ppt / routers /powerpoint.py
kevinkal's picture
Update routers/powerpoint.py
857a80e verified
from fastapi import APIRouter, Depends, HTTPException, File, UploadFile, Form
from fastapi.responses import FileResponse
from pydantic import BaseModel
from typing import Annotated, List, Literal, Optional, Union, Dict
from google import genai
from google.genai import types
from auth import verify_token
import os
import tempfile
from pptx import Presentation
from pptx.util import Inches, Pt
from pptx.enum.text import PP_ALIGN, MSO_ANCHOR
from pptx.enum.shapes import MSO_CONNECTOR
from pptx.chart.data import CategoryChartData
from pptx.enum.chart import XL_CHART_TYPE
from pptx.dml.color import RGBColor
from enum import Enum
from datetime import datetime, timedelta
from collections import deque
from threading import Lock
import json
from abc import ABC, abstractmethod
router = APIRouter(prefix="/powerpoint", tags=["powerpoint"])
# Get Gemini client
gemini_key = os.environ.get('GEMINI_KEY', '')
client = genai.Client(api_key=gemini_key)
class APIResponse(BaseModel):
success: bool
data: Optional[Union[Dict, List[Dict], str]] = None
error: Optional[str] = None
class ChartDataPoint(BaseModel):
label: str
value: float
class MultimodalModel(str, Enum):
gemini_25_flash = "gemini-2.5-flash"
gemini_25_flash_lite = "gemini-2.5-flash-lite"
gemini_20_flash = "gemini-2.0-flash"
gemini_20_flash_lite = "gemini-2.0-flash-lite"
gemini_25_pro = "gemini-2.5-pro"
class TemplateType(str, Enum):
single_graph = "single_graph"
two_column_graph = "two_column_graph"
# Base class for all presentation data models
class BasePresentationData(BaseModel, ABC):
title: str
# Single graph template (original)
class SingleGraphPresentationData(BasePresentationData):
main_text: str
graph_title: str
chart_type: Literal["bar", "column", "line", "pie", "area"]
chart_data: List[ChartDataPoint]
side_text: str
# Two column graph template (new)
class TwoColumnGraphPresentationData(BasePresentationData):
left_text: str
left_graph_title: str
left_chart_type: Literal["bar", "column", "line", "pie", "area"]
left_chart_data: List[ChartDataPoint]
left_side_text: str
right_text: str
right_graph_title: str
right_chart_type: Literal["bar", "column", "line", "pie", "area"]
right_chart_data: List[ChartDataPoint]
right_side_text: str
MULTIMODAL_MODEL_LIMITS = {
"gemini-2.5-flash": {
"rpm": 10,
"daily": 250
},
"gemini-2.5-flash-lite": {
"rpm": 30,
"daily": 1000,
},
"gemini-2.0-flash": {
"rpm": 15,
"daily": 200
},
"gemini-2.0-flash-lite": {
"rpm": 30,
"daily": 200,
},
}
# Track usage per model
model_usage = {
model: {
"daily_count": 0,
"last_daily_reset": datetime.utcnow(),
"rpm_timestamps": deque(), # Tracks timestamps of requests in the past 60 seconds
}
for model in MULTIMODAL_MODEL_LIMITS
}
usage_lock = Lock() # Thread-safe tracking
def get_available_model():
now = datetime.utcnow()
with usage_lock:
for model, limits in MULTIMODAL_MODEL_LIMITS.items():
usage = model_usage[model]
# --- Daily Limit Reset ---
if now - usage["last_daily_reset"] > timedelta(days=1):
usage["daily_count"] = 0
usage["last_daily_reset"] = now
# --- RPM Cleanup ---
rpm_window = timedelta(seconds=60)
while usage["rpm_timestamps"] and now - usage["rpm_timestamps"][0] > rpm_window:
usage["rpm_timestamps"].popleft()
# --- Limit Checks ---
if usage["daily_count"] < limits["daily"] and len(usage["rpm_timestamps"]) < limits["rpm"]:
usage["daily_count"] += 1
usage["rpm_timestamps"].append(now)
return model
raise HTTPException(
status_code=429,
detail="All Gemini multimodal models have reached their rate limits (daily or RPM)."
)
def create_chart_from_data(slide, chart_data, chart_type, x, y, cx, cy):
"""Create a chart in the slide based on the provided data and type"""
# Chart type mapping
chart_type_mapping = {
"bar": XL_CHART_TYPE.BAR_CLUSTERED,
"column": XL_CHART_TYPE.COLUMN_CLUSTERED,
"line": XL_CHART_TYPE.LINE,
"pie": XL_CHART_TYPE.PIE,
"area": XL_CHART_TYPE.AREA
}
# Create chart data
chart_data_obj = CategoryChartData()
chart_data_obj.categories = [item.label for item in chart_data]
chart_data_obj.add_series('Series 1', [item.value for item in chart_data])
# Add chart to slide
chart = slide.shapes.add_chart(
chart_type_mapping.get(chart_type, XL_CHART_TYPE.COLUMN_CLUSTERED),
x, y, cx, cy, chart_data_obj
).chart
chart.chart_title.text_frame.text = ""
return chart
# Abstract base class for template generators
class TemplateGenerator(ABC):
@abstractmethod
def generate_pptx(self, presentation_data: BasePresentationData) -> str:
pass
@abstractmethod
def get_prompt(self) -> str:
pass
@abstractmethod
def get_response_schema(self):
pass
class SingleGraphTemplateGenerator(TemplateGenerator):
def get_prompt(self) -> str:
return """
Analyze this image and create a professional presentation slide based on what you see. Don't modify or invent text.
Please provide:
1. A title from the image
2. Secondary title from the image (main_text)
3. A graph title that describes any data visualization or key metric
4. Chart data with meaningful labels and values (create realistic data if needed based on the image)
5. Choose the most appropriate chart type (bar, column, line, pie, or area)
6. Side text with bullet points •, dash - and \\n
Make sure the content is professional, data-driven, and suitable for a business presentation.
The chart data should have 3-7 data points with meaningful labels and realistic values.
"""
def get_response_schema(self):
return SingleGraphPresentationData
def generate_pptx(self, presentation_data: SingleGraphPresentationData) -> str:
"""Generate PPTX file from single graph presentation data"""
prs = Presentation()
prs.slide_width = Inches(13.33) # standard 16:9
prs.slide_height = Inches(7.5)
slide = prs.slides.add_slide(prs.slide_layouts[6])
# Shape 1: Title
title_box = slide.shapes.add_textbox(Inches(0.39), Inches(0.24), Inches(12.53), Inches(0.9))
title_tf = title_box.text_frame
title_tf.word_wrap = True
title_tf.text = presentation_data.title
title_tf.paragraphs[0].font.size = Pt(24)
title_tf.paragraphs[0].font.name = "Verdana"
# Shape 2: Sources
src_box = slide.shapes.add_textbox(Inches(0.38), Inches(6.95), Inches(12.53), Inches(0.19))
src_tf = src_box.text_frame
src_tf.text = "Sources: TBD manually"
src_tf.paragraphs[0].font.size = Pt(10)
# Shape 4: Ligne horizontale sous le titre
line1 = slide.shapes.add_connector(
MSO_CONNECTOR.STRAIGHT,
Inches(0.39), Inches(1.23),
Inches(12.92), Inches(1.23)
)
line1.line.width = Pt(0.5)
line1.shadow.inherit = False
line1.line.fill.solid()
line1.line.fill.fore_color.rgb = RGBColor(1, 1, 1)
# Shape 7: [Text] gauche
text_box = slide.shapes.add_textbox(Inches(0.37), Inches(1.66), Inches(9.61), Inches(0.35))
text_tf = text_box.text_frame
text_tf.word_wrap = True
text_tf.text = presentation_data.main_text
text_tf.paragraphs[0].font.size = Pt(16)
text_tf.paragraphs[0].font.bold = True
text_tf.paragraphs[0].font.name = "Verdana"
# Shape 6: Ligne horizontale sous [Text]
line2 = slide.shapes.add_connector(
MSO_CONNECTOR.STRAIGHT,
Inches(0.37), Inches(2.09),
Inches(9.99), Inches(2.09)
)
line2.line.width = Pt(6)
line2.shadow.inherit = False
line2.line.fill.solid()
line2.line.fill.fore_color.rgb = RGBColor(0, 32, 96)
# Shape 5: Graph title
graph_title = slide.shapes.add_textbox(Inches(0.37), Inches(2.16), Inches(9.23), Inches(0.48))
gtf = graph_title.text_frame
gtf.word_wrap = True
gtf.text = presentation_data.graph_title
gtf.paragraphs[0].font.size = Pt(14)
gtf.paragraphs[0].font.name = "Verdana"
# Shape 10: Create actual chart
try:
create_chart_from_data(slide, presentation_data.chart_data,
presentation_data.chart_type,
Inches(0.39), Inches(2.96), Inches(9.23), Inches(3.64))
except Exception as e:
# Fallback to text box if chart creation fails
graph_box = slide.shapes.add_textbox(Inches(0.39), Inches(2.96), Inches(9.23), Inches(3.64))
graph_tf = graph_box.text_frame
graph_tf.text = f"Chart ({presentation_data.chart_type}): {presentation_data.graph_title}"
graph_tf.paragraphs[0].alignment = PP_ALIGN.CENTER
graph_tf.paragraphs[0].font.italic = True
# Shape 8: Ligne verticale séparatrice
line3 = slide.shapes.add_connector(
MSO_CONNECTOR.STRAIGHT,
Inches(10.0), Inches(2.23),
Inches(10.0), Inches(6.6)
)
line3.line.width = Pt(1.5)
line3.shadow.inherit = False
line3.line.fill.solid()
line3.line.fill.fore_color.rgb = RGBColor(0, 32, 96)
# Shape 11: Texte colonne droite
side_box = slide.shapes.add_textbox(Inches(10.19), Inches(2.2), Inches(2.72), Inches(4.39))
fill = side_box.fill
fill.solid()
fill.fore_color.rgb = RGBColor(247, 248, 250)
side_tf = side_box.text_frame
side_tf.word_wrap = True
side_tf.text = presentation_data.side_text
for paragraph in side_tf.paragraphs:
paragraph.font.size = Pt(12)
paragraph.font.name = "Verdana"
side_tf.vertical_anchor = MSO_ANCHOR.MIDDLE
# Save to temporary file
temp_file = tempfile.NamedTemporaryFile(delete=False, suffix='.pptx')
prs.save(temp_file.name)
return temp_file.name
class TwoColumnGraphTemplateGenerator(TemplateGenerator):
def get_prompt(self) -> str:
return """
Analyze this image and create a professional presentation slide with two columns based on what you see. Don't modify or invent text.
Please provide:
1. A title from the image
2. Left column text (main insight or finding)
3. Left graph title that describes the first data visualization
4. Left chart data with meaningful labels and values (3-5 data points)
5. Choose appropriate chart type for left chart (bar, column, line, pie, or area)
6. Left side text with bullet points •, dash - and \\n
7. Right column text (secondary insight or comparison)
8. Right graph title that describes the second data visualization
9. Right chart data with meaningful labels and values (3-5 data points)
10. Choose appropriate chart type for right chart (bar, column, line, pie, or area)
11. Right side text with bullet points •, dash - and \\n
Make sure the content is professional, data-driven, and suitable for a business presentation.
The two columns should complement each other and tell a coherent story.
"""
def get_response_schema(self):
return TwoColumnGraphPresentationData
def generate_pptx(self, presentation_data: TwoColumnGraphPresentationData) -> str:
"""Generate PPTX file from two column graph presentation data"""
prs = Presentation()
prs.slide_width = Inches(13.33) # 16:9 widescreen
prs.slide_height = Inches(7.5)
slide = prs.slides.add_slide(prs.slide_layouts[6]) # blank layout
# === Shape 1 : Title ===
title_box = slide.shapes.add_textbox(Inches(0.39), Inches(0.24), Inches(12.53), Inches(0.9))
title_tf = title_box.text_frame
title_tf.text = presentation_data.title
title_tf.word_wrap = True
title_tf.paragraphs[0].font.name = "Verdana"
title_tf.paragraphs[0].font.size = Pt(22)
title_tf.vertical_anchor = MSO_ANCHOR.MIDDLE
# === Shape 2 : Sources ===
src_box = slide.shapes.add_textbox(Inches(0.38), Inches(6.91), Inches(12.53), Inches(0.23))
src_tf = src_box.text_frame
src_tf.text = "Sources: TBD manually"
src_tf.paragraphs[0].font.size = Pt(10)
src_tf.paragraphs[0].font.name = "Verdana"
# === Shape 3 : Placeholder (page num / legend) ===
ph_box = slide.shapes.add_textbox(Inches(10.15), Inches(7.04), Inches(3.0), Inches(0.4))
ph_tf = ph_box.text_frame
ph_tf.text = "1"
ph_tf.paragraphs[0].alignment = PP_ALIGN.RIGHT
ph_tf.paragraphs[0].font.size = Pt(10)
# Shape 4: Ligne horizontale sous le titre
line1 = slide.shapes.add_connector(
MSO_CONNECTOR.STRAIGHT,
Inches(0.39), Inches(1.23),
Inches(12.92), Inches(1.23)
)
line1.line.width = Pt(0.5)
line1.shadow.inherit = False
line1.line.fill.solid()
line1.line.fill.fore_color.rgb = RGBColor(1, 1, 1)
# === Shape 5 : Ligne horizontale gauche ===
line2 = slide.shapes.add_connector(
MSO_CONNECTOR.STRAIGHT,
Inches(0.37), Inches(2.09),
Inches(6.3), Inches(2.09)
)
line2.line.width = Pt(6)
line2.shadow.inherit = False
line2.line.fill.solid()
line2.line.fill.fore_color.rgb = RGBColor(0, 32, 96)
# === Shape 6 : Graph title gauche ===
gt_left = slide.shapes.add_textbox(Inches(0.38), Inches(2.21), Inches(5.9), Inches(0.39))
gtl_tf = gt_left.text_frame
gtl_tf.word_wrap = True
gtl_tf.text = presentation_data.left_graph_title
gtl_tf.paragraphs[0].font.size = Pt(12)
gtl_tf.paragraphs[0].font.name = "Verdana"
# === Shape 7 : Ligne horizontale droite ===
line3 = slide.shapes.add_connector(
MSO_CONNECTOR.STRAIGHT,
Inches(7.07), Inches(2.09),
Inches(12.97), Inches(2.09)
)
line3.line.width = Pt(6)
line3.shadow.inherit = False
line3.line.fill.solid()
line3.line.fill.fore_color.rgb = RGBColor(0, 32, 96)
# === Shape 8 : Graph title droite ===
gt_right = slide.shapes.add_textbox(Inches(7.07), Inches(2.21), Inches(5.9), Inches(0.39))
gtr_tf = gt_right.text_frame
gtr_tf.word_wrap = True
gtr_tf.text = presentation_data.right_graph_title
gtr_tf.paragraphs[0].font.size = Pt(12)
gtr_tf.paragraphs[0].font.name = "Verdana"
# === Shape 9 : Bloc texte gauche ===
text_left = slide.shapes.add_textbox(Inches(4.47), Inches(2.85), Inches(1.8), Inches(3.6))
tl_tf_fill = text_left.fill
tl_tf_fill.solid()
tl_tf_fill.fore_color.rgb = RGBColor(247, 248, 250)
tl_tf = text_left.text_frame
tl_tf.word_wrap = True
tl_tf.text = presentation_data.left_side_text
for paragraph in tl_tf.paragraphs:
paragraph.font.size = Pt(12)
paragraph.font.name = "Verdana"
tl_tf.vertical_anchor = MSO_ANCHOR.MIDDLE
# === Shape 10 : Bloc texte droite ===
text_right = slide.shapes.add_textbox(Inches(11.13), Inches(2.85), Inches(1.8), Inches(3.6))
tr_tf_fill = text_right.fill
tr_tf_fill.solid()
tr_tf_fill.fore_color.rgb = RGBColor(247, 248, 250)
tr_tf = text_right.text_frame
tr_tf.word_wrap = True
tr_tf.text = presentation_data.right_side_text
for paragraph in tr_tf.paragraphs:
paragraph.font.size = Pt(12)
paragraph.font.name = "Verdana"
tr_tf.vertical_anchor = MSO_ANCHOR.MIDDLE
# === Shape 11 : Ligne verticale séparation ===
line4 = slide.shapes.add_connector(
MSO_CONNECTOR.STRAIGHT,
Inches(6.69), Inches(2.23),
Inches(6.69), Inches(6.32)
)
line4.line.width = Pt(0.25)
line4.shadow.inherit = False
line4.line.fill.solid()
line4.line.fill.fore_color.rgb = RGBColor(217, 217, 217)
# === Shape 12 : Texte haut gauche ===
txt12 = slide.shapes.add_textbox(Inches(0.38), Inches(1.52), Inches(5.9), Inches(0.44))
txt12_tf = txt12.text_frame
txt12_tf.word_wrap = True
txt12_tf.text = presentation_data.left_text
txt12_tf.paragraphs[0].font.size = Pt(14)
txt12_tf.paragraphs[0].font.name = "Verdana"
txt12_tf.paragraphs[0].font.bold = True
txt12_tf.paragraphs[0].font.color.rgb = RGBColor(0, 32, 96)
# === Shape 13 : Texte haut droite ===
txt13 = slide.shapes.add_textbox(Inches(7.07), Inches(1.52), Inches(5.9), Inches(0.44))
txt13_tf = txt13.text_frame
txt13_tf.word_wrap = True
txt13_tf.text = presentation_data.right_text
txt13_tf.paragraphs[0].font.size = Pt(14)
txt13_tf.paragraphs[0].font.name = "Verdana"
txt13_tf.paragraphs[0].font.bold = True
txt13_tf.paragraphs[0].font.color.rgb = RGBColor(0, 32, 96)
# === Shape 14 : Graph gauche ===
try:
create_chart_from_data(slide, presentation_data.left_chart_data,
presentation_data.left_chart_type,
Inches(0.38), Inches(2.85), Inches(4.1), Inches(3.6))
except Exception as e:
graph_left = slide.shapes.add_shape(1, Inches(0.38), Inches(2.85), Inches(4.1), Inches(3.6))
gl_tf = graph_left.text_frame
gl_tf.text = f"Graph ({presentation_data.left_chart_type})"
gl_tf.paragraphs[0].alignment = PP_ALIGN.CENTER
gl_tf.paragraphs[0].font.italic = True
# === Shape 15 : Graph droite ===
try:
create_chart_from_data(slide, presentation_data.right_chart_data,
presentation_data.right_chart_type,
Inches(7.07), Inches(2.85), Inches(4.06), Inches(3.6))
except Exception as e:
graph_right = slide.shapes.add_shape(1, Inches(7.07), Inches(2.85), Inches(4.06), Inches(3.6))
gr_tf = graph_right.text_frame
gr_tf.text = f"Graph ({presentation_data.right_chart_type})"
gr_tf.paragraphs[0].alignment = PP_ALIGN.CENTER
gr_tf.paragraphs[0].font.italic = True
# Save to temporary file
temp_file = tempfile.NamedTemporaryFile(delete=False, suffix='.pptx')
prs.save(temp_file.name)
return temp_file.name
# Template registry
TEMPLATE_GENERATORS = {
TemplateType.single_graph: SingleGraphTemplateGenerator(),
TemplateType.two_column_graph: TwoColumnGraphTemplateGenerator()
}
@router.post("/image-to-pptx")
async def image_to_pptx(
token: Annotated[str, Depends(verify_token)],
file: UploadFile = File(..., description="Image file to convert to PowerPoint"),
template: TemplateType = Form(..., description="Template type to use for the presentation"),
model: MultimodalModel = Form(..., description="Gemini model (Gemini 2.5 Flash recommended"),
):
"""
Analyze an image with Gemini and generate a PowerPoint presentation based on the analysis using the selected template
"""
try:
# Get available model
#model = get_available_model()
# Get template generator
template_generator = TEMPLATE_GENERATORS.get(template)
if not template_generator:
raise HTTPException(status_code=400, detail=f"Unsupported template: {template}")
# Read uploaded file
file_content = await file.read()
# Send to Gemini
response = client.models.generate_content(
model=model,
contents=[
types.Part.from_bytes(
data=file_content,
mime_type=file.content_type,
),
template_generator.get_prompt()
],
config={
'response_mime_type': 'application/json',
'response_schema': template_generator.get_response_schema(),
}
)
# Display input/output tokens
print(response.usage_metadata)
# Parse the response
presentation_data = template_generator.get_response_schema().model_validate(json.loads(response.text))
# Generate PPTX file
pptx_path = template_generator.generate_pptx(presentation_data)
# Return file as download
return FileResponse(
path=pptx_path,
filename=f"presentation_{template.value}_{datetime.now().strftime('%Y%m%d_%H%M%S')}.pptx",
media_type="application/vnd.openxmlformats-officedocument.presentationml.presentation"
)
except HTTPException:
raise
except Exception as e:
return APIResponse(success=False, error=f"Failed to generate presentation: {str(e)}")
@router.get("/templates")
async def get_available_templates(token: Annotated[str, Depends(verify_token)]):
"""Get list of available presentation templates"""
templates = []
for template_type in TemplateType:
templates.append({
"name": template_type.value,
"description": {
TemplateType.single_graph: "Single graph with sidebar - ideal for focusing on one key insight",
TemplateType.two_column_graph: "Two column layout with graphs and sidebars - perfect for comparisons"
}.get(template_type, "")
})
return APIResponse(success=True, data=templates)
@router.get("/model-stats")
async def get_model_stats(token: Annotated[str, Depends(verify_token)]):
"""Get current model usage statistics"""
now = datetime.utcnow()
model_stats = []
with usage_lock:
for model, limits in MULTIMODAL_MODEL_LIMITS.items():
usage = model_usage[model]
# Reset daily count if 24h passed
if now - usage["last_daily_reset"] > timedelta(days=1):
usage["daily_count"] = 0
usage["last_daily_reset"] = now
# Clean up old RPM timestamps
rpm_window = timedelta(seconds=60)
while usage["rpm_timestamps"] and now - usage["rpm_timestamps"][0] > rpm_window:
usage["rpm_timestamps"].popleft()
model_stats.append({
"model": model,
"daily_count": usage["daily_count"],
"daily_limit": limits["daily"],
"rpm_count": len(usage["rpm_timestamps"]),
"rpm_limit": limits["rpm"],
"last_daily_reset": usage["last_daily_reset"].isoformat(),
})
return APIResponse(success=True, data=model_stats)