feat(fastapi): use OpenAI client

.dockerignore

@@ -6,4 +6,5 @@ out
 build
 .git
 .gitignore
-tmp
+tmp
+debug

.gitignore

@@ -7,4 +7,5 @@ __pycache__
 node_modules
 out
 user_data
-tmp
+tmp
+debug

Dockerfile.dev

@@ -24,9 +24,8 @@ RUN curl -fsSL https://ollama.com/install.sh | sh
 COPY servers/fastapi/requirements.txt ./
 RUN pip install -r requirements.txt
 
-
 # Install dependencies for Next.js
-WORKDIR /app/servers/nextjs
+WORKDIR /node_dependencies
 COPY servers/nextjs/package.json servers/nextjs/package-lock.json ./
 RUN npm install
 
@@ -40,4 +39,10 @@ COPY nginx.conf /etc/nginx/nginx.conf
 EXPOSE 80 3000 8000
 
 # Start the servers
-CMD ["/bin/bash", "-c", "ollama serve & service nginx start & service redis-server start && node /app/start.js"]
+CMD ["/bin/bash", "-c", "\
+  rm -rf /app/servers/nextjs/node_modules && \
+  ln -s /node_dependencies/node_modules /app/servers/nextjs/node_modules && \
+  ollama serve & \
+  service nginx start & \
+  service redis-server start && \
+  node /app/start.js"]

servers/fastapi/api/main.py

@@ -8,7 +8,8 @@ from contextlib import asynccontextmanager
 from api.routers.presentation.router import presentation_router
 from api.services.database import sql_engine
 from api.utils.supported_ollama_models import SUPPORTED_OLLAMA_MODELS
-from api.utils.utils import is_ollama_selected, update_env_with_user_config
+from api.utils.utils import update_env_with_user_config
+from api.utils.model_utils import is_ollama_selected
 
 can_change_keys = os.getenv("CAN_CHANGE_KEYS") != "false"
 

servers/fastapi/api/routers/presentation/handlers/edit.py

@@ -14,8 +14,8 @@ from api.utils.supported_ollama_models import SUPPORTED_OLLAMA_MODELS
 from api.utils.utils import (
     get_presentation_dir,
     get_presentation_images_dir,
-    is_ollama_selected,
 )
+from api.utils.model_utils import is_ollama_selected
 from image_processor.icons_vectorstore_utils import get_icons_vectorstore
 from image_processor.images_finder import generate_image
 from image_processor.icons_finder import get_icon

servers/fastapi/api/routers/presentation/handlers/generate_data.py

@@ -11,7 +11,7 @@ from api.routers.presentation.models import PresentationGenerateRequest
 from api.services.logging import LoggingService
 from api.sql_models import KeyValueSqlModel, PresentationSqlModel
 from api.services.database import get_sql_session
-from api.utils.utils import is_ollama_selected
+from api.utils.model_utils import is_ollama_selected
 from ppt_config_generator.models import PresentationMarkdownModel, SlideStructureModel
 from ppt_config_generator.structure_generator import generate_presentation_structure
 

servers/fastapi/api/routers/presentation/handlers/generate_presentation.py

@@ -17,7 +17,8 @@ from api.services.database import get_sql_session
 from api.services.instances import TEMP_FILE_SERVICE
 from api.services.logging import LoggingService
 from api.sql_models import PresentationSqlModel, SlideSqlModel
-from api.utils.utils import get_presentation_dir, is_ollama_selected
+from api.utils.utils import get_presentation_dir
+from api.utils.model_utils import is_ollama_selected
 from document_processor.loader import DocumentsLoader
 from ppt_config_generator.document_summary_generator import generate_document_summary
 from ppt_config_generator.models import PresentationMarkdownModel

servers/fastapi/api/routers/presentation/handlers/generate_stream.py

@@ -17,7 +17,8 @@ from api.routers.presentation.models import (
 from api.services.database import get_sql_session
 from api.services.logging import LoggingService
 from api.sql_models import KeyValueSqlModel, PresentationSqlModel, SlideSqlModel
-from api.utils.utils import get_presentation_dir, is_ollama_selected
+from api.utils.utils import get_presentation_dir
+from api.utils.model_utils import is_ollama_selected
 from ppt_config_generator.models import (
     PresentationMarkdownModel,
     PresentationStructureModel,

servers/fastapi/api/utils/model_utils.py

@@ -1,5 +1,7 @@
 import os
 
+from openai import AsyncOpenAI
+
 from api.models import SelectedLLMProvider
 
 
@@ -15,38 +17,56 @@ def get_model_base_url():
     selected_llm = get_selected_llm_provider()
 
     if selected_llm == SelectedLLMProvider.OLLAMA:
-        return "http://localhost:11434"
+        return "http://localhost:11434/v1"
     elif selected_llm == SelectedLLMProvider.OPENAI:
         return "https://api.openai.com/v1"
     else:
         return "https://generativelanguage.googleapis.com/v1beta/openai"
 
 
+def get_llm_api_key():
+    selected_llm = get_selected_llm_provider()
+    if selected_llm == SelectedLLMProvider.OPENAI:
+        return os.getenv("OPENAI_API_KEY")
+    elif selected_llm == SelectedLLMProvider.GOOGLE:
+        return os.getenv("GOOGLE_API_KEY")
+    else:
+        return "ollama"
+
+
+def get_llm_client():
+    client = AsyncOpenAI(
+        base_url=get_model_base_url(),
+        api_key=get_llm_api_key(),
+    )
+    return client
+
+
 def get_large_model():
     selected_llm = get_selected_llm_provider()
     if selected_llm == SelectedLLMProvider.OPENAI:
-        return ChatOpenAI(model="gpt-4.1")
+        return "gpt-4.1"
     elif selected_llm == SelectedLLMProvider.GOOGLE:
-        return ChatGoogleGenerativeAI(model="gemini-2.0-flash")
+        return "gemini-2.0-flash"
     else:
-        return ChatOllama(model=os.getenv("OLLAMA_MODEL"), temperature=0.8)
+        return os.getenv("OLLAMA_MODEL")
 
 
 def get_small_model():
-    selected_llm = os.getenv("LLM")
-    if selected_llm == "openai":
-        return ChatOpenAI(model="gpt-4.1-mini")
-    elif selected_llm == "google":
-        return ChatGoogleGenerativeAI(model="gemini-2.0-flash")
+    selected_llm = get_selected_llm_provider()
+    if selected_llm == SelectedLLMProvider.OPENAI:
+        return "gpt-4.1-mini"
+    elif selected_llm == SelectedLLMProvider.GOOGLE:
+        return "gemini-2.0-flash"
     else:
-        return ChatOllama(model=os.getenv("OLLAMA_MODEL"), temperature=0.8)
+        return os.getenv("OLLAMA_MODEL")
 
 
 def get_nano_model():
-    selected_llm = os.getenv("LLM")
-    if selected_llm == "openai":
-        return ChatOpenAI(model="gpt-4.1-nano")
-    elif selected_llm == "google":
-        return ChatGoogleGenerativeAI(model="gemini-2.0-flash")
+    selected_llm = get_selected_llm_provider()
+    if selected_llm == SelectedLLMProvider.OPENAI:
+        return "gpt-4.1-nano"
+    elif selected_llm == SelectedLLMProvider.GOOGLE:
+        return "gemini-2.0-flash"
     else:
-        return ChatOllama(model=os.getenv("OLLAMA_MODEL"), temperature=0.8)
+        return os.getenv("OLLAMA_MODEL")

servers/fastapi/image_processor/images_finder.py

@@ -9,7 +9,8 @@ from openai import OpenAI
 from ppt_generator.models.query_and_prompt_models import (
     ImagePromptWithThemeAndAspectRatio,
 )
-from api.utils.utils import download_file, get_resource, is_ollama_selected
+from api.utils.utils import download_file, get_resource
+from api.utils.model_utils import is_ollama_selected
 
 
 async def generate_image(

servers/fastapi/ppt_config_generator/document_summary_generator.py

@@ -1,14 +1,10 @@
 import asyncio
-import os
 from typing import List
 from langchain_core.documents import Document
-from langchain_google_genai import ChatGoogleGenerativeAI
-from langchain_openai import ChatOpenAI
-from langchain_core.prompts import ChatPromptTemplate
-from langchain_core.messages import BaseMessage
 from langchain_text_splitters import CharacterTextSplitter
+from openai.types.chat.chat_completion import ChatCompletion
 
-from api.utils.utils import get_nano_model
+from api.utils.model_utils import get_llm_client, get_nano_model
 
 sysmte_prompt = """
 Generate a blog-style summary of the provided document in **more than 2000 words**.
@@ -26,26 +22,28 @@ Maintain as much information as possible.
 - If **slides structure is mentioned** in document, structure the summary in the same way.
 """
 
-prompt_template = ChatPromptTemplate.from_messages(
-    [
-        ("system", sysmte_prompt),
-        ("user", "{text}"),
-    ]
-)
-
 
 async def generate_document_summary(documents: List[Document]):
+    client = get_llm_client()
     model = get_nano_model()
+
     text_splitter = CharacterTextSplitter(chunk_size=200000, chunk_overlap=0)
-    chain = prompt_template | model
 
     coroutines = []
     for document in documents:
         text = document.page_content
         truncated_text = text_splitter.split_text(text)[0]
-        coroutine = chain.ainvoke({"text": truncated_text})
+        coroutine = client.chat.completions.create(
+            model=model,
+            messages=[
+                {"role": "system", "content": sysmte_prompt},
+                {"role": "user", "content": truncated_text},
+            ],
+        )
         coroutines.append(coroutine)
 
-    completions: List[BaseMessage] = await asyncio.gather(*coroutines)
-    combined = "\n\n\n\n".join([completion.content for completion in completions])
+    completions: List[ChatCompletion] = await asyncio.gather(*coroutines)
+    combined = "\n\n\n\n".join(
+        [completion.choices[0].message.content for completion in completions]
+    )
     return combined

servers/fastapi/ppt_config_generator/ppt_outlines_generator.py

@@ -1,7 +1,8 @@
 from typing import Optional
 from langchain_core.prompts import ChatPromptTemplate
+from langchain_ollama import ChatOllama
 
-from api.utils.utils import get_large_model
+from api.utils.model_utils import get_large_model
 from api.utils.variable_length_models import (
     get_presentation_markdown_model_with_n_slides,
 )
@@ -64,7 +65,7 @@ async def generate_ppt_content(
     language: Optional[str] = None,
     content: Optional[str] = None,
 ) -> PresentationMarkdownModel:
-    model = get_large_model()
+    model = ChatOllama(model=get_large_model(), temperature=0.8)
     response_model = get_presentation_markdown_model_with_n_slides(n_slides)
 
     chain = get_prompt_template() | model.with_structured_output(

servers/fastapi/ppt_config_generator/structure_generator.py

@@ -1,6 +1,7 @@
 from langchain_core.prompts import ChatPromptTemplate
+from langchain_ollama import ChatOllama
 
-from api.utils.utils import get_small_model
+from api.utils.model_utils import get_small_model
 from api.utils.variable_length_models import (
     get_presentation_structure_model_with_n_slides,
 )
@@ -59,7 +60,7 @@ async def generate_presentation_structure(
     presentation_outline: PresentationMarkdownModel,
 ) -> PresentationStructureModel:
 
-    model = get_small_model()
+    model = ChatOllama(model=get_small_model(), temperature=0.8)
     response_model = get_presentation_structure_model_with_n_slides(
         len(presentation_outline.slides)
     )

servers/fastapi/ppt_generator/fix_validation_errors.py

@@ -1,12 +1,11 @@
 import os
 from typing import Optional
 from fastapi import HTTPException
-from langchain_google_genai import ChatGoogleGenerativeAI
-from langchain_openai import ChatOpenAI
+from langchain_ollama import ChatOllama
 from langchain_core.prompts import ChatPromptTemplate
 from pydantic import BaseModel, ValidationError
 
-from api.utils.utils import get_large_model
+from api.utils.model_utils import get_large_model
 
 
 def get_prompt_template():
@@ -41,7 +40,7 @@ def get_prompt_template():
 
 
 async def fix_validation_errors(response_model: BaseModel, response, errors):
-    model = get_large_model()
+    model = ChatOllama(model=get_large_model(), temperature=0.8)
 
     chain = get_prompt_template() | model.with_structured_output(
         response_model.model_json_schema()

servers/fastapi/ppt_generator/generator.py

@@ -5,7 +5,8 @@ from langchain_core.messages import (
     AIMessageChunk,
     AIMessage,
 )
-from api.utils.utils import get_large_model
+from langchain_ollama import ChatOllama
+from api.utils.model_utils import get_large_model
 from ppt_config_generator.models import PresentationMarkdownModel
 from ppt_generator.models.llm_models_with_validations import (
     LLMPresentationModelWithValidation,
@@ -91,7 +92,7 @@ def get_model_and_messages(
     presentation_outline: PresentationMarkdownModel,
 ):
     user_message = HumanMessage(presentation_outline.to_string())
-    model = get_large_model()
+    model = ChatOllama(model=get_large_model(), temperature=0.8)
 
     return model, system_prompt, user_message
 

servers/fastapi/ppt_generator/models/llm_models_with_validations.py

@@ -34,48 +34,34 @@ from ppt_generator.models.llm_models import (
 
 class LLMHeadingModelWithValidation(LLMHeadingModel):
     heading: str = Field(
-        description="List item heading to show in slide body",
-        min_length=10,
-        max_length=30,
+        description="List item heading to show in slide body in less than 5 words.",
     )
     description: str = Field(
         description="Description of list item in less than 20 words.",
-        min_length=80,
-        max_length=150,
     )
 
 
 class LLMHeadingModelWithImagePromptWithValidation(LLMHeadingModelWithImagePrompt):
     image_prompt: str = Field(
-        description="Prompt used to generate image for this item",
-        min_length=10,
-        max_length=50,
+        description="Prompt used to generate image for this item in less than 6 words.",
     )
 
 
 class LLMHeadingModelWithIconQueryWithValidation(LLMHeadingModelWithIconQuery):
     icon_query: str = Field(
-        description="Icon query to generate icon for this item",
-        min_length=10,
-        max_length=50,
+        description="Icon query to generate icon for this item in less than 4 words.",
     )
 
 
 class LLMType1ContentWithValidation(LLMType1Content):
     title: str = Field(
-        description="Title of the slide",
-        min_length=10,
-        max_length=50,
+        description="Title of the slide in less than 6 words.",
     )
     body: str = Field(
         description="Slide content summary in less than 30 words.",
-        min_length=100,
-        max_length=200,
     )
     image_prompt: str = Field(
-        description="Prompt used to generate image for this slide.",
-        min_length=10,
-        max_length=50,
+        description="Prompt used to generate image for this slide in less than 6 words.",
     )
 
     @classmethod
@@ -85,9 +71,7 @@ class LLMType1ContentWithValidation(LLMType1Content):
 
 class LLMType2ContentWithValidation(LLMType2Content):
     title: str = Field(
-        description="Title of the slide",
-        min_length=10,
-        max_length=50,
+        description="Title of the slide in less than 6 words.",
     )
     body: List[LLMHeadingModelWithValidation] = Field(
         description="List items to show in slide's body",
@@ -106,9 +90,7 @@ class LLMType2ContentWithValidation(LLMType2Content):
 
 class LLMType3ContentWithValidation(LLMType3Content):
     title: str = Field(
-        description="Title of the slide",
-        min_length=10,
-        max_length=50,
+        description="Title of the slide in less than 6 words.",
     )
     body: List[LLMHeadingModelWithValidation] = Field(
         description="List items to show in slide's body",
@@ -116,9 +98,7 @@ class LLMType3ContentWithValidation(LLMType3Content):
         max_length=3,
     )
     image_prompt: str = Field(
-        description="Prompt used to generate image for this slide",
-        min_length=10,
-        max_length=50,
+        description="Prompt used to generate image for this slide in less than 6 words.",
     )
 
     @classmethod
@@ -132,9 +112,7 @@ class LLMType3ContentWithValidation(LLMType3Content):
 
 class LLMType4ContentWithValidation(LLMType4Content):
     title: str = Field(
-        description="Title of the slide",
-        min_length=10,
-        max_length=50,
+        description="Title of the slide in less than 6 words.",
     )
     body: List[LLMHeadingModelWithImagePromptWithValidation] = Field(
         description="List items to show in slide's body",
@@ -153,14 +131,10 @@ class LLMType4ContentWithValidation(LLMType4Content):
 
 class LLMType5ContentWithValidation(LLMType5Content):
     title: str = Field(
-        description="Title of the slide",
-        min_length=10,
-        max_length=50,
+        description="Title of the slide in less than 6 words.",
     )
     body: str = Field(
         description="Slide content summary in less than 30 words.",
-        min_length=100,
-        max_length=250,
     )
     graph: GraphModel = Field(description="Graph to show in slide")
 
@@ -171,14 +145,10 @@ class LLMType5ContentWithValidation(LLMType5Content):
 
 class LLMType6ContentWithValidation(LLMType6Content):
     title: str = Field(
-        description="Title of the slide",
-        min_length=10,
-        max_length=50,
+        description="Title of the slide in less than 6 words.",
     )
     description: str = Field(
         description="Slide content summary in less than 20 words.",
-        min_length=80,
-        max_length=150,
     )
     body: List[LLMHeadingModelWithValidation] = Field(
         description="List items to show in slide's body",
@@ -197,9 +167,7 @@ class LLMType6ContentWithValidation(LLMType6Content):
 
 class LLMType7ContentWithValidation(LLMType7Content):
     title: str = Field(
-        description="Title of the slide",
-        min_length=10,
-        max_length=50,
+        description="Title of the slide in less than 6 words.",
     )
     body: List[LLMHeadingModelWithIconQueryWithValidation] = Field(
         description="List items to show in slide's body",
@@ -218,14 +186,10 @@ class LLMType7ContentWithValidation(LLMType7Content):
 
 class LLMType8ContentWithValidation(LLMType8Content):
     title: str = Field(
-        description="Title of the slide",
-        min_length=10,
-        max_length=50,
+        description="Title of the slide in less than 6 words.",
     )
     description: str = Field(
         description="Slide content summary in less than 20 words.",
-        min_length=80,
-        max_length=150,
     )
     body: List[LLMHeadingModelWithImagePromptWithValidation] = Field(
         description="List items to show in slide's body",
@@ -244,9 +208,7 @@ class LLMType8ContentWithValidation(LLMType8Content):
 
 class LLMType9ContentWithValidation(LLMType9Content):
     title: str = Field(
-        description="Title of the slide",
-        min_length=10,
-        max_length=50,
+        description="Title of the slide in less than 6 words.",
     )
     body: List[LLMHeadingModelWithValidation] = Field(
         description="List items to show in slide's body",

servers/fastapi/ppt_generator/slide_generator.py

@@ -1,5 +1,8 @@
 from typing import Optional
-from api.utils.utils import get_large_model, get_small_model
+
+from langchain_ollama import ChatOllama
+from openai import OpenAI
+from api.utils.model_utils import get_large_model, get_llm_client, get_small_model
 from ppt_config_generator.models import SlideMarkdownModel
 from ppt_generator.fix_validation_errors import get_validated_response
 
@@ -16,42 +19,43 @@ from ppt_generator.models.other_models import SlideTypeModel
 from ppt_generator.models.slide_model import SlideModel
 
 
-prompt_template_to_generate_slide_content = ChatPromptTemplate.from_messages(
-    [
-        (
-            "system",
-            """
-            Generate structured slide based on provided title and outline, follow mentioned steps and notes and provide structured output.
+def get_prompt_to_generate_slide_content(
+    title: str, outline: str, notes: Optional[str] = None
+):
+    return [
+        {
+            "role": "system",
+            "content": f"""
+                Generate structured slide based on provided title and outline, follow mentioned steps and notes and provide structured output.
 
 
-            # Steps
-            1. Analyze the outline and title.
-            2. Generate structured slide based on the outline and title.
-            3. Generate image prompts and icon queries if mentioned in schema.
-            4. Generate graph if mentioned in schema.
+                # Steps
+                1. Analyze the outline and title.
+                2. Generate structured slide based on the outline and title.
+                3. Generate image prompts and icon queries if mentioned in schema.
+                4. Generate graph if mentioned in schema.
 
-            # Notes
-            - Slide body should not use words like "This slide", "This presentation".
-            - Rephrase the slide body to make it flow naturally.
-            - Do not use markdown formatting in slide body.
-            - **Icon query** must be a generic single word noun.
-            - **Image prompt** should be a 2-3 words phrase.
-            - Try to make paragraphs as short as possible.
-            {notes}
+                # Notes
+                - Slide body should not use words like "This slide", "This presentation".
+                - Rephrase the slide body to make it flow naturally.
+                - Do not use markdown formatting in slide body.
+                - **Icon query** must be a generic single word noun.
+                - **Image prompt** should be a 2-3 words phrase.
+                - Try to make paragraphs as short as possible.
+                {notes}
             """,
-        ),
-        (
-            "user",
-            """
-            ## Slide Title
-            {title}
+        },
+        {
+            "role": "user",
+            "content": f"""
+                ## Slide Title
+                {title}
 
-            ## Slide Outline
-            {outline}
-        """,
-        ),
+                ## Slide Outline
+                {outline}
+            """,
+        },
     ]
-)
 
 
 prompt_template_to_edit_slide_content = ChatPromptTemplate.from_messages(
@@ -126,22 +130,26 @@ prompt_template_to_select_slide_type = ChatPromptTemplate.from_messages(
 async def get_slide_content_from_type_and_outline(
     slide_type: int, outline: SlideMarkdownModel
 ) -> LLMSlideContentModel:
-    content_type_model_type = LLM_CONTENT_TYPE_WITH_VALIDATION_MAPPING[slide_type]
-    validation_model = LLM_CONTENT_TYPE_MAPPING[slide_type]
-    model = get_small_model().with_structured_output(
-        content_type_model_type.model_json_schema()
-    )
-    chain = prompt_template_to_generate_slide_content | model
+    response_model = LLM_CONTENT_TYPE_WITH_VALIDATION_MAPPING[slide_type]
 
-    return await get_validated_response(
-        chain,
-        {
-            "title": outline.title,
-            "outline": outline.body,
-            "notes": content_type_model_type.get_notes(),
-        },
-        content_type_model_type,
-        validation_model,
+    client = get_llm_client()
+    model = get_small_model()
+
+    response = await client.beta.chat.completions.parse(
+        model=model,
+        messages=get_prompt_to_generate_slide_content(
+            outline.title,
+            outline.body,
+            response_model.get_notes(),
+        ),
+        response_format=response_model,
+    )
+
+    with open("debug/llm_response.json", "w") as f:
+        f.write(response.choices[0].message.content)
+
+    return LLM_CONTENT_TYPE_MAPPING[slide_type].model_validate_json(
+        response.choices[0].message.content
     )
 
 
@@ -152,7 +160,7 @@ async def get_edited_slide_content_model(
     theme: Optional[dict] = None,
     language: Optional[str] = None,
 ):
-    model = get_large_model()
+    model = ChatOllama(model=get_large_model(), temperature=0.8)
 
     content_type_model_type = LLM_CONTENT_TYPE_WITH_VALIDATION_MAPPING[slide_type]
     validation_model = LLM_CONTENT_TYPE_MAPPING[slide_type]
@@ -181,7 +189,7 @@ async def get_slide_type_from_prompt(
     slide: SlideModel,
 ) -> SlideTypeModel:
 
-    model = get_small_model()
+    model = ChatOllama(model=get_small_model(), temperature=0.8)
 
     chain = prompt_template_to_select_slide_type | model.with_structured_output(
         SlideTypeModel.model_json_schema()

servers/fastapi/requirements.txt

@@ -65,7 +65,7 @@ mypy_extensions==1.1.0
 numpy==2.2.5
 ollama==0.5.1
 onnxruntime==1.22.0
-openai==1.78.1
+openai==1.91.0
 orjson==3.10.18
 packaging==24.2
 pdfminer.six==20250327

servers/fastapi/server.py

@@ -1,3 +1,4 @@
+import os
 import uvicorn
 import argparse
 
@@ -8,6 +9,8 @@ from api.main import app
 app
 
 if __name__ == "__main__":
+    os.makedirs("debug", exist_ok=True)
+
     parser = argparse.ArgumentParser(description="Run the FastAPI server")
     parser.add_argument(
         "--port", type=int, required=True, help="Port number to run the server on"

servers/fastapi/server_autoreload.py

@@ -1,8 +1,11 @@
+import os
 import uvicorn
 import argparse
 
 
 if __name__ == "__main__":
+    os.makedirs("debug", exist_ok=True)
+
     parser = argparse.ArgumentParser(description="Run the FastAPI server")
     parser.add_argument(
         "--port", type=int, required=True, help="Port number to run the server on"

servers/fastapi/server_test.py

@@ -1,9 +1,12 @@
+import os
 import uvicorn
 from dotenv import load_dotenv
 
 load_dotenv()
 
 if __name__ == "__main__":
+    os.makedirs("debug", exist_ok=True)
+
     uvicorn.run(
         "api.main:app", host="0.0.0.0", port=8000, log_level="info", reload=True
     )