Small changes+

main.py

@@ -1,14 +1,15 @@
+import os
+import json
 from fastapi import FastAPI, File, UploadFile
 from fastapi.responses import PlainTextResponse
-from graphvision.extractor import GraphExtractor
-from groq import Groq
-import os
 from fastapi.middleware.cors import CORSMiddleware
+from groq import Groq
 
+# Import your newly updated PyPI library!
+from graphvision import GraphExtractor
 
 app = FastAPI(title="STEM Sight Backend")
 
-
 app.add_middleware(
     CORSMiddleware,
     allow_origins=["*"], # Allows any browser extension to connect
@@ -17,7 +18,7 @@ app.add_middleware(
     allow_headers=["*"],
 )
 
-# Initialize the Groq Client (It automatically looks for the GROQ_API_KEY environment variable)
+# Initialize the Groq Client (Looks for the GROQ_API_KEY environment variable)
 groq_client = Groq()
 
 # Initialize your custom PyPI library
@@ -38,29 +39,46 @@ async def analyze_graph(file: UploadFile = File(...)):
 
         # 2. Extract structured data using your library
         print(f"Extracting data from {file.filename}...")
-        extraction_result = vision_engine.extract_data(temp_image_path, show=False)
+        
+        # 🚨 UPDATED: Call the new extract method
+        extraction_json_string = vision_engine.extract(temp_image_path)
         
         # Clean up the temporary file immediately
-        os.remove(temp_image_path)
+        if os.path.exists(temp_image_path):
+            os.remove(temp_image_path)
+
+        # 🚨 UPDATED: Parse the JSON string back into a Python dictionary
+        extraction_result = json.loads(extraction_json_string)
 
-        if extraction_result["status"] != "Success":
-            return "I'm sorry, I couldn't clearly identify the data in this graph."
+        # 🚨 UPDATED: Check for the new error format from your library
+        if "error" in extraction_result:
+            return f"I'm sorry, I couldn't clearly identify the data in this graph. Reason: {extraction_result['error']}"
 
         # 3. Format the JSON data into a prompt
-        graph_type = extraction_result["chart_type"]
-        graph_data = extraction_result["data"]
+        graph_type = extraction_result.get("chart_type", "unknown")
+        graph_data = extraction_result.get("data", [])
+        
+        # Grab optional labels/titles if they exist (good for context!)
+        x_label = extraction_result.get("x_axis_label", "Unknown X-Axis")
+        y_label = extraction_result.get("y_axis_label", "Unknown Y-Axis")
+        title = extraction_result.get("title", "Untitled Graph")
         
         prompt = f"""
         You are an accessibility assistant for visually impaired students. 
         I am giving you extracted data from a {graph_type} chart. 
+        Title: {title}
+        X-Axis Label: {x_label}
+        Y-Axis Label: {y_label}
+        
         Please summarize this data in one short, conversational, and easy-to-understand paragraph.
-        Do not use markdown, bold text, or asterisks. Write it exactly as it should be spoken out loud.
+        Point out the largest and smallest values if relevant.
+        Do not use markdown, bold text, or asterisks. Write it exactly as it should be spoken out loud by a text-to-speech engine.
         
         Data:
         {graph_data}
         """
 
-        # 4. Send to Groq for lightning-fast inference using Llama 3.1 (8 Billion parameters)
+        # 4. Send to Groq for lightning-fast inference
         print("Generating audio script with Groq Llama 3...")
         chat_completion = groq_client.chat.completions.create(
             messages=[
@@ -70,7 +88,7 @@ async def analyze_graph(file: UploadFile = File(...)):
                 }
             ],
             model="llama-3.1-8b-instant",
-            temperature=0.5, # Keep it relatively focused and factual
+            temperature=0.4, # Lowered slightly for more factual summaries
         )
 
         # 5. Return strictly the text response for the Chrome extension to speak