Upload folder using huggingface_hub

.gitignore

@@ -0,0 +1,22 @@
+# Environment variables
+.env
+.env.*
+
+# Python
+__pycache__/
+*.pyc
+*.pyo
+*.pyd
+.Python
+env/
+venv/
+.venv/
+
+# Wrangler / Cloudflare
+.wrangler/
+
+# IDE / System
+.vscode/
+.idea/
+.DS_Store
+Thumbs.db

README.md

@@ -1,33 +1,32 @@
----
-title: Manifesto Explainer
-emoji: ⚡
-colorFrom: indigo
-colorTo: gray
-sdk: docker
-pinned: false
-app_port: 8080
-license: mit
----
-
-# Manifesto Explainer Backend
-
-This is the FastAPI backend for the Manifesto Explainer platform. 
-It provides endpoints for PDF parsing, LLM summarization, and political analysis.
-
-## Deployment on Hugging Face Spaces
-
-1. Create a new **Space** on Hugging Face.
-2. Select **Docker** as the SDK.
-3. Upload all files from this repository (or push via Git).
-4. Go to **Settings** > **Variables and secrets**.
-5. Add your `GROQ_API_KEY` and `GEMINI_API_KEY` as **Secrets**.
-6. The app will automatically build and run on port 8080.
-
-## Integration with Frontend
-
-If you are hosting the frontend on Cloudflare Pages:
-1. Copy the URL of your Hugging Face Space (e.g., `https://samarth-manifesto-explainer.hf.space`).
-2. In your `index.html` on Cloudflare, add this script tag before `script.js`:
-   ```html
-   <script>window.BACKEND_URL = "https://your-space-name.hf.space";</script>
+---
+title: Manifesto Explainer API
+emoji: 🗳️
+colorFrom: indigo
+colorTo: purple
+sdk: docker
+pinned: false
+app_port: 8080
+---
+
+# Manifesto Explainer Backend
+
+This is the FastAPI backend for the Manifesto Explainer platform. 
+It provides endpoints for PDF parsing, LLM summarization, and political analysis.
+
+## Deployment on Hugging Face Spaces
+
+1. Create a new **Space** on Hugging Face.
+2. Select **Docker** as the SDK.
+3. Upload all files from this repository (or push via Git).
+4. Go to **Settings** > **Variables and secrets**.
+5. Add your `GROQ_API_KEY` and `GEMINI_API_KEY` as **Secrets**.
+6. The app will automatically build and run on port 8080.
+
+## Integration with Frontend
+
+If you are hosting the frontend on Cloudflare Pages:
+1. Copy the URL of your Hugging Face Space (e.g., `https://samarth-manifesto-explainer.hf.space`).
+2. In your `index.html` on Cloudflare, add this script tag before `script.js`:
+   ```html
+   <script>window.BACKEND_URL = "https://your-space-name.hf.space";</script>
    ```

index.html

@@ -109,7 +109,7 @@
     </div>
 
     <!-- If deploying frontend separately (e.g. Cloudflare), set your backend URL here -->
-    <script>window.BACKEND_URL = "";</script>
+    <script>window.BACKEND_URL = "https://sa-m-manifesto-explainer-cc1d45d.hf.space";</script>
     <script src="script.js"></script>
 </body>
 </html>

main.py

@@ -83,31 +83,27 @@ class FailoverLLMClient:
         return None, ""
 
     def _get_groq_backup(self):
-        if self.groq_client: return self.groq_client, "gemma2-9b-it"
+        if self.groq_client: return self.groq_client, "llama-3.1-8b-instant"
         return None, ""
 
     def _call_gemini(self, messages: list, temperature: float, max_tokens: int):
         if not self.gemini_client: return None
-        try:
-            from google.genai import types
-            prompt = ""
-            system_instruction = "You are an expert political science analyst."
-            for msg in messages:
-                if msg["role"] == "system": system_instruction = msg["content"]
-                elif msg["role"] == "user": prompt = msg["content"]
-            response = self.gemini_client.models.generate_content(
-                model='gemini-2.5-flash',
-                contents=prompt,
-                config=types.GenerateContentConfig(
-                    system_instruction=system_instruction,
-                    temperature=temperature,
-                    max_output_tokens=max_tokens
-                )
+        from google.genai import types
+        prompt = ""
+        system_instruction = "You are an expert political science analyst."
+        for msg in messages:
+            if msg["role"] == "system": system_instruction = msg["content"]
+            elif msg["role"] == "user": prompt = msg["content"]
+        response = self.gemini_client.models.generate_content(
+            model='gemini-2.5-flash',
+            contents=prompt,
+            config=types.GenerateContentConfig(
+                system_instruction=system_instruction,
+                temperature=temperature,
+                max_output_tokens=max_tokens
             )
-            return response.text
-        except Exception as e:
-            print(f"Gemini API error: {e}")
-            return None
+        )
+        return response.text
 
     def _call_groq_with_retry(self, model, messages, temperature, max_tokens, retries=2):
         import time
@@ -118,7 +114,7 @@ class FailoverLLMClient:
                     "messages": messages,
                 }
                 if "gpt-oss" in model or "llama-4-scout" in model:
-                    params["max_completion_tokens"] = 16000
+                    params["max_completion_tokens"] = 8192
                     if "gpt-oss" in model:
                         params["reasoning_effort"] = "medium"
                     else:
@@ -252,10 +248,23 @@ def preprocess(textParty):
 def generate_summary(text):
     available, status = llm_client.is_available()
     if not available: return f"Summarization is not available. (Status: {status})"
-    if len(text) > 15000: text = text[:15000]
+    
+    # Smart sampling for large documents: Take chunks from start, middle, and end
+    # to ensure the summary covers the entire manifesto.
+    full_len = len(text)
+    if full_len > 18000:
+        chunk_size = 6000
+        start_chunk = text[:chunk_size]
+        mid_point = full_len // 2
+        mid_chunk = text[mid_point - (chunk_size // 2) : mid_point + (chunk_size // 2)]
+        end_chunk = text[-chunk_size:]
+        text_to_analyze = f"[BEGINNING]\n{start_chunk}\n\n[MIDDLE]\n{mid_chunk}\n\n[END]\n{end_chunk}"
+    else:
+        text_to_analyze = text
+
     messages = [
         {"role": "system", "content": "You are an expert political science analyst. Provide a comprehensive, objective, and structured summary of the political manifesto. Use Markdown with headings, bullet points, and bold text for key terms. Do not stop abruptly; ensure your response is complete."},
-        {"role": "user", "content": f"Summarize this political manifesto, focusing on main policy areas, promises, and core themes. Aim for 300-500 words.\n\nManifesto Text:\n{text}"}
+        {"role": "user", "content": f"Summarize this political manifesto, focusing on main policy areas, promises, and core themes. Aim for 300-500 words.\n\nManifesto Text:\n{text_to_analyze}"}
     ]
     content, provider = llm_client.chat_completion(messages=messages, temperature=0.3, max_tokens=2500)
     return content if content else f"Error generating summary: {provider}"

script.js

@@ -94,9 +94,20 @@ document.addEventListener('DOMContentLoaded', () => {
                 body: formData
             });
             
+            const contentType = response.headers.get("content-type");
             if (!response.ok) {
-                const errorData = await response.json();
-                throw new Error(errorData.detail || 'Analysis failed');
+                if (contentType && contentType.includes("application/json")) {
+                    const errorData = await response.json();
+                    throw new Error(errorData.detail || 'Analysis failed');
+                } else {
+                    const errorText = await response.text();
+                    console.error("Backend Error (Non-JSON):", errorText);
+                    throw new Error(`Server Error (${response.status}). The backend might still be starting up or is misconfigured.`);
+                }
+            }
+            
+            if (!contentType || !contentType.includes("application/json")) {
+                throw new Error("Invalid response from server. Expected JSON but received something else. Check if the Backend URL is correct.");
             }
             
             const data = await response.json();