ini @@ -0,0 +1,3 @@ +[pytest] +DJANGO_SETTINGS_MODULE = balancer_backend.settings +pythonpath = . diff --git a/server/requirements.txt b/server/requirements.txt index bbaf7bc9..001708e9 100644 --- a/server/requirements.txt +++ b/server/requirements.txt @@ -18,4 +18,6 @@ sentence_transformers PyMuPDF==1.24.0 Pillow pytesseract -anthropic \ No newline at end of file +anthropic +pytest +pytest-django \ No newline at end of file From 12b09a733a1dc79b8fbec63d358d4354764a5116 Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Fri, 13 Feb 2026 16:33:41 -0500 Subject: [PATCH 04/46] ADD Start adding tests for embedding_services" --- server/api/services/embedding_services.py | 8 +- .../api/services/test_embedding_services.py | 85 +++++++++++++++++++ 2 files changed, 89 insertions(+), 4 deletions(-) create mode 100644 server/api/services/test_embedding_services.py diff --git a/server/api/services/embedding_services.py b/server/api/services/embedding_services.py index 0720a9c8..3fa9bb68 100644 --- a/server/api/services/embedding_services.py +++ b/server/api/services/embedding_services.py @@ -60,7 +60,7 @@ def build_query(user, embedding_vector, document_name=None, guid=None, num_resul return queryset[:num_results] -def format_results(queryset): +def evaluate_query(queryset): """ Evaluate a QuerySet and return a list of result dicts. @@ -89,7 +89,7 @@ def format_results(queryset): ] -def log_search_usage( +def log_usage( results, message_data, user, guid, document_name, num_results, encoding_time, db_query_time ): """ @@ -190,10 +190,10 @@ def get_closest_embeddings( db_query_start = time.time() queryset = build_query(user, embedding_vector, document_name, guid, num_results) - results = format_results(queryset) + results = evaluate_query(queryset) db_query_time = time.time() - db_query_start - log_search_usage( + log_usage( results, message_data, user, guid, document_name, num_results, encoding_time, db_query_time ) diff --git a/server/api/services/test_embedding_services.py b/server/api/services/test_embedding_services.py new file mode 100644 index 00000000..677c1e7b --- /dev/null +++ b/server/api/services/test_embedding_services.py @@ -0,0 +1,85 @@ +from unittest.mock import MagicMock, patch + +from api.services.embedding_services import evaluate_query, log_usage + + +def test_evaluate_query_maps_fields(): + obj = MagicMock() + obj.name = "doc.pdf" + obj.text = "some text" + obj.page_num = 3 + obj.chunk_number = 1 + obj.distance = 0.42 + obj.upload_file.guid = "abc-123" + + results = evaluate_query([obj]) + + assert results == [ + { + "name": "doc.pdf", + "text": "some text", + "page_number": 3, + "chunk_number": 1, + "distance": 0.42, + "file_id": "abc-123", + } + ] + + +def test_evaluate_query_none_upload_file(): + obj = MagicMock() + obj.name = "doc.pdf" + obj.text = "some text" + obj.page_num = 1 + obj.chunk_number = 0 + obj.distance = 1.0 + obj.upload_file = None + + results = evaluate_query([obj]) + + assert results[0]["file_id"] is None + + +@patch("api.services.embedding_services.SemanticSearchUsage.objects.create") +def test_log_usage_computes_distance_stats(mock_create): + results = [{"distance": 1.0}, {"distance": 3.0}, {"distance": 2.0}] + user = MagicMock(is_authenticated=True) + + log_usage( + results, + message_data="test query", + user=user, + guid=None, + document_name=None, + num_results=10, + encoding_time=0.1, + db_query_time=0.2, + ) + + mock_create.assert_called_once() + kwargs = mock_create.call_args.kwargs + assert kwargs["min_distance"] == 1.0 + assert kwargs["max_distance"] == 3.0 + assert kwargs["median_distance"] == 2.0 + assert kwargs["num_results_returned"] == 3 + + +@patch( + "api.services.embedding_services.SemanticSearchUsage.objects.create", + side_effect=Exception("DB error"), +) +def test_log_usage_swallows_exceptions(mock_create): + results = [{"distance": 1.0}] + user = MagicMock(is_authenticated=True) + + # pytest fails the test if it catches unhandled Exception + log_usage( + results, + message_data="test query", + user=user, + guid=None, + document_name=None, + num_results=10, + encoding_time=0.1, + db_query_time=0.2, + ) From 2d2ccd10911291061d23bcdf9b56855590477494 Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Mon, 16 Feb 2026 16:33:54 -0500 Subject: [PATCH 05/46] Fix duplicate healthcheck key for db service build --- docker-compose.yml | 5 ----- 1 file changed, 5 deletions(-) diff --git a/docker-compose.yml b/docker-compose.yml index 9182cdb6..7a6e7fe9 100644 --- a/docker-compose.yml +++ b/docker-compose.yml @@ -18,11 +18,6 @@ services: networks: app_net: ipv4_address: 192.168.0.2 - healthcheck: - test: ["CMD-SHELL", "pg_isready -U balancer -d balancer_dev"] - interval: 5s - timeout: 5s - retries: 5 pgadmin: image: dpage/pgadmin4 From 01ccf9ab0995645e0ad0c0782eb9bc5efb6ed595 Mon Sep 17 00:00:00 2001 From: Akhil Bolla <129509874+AkhilRB0204@users.noreply.github.com> Date: Mon, 16 Feb 2026 22:01:40 -0500 Subject: [PATCH 06/46] Enhance input sanitization and normalize pronouns Updated the sanitizer function to improve input sanitization by removing style tags, normalizing pronouns, and increasing the maximum length limit. --- server/api/views/assistant/sanitizer.py | 62 ++++++++++++++++++++++--- 1 file changed, 56 insertions(+), 6 deletions(-) diff --git a/server/api/views/assistant/sanitizer.py b/server/api/views/assistant/sanitizer.py index bdbbc77f..fd851df6 100644 --- a/server/api/views/assistant/sanitizer.py +++ b/server/api/views/assistant/sanitizer.py @@ -1,26 +1,76 @@ import re import logging + logger = logging.getLogger(__name__) def sanitize_input(user_input:str) -> str: """ Sanitize user input to prevent injection attacks and remove unwanted characters. + Args: user_input (str): The raw input string from the user. + Returns: str: The sanitized input string. """ try: - # Remove any script tags - sanitized = re.sub(r'.*?', '', user_input, flags=re.IGNORECASE) - # Remove any HTML tags + sanitized = user_input + + # Remove any style tags + sanitized = re.sub(r'.*?', '', sanitized, flags=re.IGNORECASE) + + # Remove any HTML/script tags sanitized = re.sub(r'<.*?>', '', sanitized) + + # Remove Phone Numbers + sanitized = re.sub(r'\+?\d[\d -]{8,}\d', '[Phone Number]', sanitized) + + # Remove Email Addresses + sanitized = re.sub(r'[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}', '[Email Address]', sanitized) + + # Remove Medical Record Numbers (simple pattern) + sanitized = re.sub(r'\bMRN[:\s]*\d+\b', '[Medical Record Number]', sanitized, flags=re.IGNORECASE) + + # Normalize pronouns + sanitized = normalize_pronouns(sanitized) + # Escape special characters - sanitized = re.sub(r'["\'\\]', '', sanitized) + sanitized = re.sub(r'\s+', '', sanitized) + # Limit length to prevent buffer overflow attacks - max_length = 1000 + max_length = 5000 if len(sanitized) > max_length: sanitized = sanitized[:max_length] + return sanitized.strip() except Exception as e: logger.error(f"Error sanitizing input: {e}") - return "" \ No newline at end of file + return "" + +def normalize_pronouns(text:str) -> str: + """ + Normalize first and second person pronouns to third person clinical language. + + Converts patient centric pronouns to a more neutral form. + Args: + text (str): The input text containing pronouns. + Returns: + str: The text with normalized pronouns. + """ + # Normalize first person possessives: I, me, my, mine -> the patient + text = re.sub(r'\bMy\b', 'The patient\'s', text) + text = re.sub(r'\bmy\b', 'the patient\'s', text) + + # First person subject: I -> the patient + text = re.sub(r'\bI\b', 'the patient', text) + + # First person object: me -> the patient + text = re.sub(r'\bme\b', 'the patient', text) + + # First person reflexive: myself -> the patient + text = re.sub(r'\bmyself\b', 'the patient', text) + + # Second person: you, your -> the clinician + text = re.sub(r'\bYour\b', 'the clinician', text) + return text + + From da9afaa905a43218e1f947360aa141662359e1aa Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Tue, 17 Feb 2026 14:40:13 -0500 Subject: [PATCH 07/46] DOC Add a note about running pytest in the README --- README.md | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/README.md b/README.md index f1cea06b..15018d37 100644 --- a/README.md +++ b/README.md @@ -69,6 +69,11 @@ df = pd.read_sql(query, engine) #### Django REST - The email and password are set in `server/api/management/commands/createsu.py` +- Backend tests can be run using `pytest` by running the below command inside the running backend container: + +``` +docker compose exec backend pytest api/ -v +``` ## Local Kubernetes Deployment From b08152f94a079b24681e79203476dcbbbdcb0d83 Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Wed, 18 Feb 2026 14:18:12 -0500 Subject: [PATCH 08/46] fix: changed link to direct to balancer github page --- frontend/src/components/Footer/Footer.tsx | 2 +- frontend/src/components/Header/Header.tsx | 2 +- frontend/src/components/Header/MdNavBar.tsx | 2 +- frontend/src/pages/About/About.tsx | 2 +- 4 files changed, 4 insertions(+), 4 deletions(-) diff --git a/frontend/src/components/Footer/Footer.tsx b/frontend/src/components/Footer/Footer.tsx index d656f5ad..2f68ec49 100644 --- a/frontend/src/components/Footer/Footer.tsx +++ b/frontend/src/components/Footer/Footer.tsx @@ -62,7 +62,7 @@ function Footer() { > Leave feedback - diff --git a/frontend/src/components/Header/Header.tsx b/frontend/src/components/Header/Header.tsx index cbbd2c93..a0edc859 100644 --- a/frontend/src/components/Header/Header.tsx +++ b/frontend/src/components/Header/Header.tsx @@ -165,7 +165,7 @@ const Header: React.FC = ({ isAuthenticated, isSuperuser }) => { Leave Feedback diff --git a/frontend/src/components/Header/MdNavBar.tsx b/frontend/src/components/Header/MdNavBar.tsx index 5a8d5bce..00d45f55 100644 --- a/frontend/src/components/Header/MdNavBar.tsx +++ b/frontend/src/components/Header/MdNavBar.tsx @@ -120,7 +120,7 @@ const MdNavBar = (props: LoginFormProps) => {
  • - diff --git a/frontend/src/pages/About/About.tsx b/frontend/src/pages/About/About.tsx index c50f6705..e1c7242e 100644 --- a/frontend/src/pages/About/About.tsx +++ b/frontend/src/pages/About/About.tsx @@ -77,7 +77,7 @@ function About() {
    - + From b94e998d8c355eb19153b2d8af5527c3fa17ff14 Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Fri, 20 Feb 2026 15:33:35 -0500 Subject: [PATCH 09/46] Fix error 1, added unit tests and more logging --- server/api/views/uploadFile/test_title.py | 30 +++++++++++++++++++++++ server/api/views/uploadFile/title.py | 3 ++- server/api/views/uploadFile/views.py | 4 +++ 3 files changed, 36 insertions(+), 1 deletion(-) diff --git a/server/api/views/uploadFile/test_title.py b/server/api/views/uploadFile/test_title.py index 69979620..5391ed05 100644 --- a/server/api/views/uploadFile/test_title.py +++ b/server/api/views/uploadFile/test_title.py @@ -67,3 +67,33 @@ def test_falls_back_to_chatgpt_if_no_title_found(self, mock_openAI): title.generate_title(doc) self.assertTrue(mock_openAI.called) + + @patch("api.services.openai_services.openAIServices.openAI") + def test_strips_quotes_from_openai_title(self, mock_openAI): + doc = MagicMock() + doc.metadata = {"title": None} + doc.get_text.return_value = [] + + mock_response = MagicMock() + mock_response.choices = [MagicMock()] + mock_response.choices[0].message.content = '"Updated CANMAT/ISBD Guidelines for Treating Mixed Features in Bipolar Disorder"' + mock_openAI.return_value = mock_response + + result = title.generate_title(doc) + + self.assertEqual(result, "Updated CANMAT/ISBD Guidelines for Treating Mixed Features in Bipolar Disorder") + + @patch("api.services.openai_services.openAIServices.openAI") + def test_truncates_long_openai_title(self, mock_openAI): + doc = MagicMock() + doc.metadata = {"title": None} + doc.get_text.return_value = [] + + mock_response = MagicMock() + mock_response.choices = [MagicMock()] + mock_response.choices[0].message.content = "A" * 300 + mock_openAI.return_value = mock_response + + result = title.generate_title(doc) + + self.assertLessEqual(len(result), 255) diff --git a/server/api/views/uploadFile/title.py b/server/api/views/uploadFile/title.py index 06e0ce0c..b3f8aded 100644 --- a/server/api/views/uploadFile/title.py +++ b/server/api/views/uploadFile/title.py @@ -58,4 +58,5 @@ def summarize_pdf(pdf: fitz.Document) -> str: prompt = "Please provide a title for this document. The title should be less than 256 characters and will be displayed on a webpage." response = openAIServices.openAI( first_page_content, prompt, model='gpt-4o', temp=0.0) - return response.choices[0].message.content + title = response.choices[0].message.content.strip().strip('"').strip("'") + return title[:255] diff --git a/server/api/views/uploadFile/views.py b/server/api/views/uploadFile/views.py index 69dfb996..58bd8752 100644 --- a/server/api/views/uploadFile/views.py +++ b/server/api/views/uploadFile/views.py @@ -12,6 +12,9 @@ import fitz from django.db import transaction from .title import generate_title +import logging + +logger = logging.getLogger(__name__) class UploadFileView(APIView): @@ -124,6 +127,7 @@ def post(self, request, format=None): ) except Exception as e: # Handle potential errors + logger.exception("File upload failed for '%s': %s", uploaded_file.name, e) return Response({"message": f"Error processing file and embeddings: {str(e)}"}, status=status.HTTP_400_BAD_REQUEST) From 530b90a17afa6136e6c9de9abd472794c4058193 Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Mon, 23 Feb 2026 12:19:50 -0500 Subject: [PATCH 10/46] Changed button text from "donate" to "Support Developoment" --- frontend/src/components/Footer/Footer.tsx | 4 ++-- frontend/src/components/Header/Header.tsx | 2 +- frontend/src/components/Header/MdNavBar.tsx | 2 +- frontend/src/pages/About/About.tsx | 2 +- 4 files changed, 5 insertions(+), 5 deletions(-) diff --git a/frontend/src/components/Footer/Footer.tsx b/frontend/src/components/Footer/Footer.tsx index 2f68ec49..977c59d4 100644 --- a/frontend/src/components/Footer/Footer.tsx +++ b/frontend/src/components/Footer/Footer.tsx @@ -64,9 +64,9 @@ function Footer() { - Donate + Support Development = ({ isAuthenticated, isSuperuser }) => { target="_blank" className="header-nav-item" > - Donate + Support Development {isAuthenticated && isSuperuser && (
    { target="_blank" className="mr-9 text-black hover:border-b-2 hover:border-blue-600 hover:text-black hover:no-underline" > - Donate + Support Development
    • {isAuthenticated && diff --git a/frontend/src/pages/About/About.tsx b/frontend/src/pages/About/About.tsx index e1c7242e..9481c74d 100644 --- a/frontend/src/pages/About/About.tsx +++ b/frontend/src/pages/About/About.tsx @@ -79,7 +79,7 @@ function About() {
    From f96606d18b0f9fa2aca25033ba26079c290b54ac Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Mon, 23 Feb 2026 16:46:38 -0500 Subject: [PATCH 11/46] Fix 401 by using adminApi instead of raw axios --- frontend/src/pages/DocumentManager/UploadFile.tsx | 9 ++------- 1 file changed, 2 insertions(+), 7 deletions(-) diff --git a/frontend/src/pages/DocumentManager/UploadFile.tsx b/frontend/src/pages/DocumentManager/UploadFile.tsx index 2ee7b5db..32b727e8 100644 --- a/frontend/src/pages/DocumentManager/UploadFile.tsx +++ b/frontend/src/pages/DocumentManager/UploadFile.tsx @@ -1,5 +1,5 @@ import React, { useState, useRef } from "react"; -import axios from "axios"; +import { adminApi } from "../../api/apiClient"; import TypingAnimation from "../../components/Header/components/TypingAnimation.tsx"; import Layout from "../Layout/Layout.tsx"; @@ -22,14 +22,9 @@ const UploadFile: React.FC = () => { formData.append("file", file); try { - const response = await axios.post( + const response = await adminApi.post( `/api/v1/api/uploadFile`, formData, - { - headers: { - "Content-Type": "multipart/form-data" - }, - } ); console.log("File uploaded successfully", response.data); } catch (error) { From bbf1034d065774a252bee3cae07300584e84fbe6 Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Thu, 26 Feb 2026 15:48:51 -0500 Subject: [PATCH 12/46] Fixed wrong API url path in handleDownload --- frontend/src/pages/Files/ListOfFiles.tsx | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/frontend/src/pages/Files/ListOfFiles.tsx b/frontend/src/pages/Files/ListOfFiles.tsx index b6fff4ee..cac0836c 100644 --- a/frontend/src/pages/Files/ListOfFiles.tsx +++ b/frontend/src/pages/Files/ListOfFiles.tsx @@ -61,7 +61,7 @@ const ListOfFiles: React.FC<{ showTable?: boolean }> = ({ const handleDownload = async (guid: string, fileName: string) => { try { setDownloading(guid); - const { data } = await publicApi.get(`/v1/api/uploadFile/${guid}`, { responseType: 'blob' }); + const { data } = await publicApi.get(`/api/v1/api/uploadFile/${guid}`, { responseType: 'blob' }); const url = window.URL.createObjectURL(new Blob([data])); const link = document.createElement("a"); From 128418b5cc05a867c0419d504be42cdbd6b7cf51 Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Thu, 26 Feb 2026 16:22:45 -0500 Subject: [PATCH 13/46] Fixed API URL in handleOpen as well --- frontend/src/pages/Files/ListOfFiles.tsx | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/frontend/src/pages/Files/ListOfFiles.tsx b/frontend/src/pages/Files/ListOfFiles.tsx index cac0836c..37bd459a 100644 --- a/frontend/src/pages/Files/ListOfFiles.tsx +++ b/frontend/src/pages/Files/ListOfFiles.tsx @@ -82,7 +82,7 @@ const ListOfFiles: React.FC<{ showTable?: boolean }> = ({ const handleOpen = async (guid: string) => { try { setOpening(guid); - const { data } = await publicApi.get(`/v1/api/uploadFile/${guid}`, { responseType: 'arraybuffer' }); + const { data } = await publicApi.get(`/api/v1/api/uploadFile/${guid}`, { responseType: 'arraybuffer' }); const file = new Blob([data], { type: 'application/pdf' }); const fileURL = window.URL.createObjectURL(file); From 332af9fca17085196bdad7db35d7fbe46c981274 Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Thu, 26 Feb 2026 16:57:36 -0500 Subject: [PATCH 14/46] drf-spectacular configuration --- server/balancer_backend/settings.py | 8 ++++++++ server/requirements.txt | 3 ++- 2 files changed, 10 insertions(+), 1 deletion(-) diff --git a/server/balancer_backend/settings.py b/server/balancer_backend/settings.py index 9f917a94..cfa90dce 100644 --- a/server/balancer_backend/settings.py +++ b/server/balancer_backend/settings.py @@ -51,6 +51,7 @@ "corsheaders", "rest_framework", "djoser", + 'drf_spectacular', ] MIDDLEWARE = [ @@ -195,8 +196,15 @@ "DEFAULT_AUTHENTICATION_CLASSES": ( "rest_framework_simplejwt.authentication.JWTAuthentication", ), + 'DEFAULT_SCHEMA_CLASS': 'drf_spectacular.openapi.AutoSchema', } +SPECTACULAR_SETTINGS = { + 'TITLE': 'Balancer API', + 'DESCRIPTION': 'API for the Balancer medication decision support tool', + 'VERSION': '1.0.0', + 'SERVE_INCLUDE_SCHEMA': False, +} SIMPLE_JWT = { "AUTH_HEADER_TYPES": ("JWT",), diff --git a/server/requirements.txt b/server/requirements.txt index bbaf7bc9..880500c6 100644 --- a/server/requirements.txt +++ b/server/requirements.txt @@ -18,4 +18,5 @@ sentence_transformers PyMuPDF==1.24.0 Pillow pytesseract -anthropic \ No newline at end of file +anthropic +drf-spectacular \ No newline at end of file From a34a9f8ceda2358cebd8607c9dcb04136e590cb3 Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Thu, 26 Feb 2026 17:10:03 -0500 Subject: [PATCH 15/46] Added URL routes for API docs generation --- server/balancer_backend/urls.py | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/server/balancer_backend/urls.py b/server/balancer_backend/urls.py index c8bd290d..55bd2032 100644 --- a/server/balancer_backend/urls.py +++ b/server/balancer_backend/urls.py @@ -6,6 +6,9 @@ # Import TemplateView for rendering templates from django.views.generic import TemplateView import importlib # Import the importlib module for dynamic module importing +from drf_spectacular.views import SpectacularAPIView, SpectacularSwaggerView, SpectacularRedocView + + # Define a list of URL patterns for the application # Keep admin outside /api/ prefix @@ -50,6 +53,9 @@ # Wrap all API routes under /api/ prefix urlpatterns += [ path("api/", include(api_urlpatterns)), + path("api/schema/", SpectacularAPIView.as_view(), name="schema"), + path("api/docs/", SpectacularSwaggerView.as_view(url_name="schema"), name="swagger-ui"), + path("api/redoc/", SpectacularRedocView.as_view(url_name="schema"), name="redoc"), ] import os From fe660d2eb03db708b30b782f4fab787610640a90 Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Thu, 26 Feb 2026 17:26:44 -0500 Subject: [PATCH 16/46] Added OpenAPI security scheme --- server/balancer_backend/settings.py | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/server/balancer_backend/settings.py b/server/balancer_backend/settings.py index cfa90dce..a4ccaaae 100644 --- a/server/balancer_backend/settings.py +++ b/server/balancer_backend/settings.py @@ -204,6 +204,10 @@ 'DESCRIPTION': 'API for the Balancer medication decision support tool', 'VERSION': '1.0.0', 'SERVE_INCLUDE_SCHEMA': False, + 'SECURITY': [{'jwtAuth': []}], + 'SWAGGER_UI_SETTINGS': { + 'persistAuthorization': True, + }, } SIMPLE_JWT = { From 3c83abdc2a32cc6c300fc833671e8f7a937ba88e Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Thu, 26 Feb 2026 18:12:52 -0500 Subject: [PATCH 17/46] Added extend_schema and serializer_class to endpoints that drf-spectacular missed with default settings --- server/api/views/ai_promptStorage/views.py | 3 ++ server/api/views/ai_settings/views.py | 2 + server/api/views/assistant/views.py | 17 +++++++ server/api/views/conversations/views.py | 31 ++++++++++++ server/api/views/embeddings/embeddingsView.py | 23 ++++++++- server/api/views/feedback/views.py | 1 + server/api/views/listMeds/views.py | 48 ++++++++++++++++++- server/api/views/medRules/serializers.py | 2 + server/api/views/medRules/views.py | 25 +++++++++- .../api/views/risk/views_riskWithSources.py | 25 +++++++++- server/api/views/text_extraction/views.py | 29 +++++++++++ server/api/views/uploadFile/views.py | 46 +++++++++++++++++- server/api/views/version/views.py | 7 +++ 13 files changed, 254 insertions(+), 5 deletions(-) diff --git a/server/api/views/ai_promptStorage/views.py b/server/api/views/ai_promptStorage/views.py index 7354feb3..cc50f22e 100644 --- a/server/api/views/ai_promptStorage/views.py +++ b/server/api/views/ai_promptStorage/views.py @@ -1,10 +1,12 @@ from rest_framework import status from rest_framework.decorators import api_view from rest_framework.response import Response +from drf_spectacular.utils import extend_schema from .models import AI_PromptStorage from .serializers import AI_PromptStorageSerializer +@extend_schema(request=AI_PromptStorageSerializer, responses={201: AI_PromptStorageSerializer}) @api_view(['POST']) # @permission_classes([IsAuthenticated]) def store_prompt(request): @@ -21,6 +23,7 @@ def store_prompt(request): return Response(serializer.errors, status=status.HTTP_400_BAD_REQUEST) +@extend_schema(responses={200: AI_PromptStorageSerializer(many=True)}) @api_view(['GET']) def get_all_prompts(request): """ diff --git a/server/api/views/ai_settings/views.py b/server/api/views/ai_settings/views.py index 349b9fd9..9ee6aad7 100644 --- a/server/api/views/ai_settings/views.py +++ b/server/api/views/ai_settings/views.py @@ -2,10 +2,12 @@ from rest_framework.decorators import api_view, permission_classes from rest_framework.permissions import IsAuthenticated from rest_framework.response import Response +from drf_spectacular.utils import extend_schema from .models import AI_Settings from .serializers import AISettingsSerializer +@extend_schema(request=AISettingsSerializer, responses={200: AISettingsSerializer(many=True), 201: AISettingsSerializer}) @api_view(['GET', 'POST']) @permission_classes([IsAuthenticated]) def settings_view(request): diff --git a/server/api/views/assistant/views.py b/server/api/views/assistant/views.py index f31ab475..e3e8d6f7 100644 --- a/server/api/views/assistant/views.py +++ b/server/api/views/assistant/views.py @@ -10,6 +10,8 @@ from rest_framework.permissions import AllowAny from django.utils.decorators import method_decorator from django.views.decorators.csrf import csrf_exempt +from drf_spectacular.utils import extend_schema, inline_serializer +from rest_framework import serializers as drf_serializers from openai import OpenAI @@ -113,6 +115,21 @@ def invoke_functions_from_response( class Assistant(APIView): permission_classes = [AllowAny] + @extend_schema( + request=inline_serializer(name='AssistantRequest', fields={ + 'message': drf_serializers.CharField(help_text='User message to send to the assistant'), + 'previous_response_id': drf_serializers.CharField(required=False, allow_null=True, help_text='ID of previous response for conversation continuity'), + }), + responses={ + 200: inline_serializer(name='AssistantResponse', fields={ + 'response_output_text': drf_serializers.CharField(), + 'final_response_id': drf_serializers.CharField(), + }), + 500: inline_serializer(name='AssistantError', fields={ + 'error': drf_serializers.CharField(), + }), + } + ) def post(self, request): try: user = request.user diff --git a/server/api/views/conversations/views.py b/server/api/views/conversations/views.py index eeb68809..de927cf1 100644 --- a/server/api/views/conversations/views.py +++ b/server/api/views/conversations/views.py @@ -16,6 +16,8 @@ from .models import Conversation, Message from .serializers import ConversationSerializer from ...services.tools.tools import tools, execute_tool +from drf_spectacular.utils import extend_schema, inline_serializer +from rest_framework import serializers as drf_serializers @csrf_exempt @@ -95,6 +97,21 @@ def destroy(self, request, *args, **kwargs): self.perform_destroy(instance) return Response(status=status.HTTP_204_NO_CONTENT) + @extend_schema( + request=inline_serializer(name='ContinueConversationRequest', fields={ + 'message': drf_serializers.CharField(help_text='User message to continue the conversation'), + 'page_context': drf_serializers.CharField(required=False, help_text='Optional page context'), + }), + responses={ + 200: inline_serializer(name='ContinueConversationResponse', fields={ + 'response': drf_serializers.CharField(), + 'title': drf_serializers.CharField(), + }), + 400: inline_serializer(name='ContinueConversationBadRequest', fields={ + 'error': drf_serializers.CharField(), + }), + } + ) @action(detail=True, methods=['post']) def continue_conversation(self, request, pk=None): conversation = self.get_object() @@ -123,6 +140,20 @@ def continue_conversation(self, request, pk=None): return Response({"response": chatgpt_response, "title": conversation.title}) + @extend_schema( + request=inline_serializer(name='UpdateTitleRequest', fields={ + 'title': drf_serializers.CharField(help_text='New conversation title'), + }), + responses={ + 200: inline_serializer(name='UpdateTitleResponse', fields={ + 'status': drf_serializers.CharField(), + 'title': drf_serializers.CharField(), + }), + 400: inline_serializer(name='UpdateTitleBadRequest', fields={ + 'error': drf_serializers.CharField(), + }), + } + ) @action(detail=True, methods=['patch']) def update_title(self, request, pk=None): conversation = self.get_object() diff --git a/server/api/views/embeddings/embeddingsView.py b/server/api/views/embeddings/embeddingsView.py index d0bdd8ca..ebcf0774 100644 --- a/server/api/views/embeddings/embeddingsView.py +++ b/server/api/views/embeddings/embeddingsView.py @@ -1,8 +1,9 @@ from rest_framework.views import APIView from rest_framework.permissions import IsAuthenticated from rest_framework.response import Response -from rest_framework import status +from rest_framework import status, serializers as drf_serializers from django.http import StreamingHttpResponse +from drf_spectacular.utils import extend_schema, inline_serializer, OpenApiParameter from ...services.embedding_services import get_closest_embeddings from ...services.conversions_services import convert_uuids from ...services.openai_services import openAIServices @@ -15,6 +16,26 @@ class AskEmbeddingsAPIView(APIView): permission_classes = [IsAuthenticated] + @extend_schema( + parameters=[ + OpenApiParameter(name='guid', type=str, location=OpenApiParameter.QUERY, required=False, description='Optional file GUID to filter embeddings'), + OpenApiParameter(name='stream', type=bool, location=OpenApiParameter.QUERY, required=False, description='Enable streaming response'), + ], + request=inline_serializer(name='AskEmbeddingsRequest', fields={ + 'message': drf_serializers.CharField(help_text='Question to ask against embedded documents'), + }), + responses={ + 200: inline_serializer(name='AskEmbeddingsResponse', fields={ + 'question': drf_serializers.CharField(), + 'llm_response': drf_serializers.CharField(), + 'embeddings_info': drf_serializers.CharField(), + 'sent_to_llm': drf_serializers.CharField(), + }), + 400: inline_serializer(name='AskEmbeddingsBadRequest', fields={ + 'error': drf_serializers.CharField(), + }), + } + ) def post(self, request, *args, **kwargs): try: user = request.user diff --git a/server/api/views/feedback/views.py b/server/api/views/feedback/views.py index d0f0e1da..424e0758 100644 --- a/server/api/views/feedback/views.py +++ b/server/api/views/feedback/views.py @@ -9,6 +9,7 @@ class FeedbackView(APIView): permission_classes = [AllowAny] + serializer_class = FeedbackSerializer def post(self, request, *args, **kwargs): serializer = FeedbackSerializer(data=request.data) diff --git a/server/api/views/listMeds/views.py b/server/api/views/listMeds/views.py index fcd0edf2..1b199a7e 100644 --- a/server/api/views/listMeds/views.py +++ b/server/api/views/listMeds/views.py @@ -1,7 +1,8 @@ -from rest_framework import status +from rest_framework import status, serializers as drf_serializers from rest_framework.permissions import AllowAny from rest_framework.response import Response from rest_framework.views import APIView +from drf_spectacular.utils import extend_schema, inline_serializer from .models import Diagnosis, Medication, Suggestion from .serializers import MedicationSerializer @@ -24,6 +25,33 @@ class GetMedication(APIView): permission_classes = [AllowAny] + @extend_schema( + request=inline_serializer( + name='GetMedicationRequest', + fields={ + 'state': drf_serializers.CharField(help_text='Diagnosis state, e.g. "depressed", "manic"'), + 'suicideHistory': drf_serializers.BooleanField(default=False), + 'kidneyHistory': drf_serializers.BooleanField(default=False), + 'liverHistory': drf_serializers.BooleanField(default=False), + 'bloodPressureHistory': drf_serializers.BooleanField(default=False), + 'weightGainConcern': drf_serializers.BooleanField(default=False), + 'priorMedications': drf_serializers.CharField(required=False, default='', help_text='Comma-separated medication names'), + } + ), + responses={ + 200: inline_serializer( + name='GetMedicationResponse', + fields={ + 'first': drf_serializers.ListField(child=drf_serializers.DictField()), + 'second': drf_serializers.ListField(child=drf_serializers.DictField()), + 'third': drf_serializers.ListField(child=drf_serializers.DictField()), + } + ), + 404: inline_serializer(name='GetMedicationNotFound', fields={ + 'error': drf_serializers.CharField(), + }), + } + ) def post(self, request): data = request.data state_query = data.get('state', '') @@ -75,6 +103,7 @@ def post(self, request): class ListOrDetailMedication(APIView): permission_classes = [AllowAny] + serializer_class = MedicationSerializer def get(self, request): name_query = request.query_params.get('name', None) @@ -98,6 +127,7 @@ class AddMedication(APIView): """ API endpoint to add a medication to the database with its risks and benefits. """ + serializer_class = MedicationSerializer def post(self, request): data = request.data @@ -129,6 +159,22 @@ class DeleteMedication(APIView): API endpoint to delete medication if medication in database. """ + @extend_schema( + request=inline_serializer(name='DeleteMedicationRequest', fields={ + 'name': drf_serializers.CharField(), + }), + responses={ + 200: inline_serializer(name='DeleteMedicationSuccess', fields={ + 'success': drf_serializers.CharField(), + }), + 400: inline_serializer(name='DeleteMedicationBadRequest', fields={ + 'error': drf_serializers.CharField(), + }), + 404: inline_serializer(name='DeleteMedicationNotFound', fields={ + 'error': drf_serializers.CharField(), + }), + } + ) def delete(self, request): data = request.data name = data.get('name', '').strip() diff --git a/server/api/views/medRules/serializers.py b/server/api/views/medRules/serializers.py index df5e3663..e0d7d3f3 100644 --- a/server/api/views/medRules/serializers.py +++ b/server/api/views/medRules/serializers.py @@ -1,4 +1,5 @@ from rest_framework import serializers +from drf_spectacular.utils import extend_schema_field from ...models.model_medRule import MedRule, MedRuleSource from ..listMeds.serializers import MedicationSerializer from ...models.model_embeddings import Embeddings @@ -30,6 +31,7 @@ class Meta: "medication_sources", ] + @extend_schema_field(MedicationWithSourcesSerializer(many=True)) def get_medication_sources(self, obj): medrule_sources = MedRuleSource.objects.filter(medrule=obj).select_related( "medication", "embedding" diff --git a/server/api/views/medRules/views.py b/server/api/views/medRules/views.py index 2fae140b..2f80f8f3 100644 --- a/server/api/views/medRules/views.py +++ b/server/api/views/medRules/views.py @@ -1,9 +1,10 @@ from rest_framework.views import APIView from rest_framework.permissions import IsAuthenticated from rest_framework.response import Response -from rest_framework import status +from rest_framework import status, serializers as drf_serializers from django.utils.decorators import method_decorator from django.views.decorators.csrf import csrf_exempt +from drf_spectacular.utils import extend_schema, inline_serializer from ...models.model_medRule import MedRule from .serializers import MedRuleSerializer # You'll need to create this from ..listMeds.models import Medication @@ -13,6 +14,7 @@ @method_decorator(csrf_exempt, name='dispatch') class MedRules(APIView): permission_classes = [IsAuthenticated] + serializer_class = MedRuleSerializer def get(self, request, format=None): # Get all med rules @@ -29,6 +31,27 @@ def get(self, request, format=None): return Response(data, status=status.HTTP_200_OK) + @extend_schema( + request=inline_serializer(name='MedRuleCreateRequest', fields={ + 'rule_type': drf_serializers.CharField(help_text='INCLUDE or EXCLUDE'), + 'history_type': drf_serializers.CharField(help_text='e.g. DIAGNOSIS_DEPRESSED, DIAGNOSIS_MANIC'), + 'reason': drf_serializers.CharField(), + 'label': drf_serializers.CharField(), + 'explanation': drf_serializers.CharField(), + 'medication_names': drf_serializers.ListField(child=drf_serializers.CharField()), + 'chunk_ids': drf_serializers.ListField(child=drf_serializers.IntegerField()), + 'file_guid': drf_serializers.CharField(), + }), + responses={ + 201: MedRuleSerializer, + 400: inline_serializer(name='MedRuleCreateBadRequest', fields={ + 'error': drf_serializers.CharField(), + }), + 404: inline_serializer(name='MedRuleCreateNotFound', fields={ + 'error': drf_serializers.CharField(), + }), + } + ) def post(self, request): data = request.data diff --git a/server/api/views/risk/views_riskWithSources.py b/server/api/views/risk/views_riskWithSources.py index c02908fc..26cad9f8 100644 --- a/server/api/views/risk/views_riskWithSources.py +++ b/server/api/views/risk/views_riskWithSources.py @@ -1,7 +1,8 @@ from rest_framework.views import APIView from rest_framework.response import Response -from rest_framework import status +from rest_framework import status, serializers as drf_serializers from rest_framework.permissions import AllowAny +from drf_spectacular.utils import extend_schema, inline_serializer from api.views.listMeds.models import Medication from api.models.model_medRule import MedRule, MedRuleSource import openai @@ -11,6 +12,28 @@ class RiskWithSourcesView(APIView): permission_classes = [AllowAny] + @extend_schema( + request=inline_serializer(name='RiskWithSourcesRequest', fields={ + 'drug': drf_serializers.CharField(help_text='Medication name'), + 'source': drf_serializers.CharField(required=False, help_text='One of: include, diagnosis, diagnosis_depressed, diagnosis_manic, diagnosis_hypomanic, diagnosis_euthymic'), + }), + responses={ + 200: inline_serializer(name='RiskWithSourcesResponse', fields={ + 'benefits': drf_serializers.ListField(child=drf_serializers.CharField()), + 'risks': drf_serializers.ListField(child=drf_serializers.CharField()), + 'sources': drf_serializers.ListField(child=drf_serializers.DictField()), + 'medrules_found': drf_serializers.IntegerField(required=False), + 'source_type': drf_serializers.CharField(required=False), + 'note': drf_serializers.CharField(required=False), + }), + 400: inline_serializer(name='RiskWithSourcesBadRequest', fields={ + 'error': drf_serializers.CharField(), + }), + 404: inline_serializer(name='RiskWithSourcesNotFound', fields={ + 'error': drf_serializers.CharField(), + }), + } + ) def post(self, request): openai.api_key = os.environ.get("OPENAI_API_KEY") diff --git a/server/api/views/text_extraction/views.py b/server/api/views/text_extraction/views.py index e4122851..020740ad 100644 --- a/server/api/views/text_extraction/views.py +++ b/server/api/views/text_extraction/views.py @@ -9,6 +9,8 @@ from django.utils.decorators import method_decorator from django.views.decorators.csrf import csrf_exempt import anthropic +from drf_spectacular.utils import extend_schema, inline_serializer, OpenApiParameter +from rest_framework import serializers as drf_serializers from ...services.openai_services import openAIServices from api.models.model_embeddings import Embeddings @@ -97,6 +99,20 @@ class RuleExtractionAPIView(APIView): permission_classes = [IsAuthenticated] + @extend_schema( + parameters=[ + OpenApiParameter(name='guid', type=str, location=OpenApiParameter.QUERY, required=True, description='File GUID to extract rules from'), + ], + responses={ + 200: inline_serializer(name='RuleExtractionResponse', fields={ + 'texts': drf_serializers.CharField(), + 'cited_texts': drf_serializers.CharField(), + }), + 500: inline_serializer(name='RuleExtractionError', fields={ + 'error': drf_serializers.CharField(), + }), + } + ) def get(self, request): try: @@ -141,6 +157,19 @@ def openai_extraction(content_chunks, user_prompt): class RuleExtractionAPIOpenAIView(APIView): permission_classes = [IsAuthenticated] + @extend_schema( + parameters=[ + OpenApiParameter(name='guid', type=str, location=OpenApiParameter.QUERY, required=True, description='File GUID to extract rules from'), + ], + responses={ + 200: inline_serializer(name='RuleExtractionOpenAIResponse', fields={ + 'rules': drf_serializers.ListField(child=drf_serializers.DictField()), + }), + 500: inline_serializer(name='RuleExtractionOpenAIError', fields={ + 'error': drf_serializers.CharField(), + }), + } + ) def get(self, request): try: user_prompt = """ diff --git a/server/api/views/uploadFile/views.py b/server/api/views/uploadFile/views.py index 69dfb996..b9f0cd3d 100644 --- a/server/api/views/uploadFile/views.py +++ b/server/api/views/uploadFile/views.py @@ -1,8 +1,9 @@ from rest_framework.views import APIView from rest_framework.permissions import AllowAny, IsAuthenticated from rest_framework.response import Response -from rest_framework import status +from rest_framework import status, serializers as drf_serializers from rest_framework.generics import UpdateAPIView +from drf_spectacular.utils import extend_schema, inline_serializer, OpenApiResponse import pdfplumber from .models import UploadFile # Import your UploadFile model from .serializers import UploadFileSerializer @@ -15,6 +16,8 @@ class UploadFileView(APIView): + serializer_class = UploadFileSerializer + def get_permissions(self): if self.request.method == 'GET': return [AllowAny()] # Public access @@ -28,6 +31,23 @@ def get(self, request, format=None): serializer = UploadFileSerializer(files, many=True) return Response(serializer.data) + @extend_schema( + request={'multipart/form-data': inline_serializer( + name='UploadFileRequest', + fields={ + 'file': drf_serializers.FileField(help_text='PDF file to upload'), + } + )}, + responses={ + 201: inline_serializer(name='UploadFileSuccess', fields={ + 'message': drf_serializers.CharField(), + 'file_id': drf_serializers.IntegerField(), + }), + 400: inline_serializer(name='UploadFileBadRequest', fields={ + 'message': drf_serializers.CharField(), + }), + } + ) def post(self, request, format=None): print(request.auth) print(f"UploadFileView post called. Path: {request.path}") @@ -127,6 +147,22 @@ def post(self, request, format=None): return Response({"message": f"Error processing file and embeddings: {str(e)}"}, status=status.HTTP_400_BAD_REQUEST) + @extend_schema( + request=inline_serializer(name='DeleteFileRequest', fields={ + 'guid': drf_serializers.CharField(help_text='GUID of file to delete'), + }), + responses={ + 200: inline_serializer(name='DeleteFileSuccess', fields={ + 'message': drf_serializers.CharField(), + }), + 403: inline_serializer(name='DeleteFileForbidden', fields={ + 'message': drf_serializers.CharField(), + }), + 404: inline_serializer(name='DeleteFileNotFound', fields={ + 'message': drf_serializers.CharField(), + }), + } + ) def delete(self, request, format=None): guid = request.data.get('guid') if not guid: @@ -157,6 +193,14 @@ def delete(self, request, format=None): class RetrieveUploadFileView(APIView): permission_classes = [AllowAny] + @extend_schema( + responses={ + (200, 'application/pdf'): OpenApiResponse(description='PDF file binary content'), + 404: inline_serializer(name='RetrieveFileNotFound', fields={ + 'message': drf_serializers.CharField(), + }), + } + ) def get(self, request, guid, format=None): try: file = UploadFile.objects.get(guid=guid) diff --git a/server/api/views/version/views.py b/server/api/views/version/views.py index b79d6577..af59e9e0 100644 --- a/server/api/views/version/views.py +++ b/server/api/views/version/views.py @@ -3,11 +3,18 @@ from rest_framework.permissions import AllowAny from rest_framework.views import APIView from rest_framework.response import Response +from rest_framework import serializers as drf_serializers +from drf_spectacular.utils import extend_schema, inline_serializer class VersionView(APIView): permission_classes = [AllowAny] + @extend_schema( + responses={200: inline_serializer(name='VersionResponse', fields={ + 'version': drf_serializers.CharField(), + })} + ) def get(self, request, *args, **kwargs): version = os.environ.get("VERSION") or "dev" return Response({"version": version}) From 5ce77823c25f0a45fbcdc828e13214b515496bc2 Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Fri, 27 Feb 2026 16:09:05 -0500 Subject: [PATCH 18/46] Preload SentenceTransformer model at Django startup before traffic is routed to the application instance --- server/api/apps.py | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/server/api/apps.py b/server/api/apps.py index 66656fd2..4d502cba 100644 --- a/server/api/apps.py +++ b/server/api/apps.py @@ -4,3 +4,7 @@ class ApiConfig(AppConfig): default_auto_field = 'django.db.models.BigAutoField' name = 'api' + + def ready(self): + from .services.sentencetTransformer_model import TransformerModel + TransformerModel.get_instance() From 7085aa0c71f210bbe450034deb5ae5af1ef6cbef Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Mon, 2 Mar 2026 11:59:48 -0500 Subject: [PATCH 19/46] Requested changes: fix patch decorators to point to where openAI is used, not where it's defined --- server/api/views/uploadFile/test_title.py | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/server/api/views/uploadFile/test_title.py b/server/api/views/uploadFile/test_title.py index 5391ed05..07e1b1ba 100644 --- a/server/api/views/uploadFile/test_title.py +++ b/server/api/views/uploadFile/test_title.py @@ -53,7 +53,7 @@ def test_falls_back_to_first_page_text_if_metadata_title_does_not_match_regex(se expected_title = "Advances in Mood Disorder Pharmacotherapy: Evaluating New Antipsychotics and Mood Stabilizers for Bipolar Disorder and Schizophrenia" self.assertEqual(expected_title, title.generate_title(doc)) - @patch("api.services.openai_services.openAIServices.openAI") + @patch("api.views.uploadFile.title.openAIServices.openAI") def test_falls_back_to_chatgpt_if_no_title_found(self, mock_openAI): doc = MagicMock() doc.metadata = {"title": None} @@ -68,7 +68,7 @@ def test_falls_back_to_chatgpt_if_no_title_found(self, mock_openAI): self.assertTrue(mock_openAI.called) - @patch("api.services.openai_services.openAIServices.openAI") + @patch("api.views.uploadFile.title.openAIServices.openAI") def test_strips_quotes_from_openai_title(self, mock_openAI): doc = MagicMock() doc.metadata = {"title": None} @@ -83,7 +83,7 @@ def test_strips_quotes_from_openai_title(self, mock_openAI): self.assertEqual(result, "Updated CANMAT/ISBD Guidelines for Treating Mixed Features in Bipolar Disorder") - @patch("api.services.openai_services.openAIServices.openAI") + @patch("api.views.uploadFile.title.openAIServices.openAI") def test_truncates_long_openai_title(self, mock_openAI): doc = MagicMock() doc.metadata = {"title": None} From e6754df366d1762a1b2028f0ec296c2bbd5eb3b6 Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Mon, 2 Mar 2026 12:07:57 -0500 Subject: [PATCH 20/46] Requested changes: added comments explaining title truncation --- server/api/views/uploadFile/test_title.py | 1 + server/api/views/uploadFile/title.py | 1 + 2 files changed, 2 insertions(+) diff --git a/server/api/views/uploadFile/test_title.py b/server/api/views/uploadFile/test_title.py index 07e1b1ba..d5945da8 100644 --- a/server/api/views/uploadFile/test_title.py +++ b/server/api/views/uploadFile/test_title.py @@ -96,4 +96,5 @@ def test_truncates_long_openai_title(self, mock_openAI): result = title.generate_title(doc) + # Ensure the title is truncated to fit the UploadFile model's title field (max_length=255), since OpenAI responses may exceed this limit self.assertLessEqual(len(result), 255) diff --git a/server/api/views/uploadFile/title.py b/server/api/views/uploadFile/title.py index b3f8aded..17f52a74 100644 --- a/server/api/views/uploadFile/title.py +++ b/server/api/views/uploadFile/title.py @@ -59,4 +59,5 @@ def summarize_pdf(pdf: fitz.Document) -> str: response = openAIServices.openAI( first_page_content, prompt, model='gpt-4o', temp=0.0) title = response.choices[0].message.content.strip().strip('"').strip("'") + # Truncate to fit UploadFile model's max_length=255 title field as a final safeguard return title[:255] From 4b4d7275ed7d580c9ee7b7d51287cdb99a78b9bc Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Mon, 2 Mar 2026 12:15:32 -0500 Subject: [PATCH 21/46] Fix mock setups to match how generate_title accesses title --- server/api/views/uploadFile/test_title.py | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/server/api/views/uploadFile/test_title.py b/server/api/views/uploadFile/test_title.py index d5945da8..0ec9e1bc 100644 --- a/server/api/views/uploadFile/test_title.py +++ b/server/api/views/uploadFile/test_title.py @@ -57,7 +57,7 @@ def test_falls_back_to_first_page_text_if_metadata_title_does_not_match_regex(se def test_falls_back_to_chatgpt_if_no_title_found(self, mock_openAI): doc = MagicMock() doc.metadata = {"title": None} - doc.get_text.return_value = [] + doc[0].get_text.return_value = [] mock_response = MagicMock() mock_response.choices = [MagicMock()] @@ -72,7 +72,7 @@ def test_falls_back_to_chatgpt_if_no_title_found(self, mock_openAI): def test_strips_quotes_from_openai_title(self, mock_openAI): doc = MagicMock() doc.metadata = {"title": None} - doc.get_text.return_value = [] + doc[0].get_text.return_value = [] mock_response = MagicMock() mock_response.choices = [MagicMock()] @@ -87,7 +87,7 @@ def test_strips_quotes_from_openai_title(self, mock_openAI): def test_truncates_long_openai_title(self, mock_openAI): doc = MagicMock() doc.metadata = {"title": None} - doc.get_text.return_value = [] + doc[0].get_text.return_value = [] mock_response = MagicMock() mock_response.choices = [MagicMock()] From e8b0fc1de6be8f89c500a117a4533ad26e632200 Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Thu, 5 Mar 2026 18:56:12 -0500 Subject: [PATCH 22/46] fix: treat openAIServices.openAI() return value as string --- server/api/views/uploadFile/title.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/server/api/views/uploadFile/title.py b/server/api/views/uploadFile/title.py index 17f52a74..41e67757 100644 --- a/server/api/views/uploadFile/title.py +++ b/server/api/views/uploadFile/title.py @@ -58,6 +58,6 @@ def summarize_pdf(pdf: fitz.Document) -> str: prompt = "Please provide a title for this document. The title should be less than 256 characters and will be displayed on a webpage." response = openAIServices.openAI( first_page_content, prompt, model='gpt-4o', temp=0.0) - title = response.choices[0].message.content.strip().strip('"').strip("'") + title = response.strip().strip('"').strip("'") # Truncate to fit UploadFile model's max_length=255 title field as a final safeguard return title[:255] From e0b7c23bae2a1284202a23a19b16293cecdf4411 Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Thu, 5 Mar 2026 19:01:34 -0500 Subject: [PATCH 23/46] fix mock test setup to return string instead of mocked response object --- server/api/views/uploadFile/test_title.py | 15 +++------------ 1 file changed, 3 insertions(+), 12 deletions(-) diff --git a/server/api/views/uploadFile/test_title.py b/server/api/views/uploadFile/test_title.py index 0ec9e1bc..6d95905c 100644 --- a/server/api/views/uploadFile/test_title.py +++ b/server/api/views/uploadFile/test_title.py @@ -59,10 +59,7 @@ def test_falls_back_to_chatgpt_if_no_title_found(self, mock_openAI): doc.metadata = {"title": None} doc[0].get_text.return_value = [] - mock_response = MagicMock() - mock_response.choices = [MagicMock()] - mock_response.choices[0].message.content = "A Study Regarding The Efficacy of Drugs" - mock_openAI.return_value = mock_response + mock_openAI.return_value = "A Study Regarding The Efficacy of Drugs" title.generate_title(doc) @@ -74,10 +71,7 @@ def test_strips_quotes_from_openai_title(self, mock_openAI): doc.metadata = {"title": None} doc[0].get_text.return_value = [] - mock_response = MagicMock() - mock_response.choices = [MagicMock()] - mock_response.choices[0].message.content = '"Updated CANMAT/ISBD Guidelines for Treating Mixed Features in Bipolar Disorder"' - mock_openAI.return_value = mock_response + mock_openAI.return_value = '"Updated CANMAT/ISBD Guidelines for Treating Mixed Features in Bipolar Disorder"' result = title.generate_title(doc) @@ -89,10 +83,7 @@ def test_truncates_long_openai_title(self, mock_openAI): doc.metadata = {"title": None} doc[0].get_text.return_value = [] - mock_response = MagicMock() - mock_response.choices = [MagicMock()] - mock_response.choices[0].message.content = "A" * 300 - mock_openAI.return_value = mock_response + mock_openAI.return_value = "A" * 300 result = title.generate_title(doc) From d68fa62b03e9890cab17b46164a4bda6511a5c2f Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Thu, 5 Mar 2026 19:05:46 -0500 Subject: [PATCH 24/46] fix to make test_falls_back_to_chatgpt_if_no_title_found more robust --- server/api/views/uploadFile/test_title.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/server/api/views/uploadFile/test_title.py b/server/api/views/uploadFile/test_title.py index 6d95905c..e57ee507 100644 --- a/server/api/views/uploadFile/test_title.py +++ b/server/api/views/uploadFile/test_title.py @@ -61,9 +61,10 @@ def test_falls_back_to_chatgpt_if_no_title_found(self, mock_openAI): mock_openAI.return_value = "A Study Regarding The Efficacy of Drugs" - title.generate_title(doc) + result = title.generate_title(doc) self.assertTrue(mock_openAI.called) + self.assertEqual(result, "A Study Regarding The Efficacy of Drugs") @patch("api.views.uploadFile.title.openAIServices.openAI") def test_strips_quotes_from_openai_title(self, mock_openAI): From 03b7639bcddf837a78f542ee3eda29c75bc669c3 Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Thu, 5 Mar 2026 22:25:50 -0500 Subject: [PATCH 25/46] Replace block-position title extraction with font-size-based approach The old "scan first couple pages" logic used get_text("blocks") and picked the first block matching a title regex, which frequently selected preambles, journal names, and article headers instead of the actual title. The new approach uses get_text("dict") to find the largest font size across the first few pages and collects contiguous runs of text at that size, since research paper titles are typically the largest font. --- server/api/views/uploadFile/title.py | 91 +++++++++++++++++++++------- 1 file changed, 68 insertions(+), 23 deletions(-) diff --git a/server/api/views/uploadFile/title.py b/server/api/views/uploadFile/title.py index 17f52a74..76afbfd3 100644 --- a/server/api/views/uploadFile/title.py +++ b/server/api/views/uploadFile/title.py @@ -15,35 +15,80 @@ def generate_title(pdf: fitz.Document) -> str | None: document_metadata_title = pdf.metadata["title"] if document_metadata_title is not None and document_metadata_title != "": if title_regex.match(document_metadata_title): - print("suitable title was found in metadata") return document_metadata_title.strip() - else: - print("metadata title did not match regex") - print("Looking for title in first page text") - first_page = pdf[0] - first_page_blocks = first_page.get_text("blocks") - text_blocks = [ - block[4].strip().replace("\n", " ") - for block in first_page_blocks - if block[6] == 0 # only include text blocks. - ] - - # For some reason, extracted PDF text has extra spaces. Collapse them here. - regex = r"\s{2,}" - text_blocks = [re.sub(regex, " ", text) for text in text_blocks] - - if len(text_blocks) != 0: - for text in text_blocks: - if title_regex.match(text): - return text - - print( - "no suitable title found in first page text. Using GPT-4 to summarize the PDF") + font_title = extract_title_by_font_size(pdf) + if font_title: + return font_title + gpt_title = summarize_pdf(pdf) return gpt_title or None +def extract_title_by_font_size(pdf: fitz.Document, max_pages: int = 3) -> str | None: + """ + Extract the title by finding the largest font size across the first few pages + and collecting contiguous runs of text at that size. + """ + pages_to_scan = min(max_pages, len(pdf)) + + # First pass: collect all spans with their font size, and find the max font size. + all_spans = [] + max_font_size = 0.0 + + for page_idx in range(pages_to_scan): + page_dict = pdf[page_idx].get_text("dict") + for block in page_dict["blocks"]: + if block.get("type") != 0: + continue + for line in block["lines"]: + for span in line["spans"]: + text = span["text"].strip() + size = span["size"] + if len(text) < 2 or size < 6.0: + continue + all_spans.append({"text": text, "size": size}) + if size > max_font_size: + max_font_size = size + + if max_font_size == 0.0: + return None + + # Second pass: gather contiguous runs of spans at the max font size. + # Runs continue across block boundaries so multi-block titles (e.g., + # "BIPOLAR DISORDER IN PRIMARY CARE:" in one block and "DIAGNOSIS AND + # MANAGEMENT" in the next) are joined into a single candidate. + # A run only ends when a non-max-size span interrupts it. + candidates = [] + current_run = [] + + for span in all_spans: + if span["size"] == max_font_size: + current_run.append(span["text"]) + else: + if current_run: + candidates.append(" ".join(current_run)) + current_run = [] + + if current_run: + candidates.append(" ".join(current_run)) + + # Collapse extra whitespace, validate against title regex, and pick the longest match. + # Longest wins because real titles are typically longer than section headers + # (e.g., "About the Author") that may share the same max font size. + best = None + for candidate in candidates: + cleaned = re.sub(r"\s{2,}", " ", candidate).strip() + if title_regex.match(cleaned): + if best is None or len(cleaned) > len(best): + best = cleaned + + if best: + return best[:255] + + return None + + def summarize_pdf(pdf: fitz.Document) -> str: """ Summarize a PDF document using OpenAI's GPT-4 model. From 2a822f60f86857672af5a8d5f688ed3ee5208326 Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Thu, 5 Mar 2026 22:26:27 -0500 Subject: [PATCH 26/46] loosens the title regex to allow years, question marks, apostrophes, and non-breaking spaces in titles. --- server/api/views/uploadFile/title.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/server/api/views/uploadFile/title.py b/server/api/views/uploadFile/title.py index 76afbfd3..f4e562c1 100644 --- a/server/api/views/uploadFile/title.py +++ b/server/api/views/uploadFile/title.py @@ -6,9 +6,9 @@ # regular expression to match common research white paper titles. Created by Chat-gpt -# requires at least 3 words, no dates, no version numbers. +# requires at least 3 words, no version numbers. title_regex = re.compile( - r'^(?=(?:\b\w+\b[\s:,\-\(\)]*){3,})(?!.*\b(?:19|20)\d{2}\b)(?!.*\bv\d+\b)[A-Za-z0-9][\w\s:,\-\(\)]*[A-Za-z\)]$', re.IGNORECASE) + r"^(?=(?:\b\w+\b[^A-Za-z0-9]*){3,})(?!.*\bv\d+\b)[A-Za-z0-9].+[A-Za-z\)?!]$", re.IGNORECASE) def generate_title(pdf: fitz.Document) -> str | None: From edf1eb64a77f4dd153a0c075933693831dfaed9f Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Thu, 5 Mar 2026 22:26:58 -0500 Subject: [PATCH 27/46] Update tests for font-size-based title extraction Refactor test helpers to use get_text("dict") structure instead of get_text("blocks"). Add tests for multi-span joining, short span filtering, regex rejection, and multi-page title detection. --- server/api/views/uploadFile/test_title.py | 153 +++++++++++++++------- 1 file changed, 108 insertions(+), 45 deletions(-) diff --git a/server/api/views/uploadFile/test_title.py b/server/api/views/uploadFile/test_title.py index 0ec9e1bc..531a3f07 100644 --- a/server/api/views/uploadFile/test_title.py +++ b/server/api/views/uploadFile/test_title.py @@ -4,6 +4,39 @@ from . import title +def make_page_dict(blocks): + """Helper to build a get_text("dict") return value from a simple list of blocks. + Each block is a list of (text, font_size) tuples representing spans. + """ + dict_blocks = [] + for spans in blocks: + dict_blocks.append({ + "type": 0, + "lines": [{ + "spans": [{"text": text, "size": size} for text, size in spans] + }] + }) + return {"blocks": dict_blocks} + + +def make_mock_doc(pages_data, metadata=None): + """Build a mock fitz.Document. + pages_data: list of block lists, one per page. Each block is a list of (text, size) tuples. + """ + doc = MagicMock() + doc.metadata = metadata or {"title": None} + doc.__len__ = lambda self: len(pages_data) + + mock_pages = [] + for page_blocks in pages_data: + page = MagicMock() + page.get_text.return_value = make_page_dict(page_blocks) + mock_pages.append(page) + + doc.__getitem__ = lambda self, idx: mock_pages[idx] + return doc + + class TestGenerateTitle(unittest.TestCase): def test_prefers_metadata_title_if_valid(self): doc = MagicMock() @@ -11,53 +44,35 @@ def test_prefers_metadata_title_if_valid(self): self.assertEqual( "A Study Regarding The Efficacy of Drugs", title.generate_title(doc)) - def test_falls_back_to_first_page_text_if_metadata_title_is_empty(self): - doc = MagicMock() - doc.metadata = {"title": ""} - doc[0].get_text = MagicMock() - - foo_block = [None] * 7 - foo_block[4] = "foo" - foo_block[6] = 0 - - title_block = [None] * 7 - title_block[4] = "Advances in Mood Disorder Pharmacotherapy: Evaluating New Antipsychotics and Mood Stabilizers for Bipolar Disorder and Schizophrenia" - title_block[6] = 0 - - bar_block = [None] * 7 - bar_block[4] = "bar" - bar_block[6] = 0 - doc[0].get_text.return_value = [foo_block, title_block, bar_block] - + def test_falls_back_to_font_size_if_metadata_title_is_empty(self): + doc = make_mock_doc( + pages_data=[[ + [("foo", 10.0)], + [("Advances in Mood Disorder Pharmacotherapy: Evaluating New Antipsychotics and Mood Stabilizers for Bipolar Disorder and Schizophrenia", 18.0)], + [("bar", 10.0)], + ]], + metadata={"title": ""}, + ) expected_title = "Advances in Mood Disorder Pharmacotherapy: Evaluating New Antipsychotics and Mood Stabilizers for Bipolar Disorder and Schizophrenia" self.assertEqual(expected_title, title.generate_title(doc)) - def test_falls_back_to_first_page_text_if_metadata_title_does_not_match_regex(self): - doc = MagicMock() - doc.metadata = {"title": "abcd1234"} - doc[0].get_text = MagicMock() - - foo_block = [None] * 7 - foo_block[4] = "foo" - foo_block[6] = 0 - - title_block = [None] * 7 - title_block[4] = "Advances in Mood Disorder Pharmacotherapy: Evaluating New Antipsychotics and Mood Stabilizers for Bipolar Disorder and Schizophrenia" - title_block[6] = 0 - - bar_block = [None] * 7 - bar_block[4] = "bar" - bar_block[6] = 0 - doc[0].get_text.return_value = [foo_block, title_block, bar_block] - + def test_falls_back_to_font_size_if_metadata_title_does_not_match_regex(self): + doc = make_mock_doc( + pages_data=[[ + [("foo", 10.0)], + [("Advances in Mood Disorder Pharmacotherapy: Evaluating New Antipsychotics and Mood Stabilizers for Bipolar Disorder and Schizophrenia", 18.0)], + [("bar", 10.0)], + ]], + metadata={"title": "abcd1234"}, + ) expected_title = "Advances in Mood Disorder Pharmacotherapy: Evaluating New Antipsychotics and Mood Stabilizers for Bipolar Disorder and Schizophrenia" self.assertEqual(expected_title, title.generate_title(doc)) @patch("api.views.uploadFile.title.openAIServices.openAI") def test_falls_back_to_chatgpt_if_no_title_found(self, mock_openAI): - doc = MagicMock() - doc.metadata = {"title": None} - doc[0].get_text.return_value = [] + doc = make_mock_doc( + pages_data=[[]] # no blocks at all + ) mock_response = MagicMock() mock_response.choices = [MagicMock()] @@ -70,9 +85,7 @@ def test_falls_back_to_chatgpt_if_no_title_found(self, mock_openAI): @patch("api.views.uploadFile.title.openAIServices.openAI") def test_strips_quotes_from_openai_title(self, mock_openAI): - doc = MagicMock() - doc.metadata = {"title": None} - doc[0].get_text.return_value = [] + doc = make_mock_doc(pages_data=[[]]) mock_response = MagicMock() mock_response.choices = [MagicMock()] @@ -85,9 +98,7 @@ def test_strips_quotes_from_openai_title(self, mock_openAI): @patch("api.views.uploadFile.title.openAIServices.openAI") def test_truncates_long_openai_title(self, mock_openAI): - doc = MagicMock() - doc.metadata = {"title": None} - doc[0].get_text.return_value = [] + doc = make_mock_doc(pages_data=[[]]) mock_response = MagicMock() mock_response.choices = [MagicMock()] @@ -98,3 +109,55 @@ def test_truncates_long_openai_title(self, mock_openAI): # Ensure the title is truncated to fit the UploadFile model's title field (max_length=255), since OpenAI responses may exceed this limit self.assertLessEqual(len(result), 255) + + def test_font_size_joins_adjacent_spans_in_same_block(self): + """A title split across multiple spans in the same block should be joined.""" + doc = make_mock_doc( + pages_data=[[ + [("Author Name", 10.0)], + [("Advances in Mood Disorder", 18.0), ("Pharmacotherapy", 18.0)], + [("Some journal info", 10.0)], + ]], + ) + result = title.extract_title_by_font_size(doc) + self.assertEqual(result, "Advances in Mood Disorder Pharmacotherapy") + + def test_font_size_ignores_short_spans(self): + """Superscript markers and other tiny spans should be filtered out.""" + doc = make_mock_doc( + pages_data=[[ + [("Advances in Mood Disorder Pharmacotherapy", 18.0), ("*", 18.0)], + [("Author Name et al.", 10.0)], + ]], + ) + # The "*" span is < 2 chars, so it should be ignored; title is just the real text + result = title.extract_title_by_font_size(doc) + self.assertEqual(result, "Advances in Mood Disorder Pharmacotherapy") + + def test_font_size_returns_none_when_no_regex_match(self): + """If the largest-font text doesn't match the title regex, return None.""" + doc = make_mock_doc( + pages_data=[[ + # Only 2 words — regex requires at least 3 + [("Psychiatry Research", 18.0)], + [("Author Name et al.", 10.0)], + ]], + ) + result = title.extract_title_by_font_size(doc) + self.assertIsNone(result) + + def test_font_size_finds_title_on_later_page(self): + """Title on page 2 should still be found if it has the largest font.""" + doc = make_mock_doc( + pages_data=[ + [ # page 1: cover page with smaller text + [("Some preamble text here", 12.0)], + ], + [ # page 2: actual title in larger font + [("Advances in Mood Disorder Pharmacotherapy", 18.0)], + [("Author Name et al.", 10.0)], + ], + ], + ) + result = title.extract_title_by_font_size(doc) + self.assertEqual(result, "Advances in Mood Disorder Pharmacotherapy") From 4bae7460cc4a2f4805a3e49286dc49d340fd8163 Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Tue, 10 Mar 2026 12:28:50 -0400 Subject: [PATCH 28/46] update documentation to include instructions about how to use the API docs --- CLAUDE.md | 10 ++++++++++ README.md | 17 +++++++++++++++++ 2 files changed, 27 insertions(+) diff --git a/CLAUDE.md b/CLAUDE.md index 712082e7..b9f417e7 100644 --- a/CLAUDE.md +++ b/CLAUDE.md @@ -147,6 +147,16 @@ Each module contains: - Auth endpoints via Djoser: `/auth/` - JWT token lifetime: 60 minutes (access), 1 day (refresh) +#### API Documentation +- Auto-generated using **drf-spectacular** (OpenAPI 3.0) +- **Swagger UI**: `http://localhost:8000/api/docs/` — interactive API explorer +- **ReDoc**: `http://localhost:8000/api/redoc/` — readable reference docs +- **Raw schema**: `http://localhost:8000/api/schema/` +- Configuration in `SPECTACULAR_SETTINGS` in `settings.py` +- Views use `@extend_schema` decorators and `serializer_class` attributes for schema generation +- JWT auth is configured in the schema — use `JWT ` (not `Bearer`) in Swagger UI's Authorize dialog +- To document a new endpoint: add `serializer_class` to the view if it has one, or add `@extend_schema` with `inline_serializer` for views returning raw dicts + #### Key Data Models - **Medication** (`api.views.listMeds.models`) - Medication catalog with benefits/risks - **MedRule** (`api.models.model_medRule`) - Include/Exclude rules for medications based on patient history diff --git a/README.md b/README.md index e5a246b1..8988ea75 100644 --- a/README.md +++ b/README.md @@ -74,6 +74,23 @@ df = pd.read_sql(query, engine) #### Django REST - The email and password are set in `server/api/management/commands/createsu.py` +## API Documentation + +Interactive API docs are auto-generated using [drf-spectacular](https://drf-spectacular.readthedocs.io/) and available at: + +- **Swagger UI**: [http://localhost:8000/api/docs/](http://localhost:8000/api/docs/) — interactive explorer with "Try it out" functionality +- **ReDoc**: [http://localhost:8000/api/redoc/](http://localhost:8000/api/redoc/) — clean, readable reference docs +- **Raw schema**: [http://localhost:8000/api/schema/](http://localhost:8000/api/schema/) — OpenAPI 3.0 JSON/YAML + +### Testing authenticated endpoints + +Most endpoints require JWT authentication. To test them in Swagger UI: + +1. **Get a token**: Find the `POST /auth/jwt/create/` endpoint in Swagger UI, click **Try it out**, enter an authorized `email` and `password`, and click **Execute**. Copy the `access` token from the response. +2. **Authorize**: Click the **Authorize** button (lock icon) at the top of the page. Enter `JWT ` in the value field. The prefix must be `JWT`, not `Bearer`. +3. **Test endpoints**: All subsequent requests will include your token. Use **Try it out** on any protected endpoint. +4. **Token refresh**: Access tokens expire after 60 minutes. Use `POST /auth/jwt/refresh/` with your `refresh` token, or repeat step 1. + ## Architecture The Balancer website is a Postgres, Django REST, and React project. The source code layout is: From 6f0deedb5505abf5917320daa9539167adda0447 Mon Sep 17 00:00:00 2001 From: anishamahuli Date: Tue, 10 Mar 2026 19:40:16 -0400 Subject: [PATCH 29/46] update site links on README The links were for the old site and needed an update --- README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index e5a246b1..4603df49 100644 --- a/README.md +++ b/README.md @@ -5,7 +5,7 @@ for patients with bipolar disorder, helping them shorten their journey to stabil ## Usage -You can view the current build of the website here: [https://balancertestsite.com](https://balancertestsite.com/) +You can view the current build of the website here: [https://balancerproject.org/](https://balancerproject.org/) ## Contributing @@ -53,7 +53,7 @@ The application supports connecting to PostgreSQL databases via: See [Database Connection Documentation](./docs/DATABASE_CONNECTION.md) for detailed configuration. **Local Development:** -- Download a sample of papers to upload from [https://balancertestsite.com](https://balancertestsite.com/) +- Download a sample of papers to upload from [https://balancerproject.org/](https://balancerproject.org/) - The email and password of `pgAdmin` are specified in `balancer-main/docker-compose.yml` - The first time you use `pgAdmin` after building the Docker containers you will need to register the server. - The `Host name/address` is the Postgres server service name in the Docker Compose file From 795f21885e766de9f6379319c5555ef7db48d989 Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Wed, 11 Mar 2026 13:09:14 -0400 Subject: [PATCH 30/46] Run python-app workflow on pushes and PRs to develop branch --- .github/workflows/python-app.yml | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/.github/workflows/python-app.yml b/.github/workflows/python-app.yml index 41bc74d9..a6c07075 100644 --- a/.github/workflows/python-app.yml +++ b/.github/workflows/python-app.yml @@ -5,9 +5,9 @@ name: Python application on: push: - branches: [ "listOfMed" ] + branches: [ "develop" ] pull_request: - branches: [ "listOfMed" ] + branches: [ "develop" ] permissions: contents: read From d498a0057008b1b2c80bdb176e5940940899639b Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Thu, 19 Mar 2026 13:15:47 -0400 Subject: [PATCH 31/46] =?UTF-8?q?Pytest=20won=E2=80=99t=20automatically=20?= =?UTF-8?q?discover=20config=20files=20in=20subdirectories?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .github/workflows/python-app.yml | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/.github/workflows/python-app.yml b/.github/workflows/python-app.yml index a6c07075..2afa2828 100644 --- a/.github/workflows/python-app.yml +++ b/.github/workflows/python-app.yml @@ -29,6 +29,6 @@ jobs: run: ruff check --output-format=github --target-version=py39 - name: Install test dependencies run: pip install -r server/requirements.txt - # Discover and run all files matching test_*.py or *_test.py under server/ + # Pytest won’t automatically discover config files in subdirectories - name: Run tests - run: pytest server/ -v + run: pytest -c server/pytest.ini server/ -v From 3824d81ae0fb722383f7e80df07ca7ba28d0c4e6 Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Thu, 19 Mar 2026 13:47:52 -0400 Subject: [PATCH 32/46] Suppress E402 import violations --- evaluation/evals.py | 16 ++++++++-------- server/balancer_backend/urls.py | 6 +++--- 2 files changed, 11 insertions(+), 11 deletions(-) diff --git a/evaluation/evals.py b/evaluation/evals.py index 8eb7e9e6..5110076f 100755 --- a/evaluation/evals.py +++ b/evaluation/evals.py @@ -21,18 +21,18 @@ # Ensure the parent directory is in the path to import ModelFactory sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))) -import argparse -import logging -import asyncio -import time +import argparse # noqa: E402 +import logging # noqa: E402 +import asyncio # noqa: E402 +import time # noqa: E402 -import pandas as pd +import pandas as pd # noqa: E402 # lighteval depends on `sentencepiece` and it only has prebuilt wheels for Python 3.11 or below -from lighteval.tasks.requests import Doc -from lighteval.metrics.metrics_sample import Extractiveness +from lighteval.tasks.requests import Doc # noqa: E402 +from lighteval.metrics.metrics_sample import Extractiveness # noqa: E402 -from server.api.services.llm_services import ModelFactory +from server.api.services.llm_services import ModelFactory # noqa: E402 logging.basicConfig( level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s" diff --git a/server/balancer_backend/urls.py b/server/balancer_backend/urls.py index 55bd2032..cdb92dbb 100644 --- a/server/balancer_backend/urls.py +++ b/server/balancer_backend/urls.py @@ -58,9 +58,9 @@ path("api/redoc/", SpectacularRedocView.as_view(url_name="schema"), name="redoc"), ] -import os -from django.conf import settings -from django.http import HttpResponseNotFound +import os # noqa: E402 +from django.conf import settings # noqa: E402 +from django.http import HttpResponseNotFound # noqa: E402 def spa_fallback(request): From 46e9969dade55777098286cd6316bc18444e5b1f Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Fri, 20 Mar 2026 12:00:44 -0400 Subject: [PATCH 33/46] Add build_query tests and document coverage gaps in embedding_services --- server/api/services/embedding_services.py | 1 + .../api/services/test_embedding_services.py | 202 +++++++++++++++++- 2 files changed, 201 insertions(+), 2 deletions(-) diff --git a/server/api/services/embedding_services.py b/server/api/services/embedding_services.py index 3fa9bb68..aca99133 100644 --- a/server/api/services/embedding_services.py +++ b/server/api/services/embedding_services.py @@ -2,6 +2,7 @@ import logging from statistics import median +# filter() only does ADD logic from django.db.models import Q from pgvector.django import L2Distance diff --git a/server/api/services/test_embedding_services.py b/server/api/services/test_embedding_services.py index 677c1e7b..ea322645 100644 --- a/server/api/services/test_embedding_services.py +++ b/server/api/services/test_embedding_services.py @@ -1,9 +1,175 @@ from unittest.mock import MagicMock, patch -from api.services.embedding_services import evaluate_query, log_usage +from django.db.models import Q +from api.services.embedding_services import build_query, evaluate_query, log_usage + +# --------------------------------------------------------------------------- +# build_query tests +# +# build_query only constructs a lazy Django QuerySet — it never evaluates it +# (no iteration, .get(), .exists(), etc.), so no database is needed. +# +# We patch Embeddings.objects so every chained ORM call (.filter, .annotate, +# .order_by, __getitem__) returns a MagicMock instead of hitting the DB. +# All assertions inspect which methods were called with which arguments. +# --------------------------------------------------------------------------- + +# Only forwarded to L2Distance +EMBEDDING_VECTOR = [0.1, 0.2, 0.3] + +# Test authenticated/unauthenticated user access control + +@patch("api.services.embedding_services.Embeddings.objects") +def test_build_query_authenticated_uses_or_filter(mock_objects): + # An authenticated user should see their own files OR files uploaded by a + # superuser. The initial filter must use an OR-connected Q expression. + user = MagicMock(is_authenticated=True) + + build_query(user, EMBEDDING_VECTOR) + + # Q objects support equality comparison in pure Python — no DB needed. + expected_q = Q(upload_file__uploaded_by=user) | Q(upload_file__uploaded_by__is_superuser=True) + actual_q = mock_objects.filter.call_args.args[0] + assert actual_q == expected_q + + +@patch("api.services.embedding_services.Embeddings.objects") +def test_build_query_unauthenticated_uses_superuser_only_filter(mock_objects): + # An unauthenticated user may only see files uploaded by superusers. + # The OR branch for the user's own files must NOT be present. + user = MagicMock(is_authenticated=False) + + build_query(user, EMBEDDING_VECTOR) + + expected_q = Q(upload_file__uploaded_by__is_superuser=True) + actual_q = mock_objects.filter.call_args.args[0] + assert actual_q == expected_q + +# Test application of annotate and order_by + +# TODO: Strengthen test_build_query_annotates_and_orders_by_distance to also +# assert the *arguments* to annotate — specifically that it receives +# distance=L2Distance("embedding_sentence_transformers", EMBEDDING_VECTOR). +# Currently only the call count is checked, so a wrong field name or a +# dropped vector would go undetected. + +@patch("api.services.embedding_services.Embeddings.objects") +def test_build_query_annotates_and_orders_by_distance(mock_objects): + # Regardless of other arguments, annotate(distance=L2Distance(...)) and + # order_by("distance") must always be applied to the queryset. + user = MagicMock(is_authenticated=True) + + build_query(user, EMBEDDING_VECTOR) + + # Retrieve the mock chain that .filter() returned, then check its methods. + filtered_qs = mock_objects.filter.return_value + filtered_qs.annotate.assert_called_once() + filtered_qs.annotate.return_value.order_by.assert_called_once_with("distance") + +# Test guid-over-document precedence logic + +@patch("api.services.embedding_services.Embeddings.objects") +def test_build_query_no_document_filter_when_both_none(mock_objects): + # When neither guid nor document_name is provided, only the access-control + # filter should fire — no secondary filter call for a document. + user = MagicMock(is_authenticated=True) + + build_query(user, EMBEDDING_VECTOR, document_name=None, guid=None) + + # Exactly one filter call: the auth/access-control filter. + assert mock_objects.filter.call_count == 1 + + + +@patch("api.services.embedding_services.Embeddings.objects") +def test_build_query_guid_takes_precedence_over_document_name(mock_objects): + # When both guid and document_name are provided, the guid branch runs and + # the document_name branch is skipped entirely (only two filter calls total). + user = MagicMock(is_authenticated=True) + + build_query(user, EMBEDDING_VECTOR, guid="abc-123", document_name="study.pdf") + + # Two calls: auth filter + guid filter. No third call for document_name. + assert mock_objects.filter.call_count == 2 + + # The second filter must use upload_file__guid, not name. + # We follow the mock chain to the queryset that .annotate().order_by() returned. + ordered_qs = mock_objects.filter.return_value.annotate.return_value.order_by.return_value + ordered_qs.filter.assert_called_once_with(upload_file__guid="abc-123") + + +@patch("api.services.embedding_services.Embeddings.objects") +def test_build_query_guid_filter_applied(mock_objects): + # When only guid is given, a second filter on upload_file__guid is applied. + user = MagicMock(is_authenticated=True) + + build_query(user, EMBEDDING_VECTOR, guid="doc-guid-456") + + ordered_qs = mock_objects.filter.return_value.annotate.return_value.order_by.return_value + ordered_qs.filter.assert_called_once_with(upload_file__guid="doc-guid-456") + + +@patch("api.services.embedding_services.Embeddings.objects") +def test_build_query_document_name_filter_applied(mock_objects): + # When only document_name is given (guid is None), a second filter on + # name is applied instead of upload_file__guid. + user = MagicMock(is_authenticated=True) + + build_query(user, EMBEDDING_VECTOR, document_name="study.pdf", guid=None) + + ordered_qs = mock_objects.filter.return_value.annotate.return_value.order_by.return_value + ordered_qs.filter.assert_called_once_with(name="study.pdf") + + +@patch("api.services.embedding_services.Embeddings.objects") +def test_build_query_empty_string_guid_falls_back_to_document_name(mock_objects): + # An empty-string guid is falsy in Python, so it should not trigger the + # guid branch. The document_name filter should fire instead. This guards + # against callers passing guid="" from an unset form field. + user = MagicMock(is_authenticated=True) + + build_query(user, EMBEDDING_VECTOR, guid="", document_name="fallback.pdf") + + ordered_qs = mock_objects.filter.return_value.annotate.return_value.order_by.return_value + ordered_qs.filter.assert_called_once_with(name="fallback.pdf") + +# Cover LIMIT slicing + +@patch("api.services.embedding_services.Embeddings.objects") +def test_build_query_respects_num_results(mock_objects): + # num_results controls the SQL LIMIT via queryset slicing. Verify that a + # non-default value propagates correctly to the __getitem__ call. + user = MagicMock(is_authenticated=True) + + build_query(user, EMBEDDING_VECTOR, num_results=5) + + # Django translates qs[:5] into qs.__getitem__(slice(None, 5, None)). + ordered_qs = mock_objects.filter.return_value.annotate.return_value.order_by.return_value + ordered_qs.__getitem__.assert_called_once_with(slice(None, 5, None)) + +@patch("api.services.embedding_services.Embeddings.objects") +def test_build_query_returns_unevaluated_queryset(mock_objects): + # build_query must NOT evaluate the queryset (no list(), no iteration). + # The return value should be the mock produced by the final __getitem__ call. + user = MagicMock(is_authenticated=True) + + result = build_query(user, EMBEDDING_VECTOR) + + ordered_qs = mock_objects.filter.return_value.annotate.return_value.order_by.return_value + assert result is ordered_qs.__getitem__.return_value + assert not isinstance(result, list) + + +# --------------------------------------------------------------------------- +# evaluate_query tests +# --------------------------------------------------------------------------- + +# TODO: Add test for empty queryset — evaluate_query([]) should return []. def test_evaluate_query_maps_fields(): + # Verify that each Embeddings model attribute is mapped to the correct + # output dict key. Note the rename: obj.page_num -> result["page_number"]. obj = MagicMock() obj.name = "doc.pdf" obj.text = "some text" @@ -27,6 +193,8 @@ def test_evaluate_query_maps_fields(): def test_evaluate_query_none_upload_file(): + # When upload_file is None (e.g. the FK was deleted), file_id must be None + # rather than raising an AttributeError on None.guid. obj = MagicMock() obj.name = "doc.pdf" obj.text = "some text" @@ -39,9 +207,26 @@ def test_evaluate_query_none_upload_file(): assert results[0]["file_id"] is None +# --------------------------------------------------------------------------- +# log_usage tests +# --------------------------------------------------------------------------- + +# TODO: Add test for empty results list — log_usage([]) hits the else branch and +# should call SemanticSearchUsage.objects.create with num_results_returned=0 +# and max_distance=None, median_distance=None, min_distance=None. + +# TODO: Add test for unauthenticated user — user.is_authenticated=False should +# result in user=None being stored in the SemanticSearchUsage record. + +# TODO: Add test for user=None — passing None directly as the user argument +# should also store user=None (the expression `user if (user and +# user.is_authenticated) else None` handles both cases, but only the +# authenticated path is currently exercised). @patch("api.services.embedding_services.SemanticSearchUsage.objects.create") def test_log_usage_computes_distance_stats(mock_create): + # Verify min, max, and median are computed correctly from the distance + # values in the results list and forwarded to the DB record. results = [{"distance": 1.0}, {"distance": 3.0}, {"distance": 2.0}] user = MagicMock(is_authenticated=True) @@ -69,10 +254,12 @@ def test_log_usage_computes_distance_stats(mock_create): side_effect=Exception("DB error"), ) def test_log_usage_swallows_exceptions(mock_create): + # log_usage must not propagate exceptions — a logging failure should never + # interrupt the caller's search flow. + # pytest fails the test if it catches unhandled Exception results = [{"distance": 1.0}] user = MagicMock(is_authenticated=True) - # pytest fails the test if it catches unhandled Exception log_usage( results, message_data="test query", @@ -83,3 +270,14 @@ def test_log_usage_swallows_exceptions(mock_create): encoding_time=0.1, db_query_time=0.2, ) + + +# --------------------------------------------------------------------------- +# get_closest_embeddings tests +# --------------------------------------------------------------------------- + +# TODO: Add smoke test for get_closest_embeddings verifying the wiring between +# its three steps: encode → build_query → evaluate_query → log_usage. +# Patch TransformerModel.get_instance, build_query, evaluate_query, and +# log_usage. Assert that evaluate_query receives the queryset returned by +# build_query, and that the function returns evaluate_query's result. From 64a19ef56dad988d1a2ea61a0b409c7878d0370c Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Fri, 20 Mar 2026 12:39:51 -0400 Subject: [PATCH 34/46] Fill test gaps in test_embedding_services --- server/api/services/embedding_services.py | 2 +- .../api/services/test_embedding_services.py | 149 ++++++++++++++---- 2 files changed, 119 insertions(+), 32 deletions(-) diff --git a/server/api/services/embedding_services.py b/server/api/services/embedding_services.py index aca99133..dada28a2 100644 --- a/server/api/services/embedding_services.py +++ b/server/api/services/embedding_services.py @@ -2,7 +2,7 @@ import logging from statistics import median -# filter() only does ADD logic +# Django filter() only does ADD logic from django.db.models import Q from pgvector.django import L2Distance diff --git a/server/api/services/test_embedding_services.py b/server/api/services/test_embedding_services.py index ea322645..8cbc1be9 100644 --- a/server/api/services/test_embedding_services.py +++ b/server/api/services/test_embedding_services.py @@ -1,20 +1,21 @@ from unittest.mock import MagicMock, patch from django.db.models import Q +from pgvector.django import L2Distance -from api.services.embedding_services import build_query, evaluate_query, log_usage +from api.services.embedding_services import ( + build_query, + evaluate_query, + get_closest_embeddings, + log_usage, +) # --------------------------------------------------------------------------- # build_query tests -# -# build_query only constructs a lazy Django QuerySet — it never evaluates it -# (no iteration, .get(), .exists(), etc.), so no database is needed. -# -# We patch Embeddings.objects so every chained ORM call (.filter, .annotate, -# .order_by, __getitem__) returns a MagicMock instead of hitting the DB. -# All assertions inspect which methods were called with which arguments. # --------------------------------------------------------------------------- +# All assertions inspect which methods and arguments were called on Embeddings.objects + # Only forwarded to L2Distance EMBEDDING_VECTOR = [0.1, 0.2, 0.3] @@ -48,12 +49,6 @@ def test_build_query_unauthenticated_uses_superuser_only_filter(mock_objects): # Test application of annotate and order_by -# TODO: Strengthen test_build_query_annotates_and_orders_by_distance to also -# assert the *arguments* to annotate — specifically that it receives -# distance=L2Distance("embedding_sentence_transformers", EMBEDDING_VECTOR). -# Currently only the call count is checked, so a wrong field name or a -# dropped vector would go undetected. - @patch("api.services.embedding_services.Embeddings.objects") def test_build_query_annotates_and_orders_by_distance(mock_objects): # Regardless of other arguments, annotate(distance=L2Distance(...)) and @@ -67,6 +62,12 @@ def test_build_query_annotates_and_orders_by_distance(mock_objects): filtered_qs.annotate.assert_called_once() filtered_qs.annotate.return_value.order_by.assert_called_once_with("distance") + # L2Distance is a Django Func subclass, which implements __eq__ by comparing + # class and source expressions — so we can assert the exact field name and + # vector without patching L2Distance itself. + actual_distance_expr = filtered_qs.annotate.call_args.kwargs["distance"] + assert actual_distance_expr == L2Distance("embedding_sentence_transformers", EMBEDDING_VECTOR) + # Test guid-over-document precedence logic @patch("api.services.embedding_services.Embeddings.objects") @@ -165,7 +166,10 @@ def test_build_query_returns_unevaluated_queryset(mock_objects): # evaluate_query tests # --------------------------------------------------------------------------- -# TODO: Add test for empty queryset — evaluate_query([]) should return []. +def test_evaluate_query_empty_queryset(): + # An empty iterable should return an empty list, not raise an exception. + assert evaluate_query([]) == [] + def test_evaluate_query_maps_fields(): # Verify that each Embeddings model attribute is mapped to the correct @@ -193,8 +197,8 @@ def test_evaluate_query_maps_fields(): def test_evaluate_query_none_upload_file(): - # When upload_file is None (e.g. the FK was deleted), file_id must be None - # rather than raising an AttributeError on None.guid. + # When upload_file is None, file_id must be None rather than raising + # an AttributeError on None.guid. obj = MagicMock() obj.name = "doc.pdf" obj.text = "some text" @@ -211,17 +215,71 @@ def test_evaluate_query_none_upload_file(): # log_usage tests # --------------------------------------------------------------------------- -# TODO: Add test for empty results list — log_usage([]) hits the else branch and -# should call SemanticSearchUsage.objects.create with num_results_returned=0 -# and max_distance=None, median_distance=None, min_distance=None. +@patch("api.services.embedding_services.SemanticSearchUsage.objects.create") +def test_log_usage_empty_results(mock_create): + # Empty results hits the else branch. The record should still be created + # with num_results_returned=0 and all distance fields set to None. + user = MagicMock(is_authenticated=True) + + log_usage( + [], + message_data="test query", + user=user, + guid=None, + document_name=None, + num_results=10, + encoding_time=0.1, + db_query_time=0.2, + ) + + mock_create.assert_called_once() + kwargs = mock_create.call_args.kwargs + assert kwargs["num_results_returned"] == 0 + assert kwargs["max_distance"] is None + assert kwargs["median_distance"] is None + assert kwargs["min_distance"] is None + + +@patch("api.services.embedding_services.SemanticSearchUsage.objects.create") +def test_log_usage_unauthenticated_user_stored_as_none(mock_create): + # An unauthenticated user should be stored as None in the DB record, not as + # the user object itself, so the FK constraint is not violated. + user = MagicMock(is_authenticated=False) + + log_usage( + [{"distance": 1.0}], + message_data="test query", + user=user, + guid=None, + document_name=None, + num_results=10, + encoding_time=0.1, + db_query_time=0.2, + ) + + kwargs = mock_create.call_args.kwargs + assert kwargs["user"] is None + + +@patch("api.services.embedding_services.SemanticSearchUsage.objects.create") +def test_log_usage_none_user_stored_as_none(mock_create): + # Passing user=None directly (e.g. from an anonymous request) should also + # store None — the expression `user if (user and user.is_authenticated)` + # short-circuits on the falsy None before accessing .is_authenticated. + log_usage( + [{"distance": 1.0}], + message_data="test query", + user=None, + guid=None, + document_name=None, + num_results=10, + encoding_time=0.1, + db_query_time=0.2, + ) -# TODO: Add test for unauthenticated user — user.is_authenticated=False should -# result in user=None being stored in the SemanticSearchUsage record. + kwargs = mock_create.call_args.kwargs + assert kwargs["user"] is None -# TODO: Add test for user=None — passing None directly as the user argument -# should also store user=None (the expression `user if (user and -# user.is_authenticated) else None` handles both cases, but only the -# authenticated path is currently exercised). @patch("api.services.embedding_services.SemanticSearchUsage.objects.create") def test_log_usage_computes_distance_stats(mock_create): @@ -276,8 +334,37 @@ def test_log_usage_swallows_exceptions(mock_create): # get_closest_embeddings tests # --------------------------------------------------------------------------- -# TODO: Add smoke test for get_closest_embeddings verifying the wiring between -# its three steps: encode → build_query → evaluate_query → log_usage. -# Patch TransformerModel.get_instance, build_query, evaluate_query, and -# log_usage. Assert that evaluate_query receives the queryset returned by -# build_query, and that the function returns evaluate_query's result. +@patch("api.services.embedding_services.log_usage") +@patch("api.services.embedding_services.evaluate_query") +@patch("api.services.embedding_services.build_query") +@patch("api.services.embedding_services.TransformerModel") +def test_get_closest_embeddings_wiring(mock_transformer, mock_build, mock_evaluate, mock_log): + # Smoke test verifying that get_closest_embeddings correctly wires together + # encode → build_query → evaluate_query → log_usage and returns the results. + user = MagicMock(is_authenticated=True) + + # Simulate the model encoding the message to a vector. + fake_vector = [0.1, 0.2, 0.3] + mock_transformer.get_instance.return_value.model.encode.return_value = fake_vector + + # build_query returns a queryset; evaluate_query turns it into a results list. + fake_queryset = MagicMock() + mock_build.return_value = fake_queryset + fake_results = [{"name": "doc.pdf", "distance": 0.5}] + mock_evaluate.return_value = fake_results + + result = get_closest_embeddings(user, "some query", document_name="doc.pdf", guid=None, num_results=5) + + # The encoded vector must be forwarded to build_query. + mock_build.assert_called_once_with(user, fake_vector, "doc.pdf", None, 5) + + # evaluate_query must receive the queryset that build_query returned. + mock_evaluate.assert_called_once_with(fake_queryset) + + # log_usage must be called with the results and original parameters. + mock_log.assert_called_once() + log_kwargs = mock_log.call_args.args + assert log_kwargs[0] is fake_results + + # The function must return evaluate_query's result unchanged. + assert result is fake_results From dec3c12a71c1fefc81f30768fe2aec8e48df2fb8 Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Fri, 20 Mar 2026 14:05:17 -0400 Subject: [PATCH 35/46] Fix incorrect build_query test assertions --- .../api/services/test_embedding_services.py | 19 ++++++++++--------- 1 file changed, 10 insertions(+), 9 deletions(-) diff --git a/server/api/services/test_embedding_services.py b/server/api/services/test_embedding_services.py index 8cbc1be9..dcbb2fc7 100644 --- a/server/api/services/test_embedding_services.py +++ b/server/api/services/test_embedding_services.py @@ -38,14 +38,13 @@ def test_build_query_authenticated_uses_or_filter(mock_objects): @patch("api.services.embedding_services.Embeddings.objects") def test_build_query_unauthenticated_uses_superuser_only_filter(mock_objects): # An unauthenticated user may only see files uploaded by superusers. - # The OR branch for the user's own files must NOT be present. + # The source uses a plain kwarg here (not a positional Q object), so the + # value lives in call_args.kwargs, not call_args.args. user = MagicMock(is_authenticated=False) build_query(user, EMBEDDING_VECTOR) - expected_q = Q(upload_file__uploaded_by__is_superuser=True) - actual_q = mock_objects.filter.call_args.args[0] - assert actual_q == expected_q + assert mock_objects.filter.call_args.kwargs == {"upload_file__uploaded_by__is_superuser": True} # Test application of annotate and order_by @@ -86,16 +85,18 @@ def test_build_query_no_document_filter_when_both_none(mock_objects): @patch("api.services.embedding_services.Embeddings.objects") def test_build_query_guid_takes_precedence_over_document_name(mock_objects): # When both guid and document_name are provided, the guid branch runs and - # the document_name branch is skipped entirely (only two filter calls total). + # the document_name branch is skipped entirely. user = MagicMock(is_authenticated=True) build_query(user, EMBEDDING_VECTOR, guid="abc-123", document_name="study.pdf") - # Two calls: auth filter + guid filter. No third call for document_name. - assert mock_objects.filter.call_count == 2 + # The auth filter fires on mock_objects.filter (call_count == 1). + # The document filter fires on the chained ordered_qs.filter — a different + # mock object — so mock_objects.filter.call_count stays at 1. + assert mock_objects.filter.call_count == 1 - # The second filter must use upload_file__guid, not name. - # We follow the mock chain to the queryset that .annotate().order_by() returned. + # The document filter must use upload_file__guid, not name, and must be + # called exactly once (confirming document_name branch was skipped). ordered_qs = mock_objects.filter.return_value.annotate.return_value.order_by.return_value ordered_qs.filter.assert_called_once_with(upload_file__guid="abc-123") From f9e890a21a3c716fe2bcfb17b8dfef92fcebb905 Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Mon, 23 Mar 2026 14:43:39 -0400 Subject: [PATCH 36/46] Guard TransformerModel preload to runserver processes only --- server/api/apps.py | 23 +++++++++++++++++++++++ 1 file changed, 23 insertions(+) diff --git a/server/api/apps.py b/server/api/apps.py index 4d502cba..d8b9eaa7 100644 --- a/server/api/apps.py +++ b/server/api/apps.py @@ -6,5 +6,28 @@ class ApiConfig(AppConfig): name = 'api' def ready(self): + import os + import sys + + # ready() runs in every Django process: migrate, test, shell, runserver, etc. + # Only preload the model when we're actually going to serve requests. + # Dev (docker-compose.yml) runs `manage.py runserver 0.0.0.0:8000`. + # Prod (Dockerfile.prod CMD) runs `manage.py runserver 0.0.0.0:8000 --noreload`. + # entrypoint.prod.sh also runs migrate, createsu, and populatedb before exec'ing + # runserver — the guard below correctly skips model loading for those commands too. + if sys.argv[1:2] != ['runserver']: + return + + # Dev's autoreloader spawns two processes: a parent file-watcher and a child + # server. ready() runs in both, but only the child (RUN_MAIN=true) serves + # requests. Skip the parent to avoid loading the model twice on each file change. + # Prod uses --noreload so RUN_MAIN is never set; 'noreload' in sys.argv handles that case. + if os.environ.get('RUN_MAIN') != 'true' and '--noreload' not in sys.argv: + return + + # Note: paraphrase-MiniLM-L6-v2 (~80MB) is downloaded from HuggingFace on first + # use and cached to ~/.cache/torch/sentence_transformers/ inside the container. + # That cache is ephemeral — every container rebuild re-downloads the model unless + # a volume is mounted at that path. from .services.sentencetTransformer_model import TransformerModel TransformerModel.get_instance() From 67176a8541be4d2862d213783873e1651f6fe761 Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Wed, 25 Mar 2026 16:42:20 -0400 Subject: [PATCH 37/46] Revert GitHub Workflow changes --- .github/workflows/python-app.yml | 9 ++------- evaluation/evals.py | 16 ++++++++-------- server/balancer_backend/urls.py | 6 +++--- 3 files changed, 13 insertions(+), 18 deletions(-) diff --git a/.github/workflows/python-app.yml b/.github/workflows/python-app.yml index 2afa2828..dcb7a4bb 100644 --- a/.github/workflows/python-app.yml +++ b/.github/workflows/python-app.yml @@ -5,9 +5,9 @@ name: Python application on: push: - branches: [ "develop" ] + branches: [ "listOfMed" ] pull_request: - branches: [ "develop" ] + branches: [ "listOfMed" ] permissions: contents: read @@ -27,8 +27,3 @@ jobs: run: pipx install ruff - name: Lint code with Ruff run: ruff check --output-format=github --target-version=py39 - - name: Install test dependencies - run: pip install -r server/requirements.txt - # Pytest won’t automatically discover config files in subdirectories - - name: Run tests - run: pytest -c server/pytest.ini server/ -v diff --git a/evaluation/evals.py b/evaluation/evals.py index 5110076f..8eb7e9e6 100755 --- a/evaluation/evals.py +++ b/evaluation/evals.py @@ -21,18 +21,18 @@ # Ensure the parent directory is in the path to import ModelFactory sys.path.insert(0, os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))) -import argparse # noqa: E402 -import logging # noqa: E402 -import asyncio # noqa: E402 -import time # noqa: E402 +import argparse +import logging +import asyncio +import time -import pandas as pd # noqa: E402 +import pandas as pd # lighteval depends on `sentencepiece` and it only has prebuilt wheels for Python 3.11 or below -from lighteval.tasks.requests import Doc # noqa: E402 -from lighteval.metrics.metrics_sample import Extractiveness # noqa: E402 +from lighteval.tasks.requests import Doc +from lighteval.metrics.metrics_sample import Extractiveness -from server.api.services.llm_services import ModelFactory # noqa: E402 +from server.api.services.llm_services import ModelFactory logging.basicConfig( level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s" diff --git a/server/balancer_backend/urls.py b/server/balancer_backend/urls.py index cdb92dbb..55bd2032 100644 --- a/server/balancer_backend/urls.py +++ b/server/balancer_backend/urls.py @@ -58,9 +58,9 @@ path("api/redoc/", SpectacularRedocView.as_view(url_name="schema"), name="redoc"), ] -import os # noqa: E402 -from django.conf import settings # noqa: E402 -from django.http import HttpResponseNotFound # noqa: E402 +import os +from django.conf import settings +from django.http import HttpResponseNotFound def spa_fallback(request): From d27392190c45e465dc81bf363f7bb381ec9c7aad Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Thu, 26 Mar 2026 14:35:35 -0400 Subject: [PATCH 38/46] Add section header comments to all four test groups in test_embedding_services.py --- .../api/services/test_embedding_services.py | 28 +++++++++++++++++-- 1 file changed, 26 insertions(+), 2 deletions(-) diff --git a/server/api/services/test_embedding_services.py b/server/api/services/test_embedding_services.py index dcbb2fc7..d1095366 100644 --- a/server/api/services/test_embedding_services.py +++ b/server/api/services/test_embedding_services.py @@ -12,10 +12,14 @@ # --------------------------------------------------------------------------- # build_query tests +# +# build_query is responsible for access control, annotate/order, document filter +# and slicing and only constructs a lazy Django QuerySet without evaluating it +# +# We can test build_query by patching Embeddings.objects and inspecting which +# methods and arguments were called on Embeddings.objects # --------------------------------------------------------------------------- -# All assertions inspect which methods and arguments were called on Embeddings.objects - # Only forwarded to L2Distance EMBEDDING_VECTOR = [0.1, 0.2, 0.3] @@ -165,6 +169,13 @@ def test_build_query_returns_unevaluated_queryset(mock_objects): # --------------------------------------------------------------------------- # evaluate_query tests +# +# evaluate_query is responsible for iterating the queryset and mapping each +# Embeddings object's attributes to a result dict, including the rename +# page_num -> page_number and the None-safe file_id lookup +# +# We can test evaluate_query by passing plain MagicMock objects directly as +# the iterable and asserting on the shape and values of the returned list # --------------------------------------------------------------------------- def test_evaluate_query_empty_queryset(): @@ -214,6 +225,13 @@ def test_evaluate_query_none_upload_file(): # --------------------------------------------------------------------------- # log_usage tests +# +# log_usage is responsible for computing distance stats, storing the correct +# user (None for unauthenticated), handling empty results, and swallowing +# exceptions so search is never interrupted +# +# We can test log_usage by patching SemanticSearchUsage.objects.create and +# inspecting the keyword arguments it was called with # --------------------------------------------------------------------------- @patch("api.services.embedding_services.SemanticSearchUsage.objects.create") @@ -333,6 +351,12 @@ def test_log_usage_swallows_exceptions(mock_create): # --------------------------------------------------------------------------- # get_closest_embeddings tests +# +# get_closest_embeddings is responsible for wiring together encode, +# build_query, evaluate_query, and log_usage and returning the results +# +# We can test get_closest_embeddings by patching all four collaborators and +# asserting that each is called with the correct arguments in the correct order # --------------------------------------------------------------------------- @patch("api.services.embedding_services.log_usage") From 8198574dc63e04e03d0c7198dfa2186c458c206e Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Thu, 26 Mar 2026 15:24:12 -0400 Subject: [PATCH 39/46] Document why tests are split by responsibility --- server/api/services/test_embedding_services.py | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/server/api/services/test_embedding_services.py b/server/api/services/test_embedding_services.py index d1095366..b1e5d5a7 100644 --- a/server/api/services/test_embedding_services.py +++ b/server/api/services/test_embedding_services.py @@ -10,6 +10,11 @@ log_usage, ) +# Each function is tested one responsibility at a time. One test for the whole +# function collapses all responsibilities into a single assertion block — when +# it fails you know something is broken but not which responsibility. You have +# to debug to find out. + # --------------------------------------------------------------------------- # build_query tests # From 5d8c8b37403ccc98e0b02dc0e491f0f30a61bd9c Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Tue, 31 Mar 2026 11:00:08 -0400 Subject: [PATCH 40/46] Improve logging and comments --- server/api/services/embedding_services.py | 6 +-- .../api/services/test_embedding_services.py | 2 +- server/balancer_backend/settings.py | 48 +++++++++---------- 3 files changed, 28 insertions(+), 28 deletions(-) diff --git a/server/api/services/embedding_services.py b/server/api/services/embedding_services.py index dada28a2..213519e5 100644 --- a/server/api/services/embedding_services.py +++ b/server/api/services/embedding_services.py @@ -2,7 +2,7 @@ import logging from statistics import median -# Django filter() only does ADD logic +# Use Q objects to express OR conditions in Django queries from django.db.models import Q from pgvector.django import L2Distance @@ -146,8 +146,8 @@ def log_usage( median_distance=None, min_distance=None, ) - except Exception as e: - logger.error(f"Failed to create semantic search usage database record: {e}") + except Exception: + logger.exception("Failed to create semantic search usage database record") def get_closest_embeddings( diff --git a/server/api/services/test_embedding_services.py b/server/api/services/test_embedding_services.py index b1e5d5a7..e43c0d74 100644 --- a/server/api/services/test_embedding_services.py +++ b/server/api/services/test_embedding_services.py @@ -26,7 +26,7 @@ # --------------------------------------------------------------------------- # Only forwarded to L2Distance -EMBEDDING_VECTOR = [0.1, 0.2, 0.3] +EMBEDDING_VECTOR = [0.1, 0.2, 0.3] # Test authenticated/unauthenticated user access control diff --git a/server/balancer_backend/settings.py b/server/balancer_backend/settings.py index a4ccaaae..c56a3435 100644 --- a/server/balancer_backend/settings.py +++ b/server/balancer_backend/settings.py @@ -254,27 +254,27 @@ # Logging configuration -# LOGGING = { -# "version": 1, -# "disable_existing_loggers": False, -# "formatters": { -# "verbose": { -# "format": "{levelname} {asctime} {module} {process:d} {thread:d} {message}", -# "style": "{", -# }, -# "simple": { -# "format": "{levelname} {message}", -# "style": "{", -# }, -# }, -# "handlers": { -# "console": { -# "class": "logging.StreamHandler", -# "formatter": "verbose", -# }, -# }, -# "root": { -# "handlers": ["console"], -# "level": "INFO", -# }, -# } +LOGGING = { + "version": 1, + "disable_existing_loggers": False, + "formatters": { + "verbose": { + "format": "{levelname} {asctime} {module} {process:d} {thread:d} {message}", + "style": "{", + }, + "simple": { + "format": "{levelname} {message}", + "style": "{", + }, + }, + "handlers": { + "console": { + "class": "logging.StreamHandler", + "formatter": "verbose", + }, + }, + "root": { + "handlers": ["console"], + "level": "INFO", + }, +} From 31498dcb62874899302acc72bfc9031d06800f05 Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Tue, 31 Mar 2026 11:06:03 -0400 Subject: [PATCH 41/46] Fall back to lazy load using try except block --- server/api/apps.py | 54 +++++++++++++++++++++++++--------------------- 1 file changed, 29 insertions(+), 25 deletions(-) diff --git a/server/api/apps.py b/server/api/apps.py index d8b9eaa7..c7eccc53 100644 --- a/server/api/apps.py +++ b/server/api/apps.py @@ -6,28 +6,32 @@ class ApiConfig(AppConfig): name = 'api' def ready(self): - import os - import sys - - # ready() runs in every Django process: migrate, test, shell, runserver, etc. - # Only preload the model when we're actually going to serve requests. - # Dev (docker-compose.yml) runs `manage.py runserver 0.0.0.0:8000`. - # Prod (Dockerfile.prod CMD) runs `manage.py runserver 0.0.0.0:8000 --noreload`. - # entrypoint.prod.sh also runs migrate, createsu, and populatedb before exec'ing - # runserver — the guard below correctly skips model loading for those commands too. - if sys.argv[1:2] != ['runserver']: - return - - # Dev's autoreloader spawns two processes: a parent file-watcher and a child - # server. ready() runs in both, but only the child (RUN_MAIN=true) serves - # requests. Skip the parent to avoid loading the model twice on each file change. - # Prod uses --noreload so RUN_MAIN is never set; 'noreload' in sys.argv handles that case. - if os.environ.get('RUN_MAIN') != 'true' and '--noreload' not in sys.argv: - return - - # Note: paraphrase-MiniLM-L6-v2 (~80MB) is downloaded from HuggingFace on first - # use and cached to ~/.cache/torch/sentence_transformers/ inside the container. - # That cache is ephemeral — every container rebuild re-downloads the model unless - # a volume is mounted at that path. - from .services.sentencetTransformer_model import TransformerModel - TransformerModel.get_instance() + + try: + import os + import sys + + # ready() runs in every Django process: migrate, test, shell, runserver, etc. + # Only preload the model when we're actually going to serve requests. + # Dev (docker-compose.yml) runs `manage.py runserver 0.0.0.0:8000`. + # Prod (Dockerfile.prod CMD) runs `manage.py runserver 0.0.0.0:8000 --noreload`. + # entrypoint.prod.sh also runs migrate, createsu, and populatedb before exec'ing + # runserver — the guard below correctly skips model loading for those commands too. + if sys.argv[1:2] != ['runserver']: + return + + # Dev's autoreloader spawns two processes: a parent file-watcher and a child + # server. ready() runs in both, but only the child (RUN_MAIN=true) serves + # requests. Skip the parent to avoid loading the model twice on each file change. + # Prod uses --noreload so RUN_MAIN is never set; 'noreload' in sys.argv handles that case. + if os.environ.get('RUN_MAIN') != 'true' and '--noreload' not in sys.argv: + return + + # Note: paraphrase-MiniLM-L6-v2 (~80MB) is downloaded from HuggingFace on first + # use and cached to ~/.cache/torch/sentence_transformers/ inside the container. + # That cache is ephemeral — every container rebuild re-downloads the model unless + # a volume is mounted at that path. + from .services.sentencetTransformer_model import TransformerModel + TransformerModel.get_instance() + except Exception: + logger.exception("Failed to preload the embedding model at startup") From a39d33c75564b564b3740134223eaa09333331b6 Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Tue, 31 Mar 2026 11:17:04 -0400 Subject: [PATCH 42/46] Revert settings.py to develop state --- server/balancer_backend/settings.py | 48 ++++++++++++++--------------- 1 file changed, 24 insertions(+), 24 deletions(-) diff --git a/server/balancer_backend/settings.py b/server/balancer_backend/settings.py index c56a3435..a4ccaaae 100644 --- a/server/balancer_backend/settings.py +++ b/server/balancer_backend/settings.py @@ -254,27 +254,27 @@ # Logging configuration -LOGGING = { - "version": 1, - "disable_existing_loggers": False, - "formatters": { - "verbose": { - "format": "{levelname} {asctime} {module} {process:d} {thread:d} {message}", - "style": "{", - }, - "simple": { - "format": "{levelname} {message}", - "style": "{", - }, - }, - "handlers": { - "console": { - "class": "logging.StreamHandler", - "formatter": "verbose", - }, - }, - "root": { - "handlers": ["console"], - "level": "INFO", - }, -} +# LOGGING = { +# "version": 1, +# "disable_existing_loggers": False, +# "formatters": { +# "verbose": { +# "format": "{levelname} {asctime} {module} {process:d} {thread:d} {message}", +# "style": "{", +# }, +# "simple": { +# "format": "{levelname} {message}", +# "style": "{", +# }, +# }, +# "handlers": { +# "console": { +# "class": "logging.StreamHandler", +# "formatter": "verbose", +# }, +# }, +# "root": { +# "handlers": ["console"], +# "level": "INFO", +# }, +# } From fe1eecaf133c8e82f5df765fd2161874a0def2a7 Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Tue, 31 Mar 2026 12:27:29 -0400 Subject: [PATCH 43/46] Manually test fall back to lazy loading --- server/api/apps.py | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/server/api/apps.py b/server/api/apps.py index c7eccc53..13977850 100644 --- a/server/api/apps.py +++ b/server/api/apps.py @@ -7,7 +7,7 @@ class ApiConfig(AppConfig): def ready(self): - try: + try: import os import sys @@ -34,4 +34,8 @@ def ready(self): from .services.sentencetTransformer_model import TransformerModel TransformerModel.get_instance() except Exception: + # TransformerModel._instance stays None on failure, so the first actual request + # that calls get_instance() will attempt to load the model again. + import logging + logger = logging.getLogger(__name__) logger.exception("Failed to preload the embedding model at startup") From 88adfb1cac637a1f2913857918e64924f44c23ee Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Tue, 14 Apr 2026 14:33:16 -0400 Subject: [PATCH 44/46] Add issue template --- .github/ISSUE_TEMPLATE/issue.md | 23 +++++++++++++++++++++++ 1 file changed, 23 insertions(+) create mode 100644 .github/ISSUE_TEMPLATE/issue.md diff --git a/.github/ISSUE_TEMPLATE/issue.md b/.github/ISSUE_TEMPLATE/issue.md new file mode 100644 index 00000000..70fca447 --- /dev/null +++ b/.github/ISSUE_TEMPLATE/issue.md @@ -0,0 +1,23 @@ +## Title + + +## Background + + +## Existing Behavior + + +## Acceptance Criteria +- [] + +## Approach + + +## References + + +## Risks and Rollback + + +## Screenshots / Recordings + \ No newline at end of file From 55acb2ead6cd75b407501b43742645379e93da97 Mon Sep 17 00:00:00 2001 From: Sahil D Shah Date: Tue, 14 Apr 2026 15:01:34 -0400 Subject: [PATCH 45/46] Update README --- README.md | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index 0d7d531a..34947225 100644 --- a/README.md +++ b/README.md @@ -6,6 +6,7 @@ for patients with bipolar disorder, helping them shorten their journey to stabil ## Usage You can view the current build of the website here: [https://balancerproject.org/](https://balancerproject.org/) +You can view the website in a sandbox here: [https://sandbox.balancerproject.org/](https://sandbox.balancerproject.org/) ## Contributing @@ -31,11 +32,9 @@ Get the code using git by either forking or cloning `CodeForPhilly/balancer-main ``` 2. (Optional) Add your API keys to `config/env/dev.env`: - `OpenAI API` - - `Anthropic API` Tools used for development: 1. `Docker`: Install Docker Desktop -2. `Postman`: Ask to get invited to the Balancer Postman team `balancer_dev` 3. `npm`: In the terminal run 1) 'cd frontend' 2) 'npm install' 3) 'cd ..' ### Running Balancer for development @@ -91,6 +90,14 @@ Most endpoints require JWT authentication. To test them in Swagger UI: 3. **Test endpoints**: All subsequent requests will include your token. Use **Try it out** on any protected endpoint. 4. **Token refresh**: Access tokens expire after 60 minutes. Use `POST /auth/jwt/refresh/` with your `refresh` token, or repeat step 1. +### Deployment + +1. Merging your PR into develop automatically triggers a GitHub Release +2. The release triggers a container build workflow that builds and pushes the Docker image +3. [Go to GitHub Packages](https://github.com/CodeForPhilly/balancer-main/pkgs/container/balancer-main%2Fapp) to find the new image tag +4. Update newTag in kustomization.yaml [in the cluster repo](https://github.com/CodeForPhilly/cfp-live-cluster/blob/main/balancer/kustomization.yaml) +5. Open a PR to [cfp-sandbox-cluster](https://github.com/CodeForPhilly/cfp-sandbox-cluster) (or [cfp-live-cluster](https://github.com/CodeForPhilly/cfp-live-cluster)) + ## Architecture The Balancer website is a Postgres, Django REST, and React project. The source code layout is: From cedc58ad2ba8008787ea1a902f63452febf284db Mon Sep 17 00:00:00 2001 From: Sahil Shah Date: Tue, 14 Apr 2026 15:04:38 -0400 Subject: [PATCH 46/46] Add sandbox link to README usage section --- README.md | 1 + 1 file changed, 1 insertion(+) diff --git a/README.md b/README.md index b0480844..fe765910 100644 --- a/README.md +++ b/README.md @@ -6,6 +6,7 @@ for patients with bipolar disorder, helping them shorten their journey to stabil ## Usage You can view the current build of the website here: [https://balancerproject.org/](https://balancerproject.org/) + You can view the website in a sandbox here: [https://sandbox.balancerproject.org/](https://sandbox.balancerproject.org/) ## Contributing