improve url grounding by including source urls in context and citation instructions

academic_templates.json

@@ -49,7 +49,7 @@
     "name": "Course Assistant Example",
     "description": "Python support for cultural analytics students",
     "theme": "Monochrome",
-    "system_prompt": "You're a Python guide for CCNY's CSC 10800 where September covers foundations (command line, Jupyter, script anatomy), October builds programming basics (data types through functions) with Activities 1-2, and November-December advances to pandas, network analysis, and data collection with Activities 3-5, culminating in a Social Coding Portfolio. Support diverse learners by first assessing their comfort level and adapt your explanations accordingly. Always provide multiple entry points to concepts: start with the simplest working example that accomplishes the goal, then show incremental improvements and allow students to work and learn at their comfort level while, giving advanced learners paths to explore new concept and expand their programming repertoire.",
+    "system_prompt": "You're a Python guide for CCNY's CSC 10800 where September covers foundations (command line, Jupyter, script anatomy), October builds programming basics (data types through functions) with Activities 1-2, and November-December advances to pandas, network analysis, and data collection with Activities 3-5, culminating in a Social Coding Portfolio. Support diverse learners by first assessing their comfort level and adapt your explanations accordingly. Always provide multiple entry points to concepts: start with the simplest working example that accomplishes the goal, then show incremental improvements and allow students to work and learn at their comfort level while, giving advanced learners paths to explore new concept and expand their programming repertoire. Expect to complete all responses in under 750 tokens.",
     "examples": [
       "How do I set up a interactive development environment?",
       "Where can I find the course schedule and syllabus?",

app.py

@@ -551,7 +551,7 @@ class SpaceGenerator:
                                         try:
                                             content = fetch_url_content(url, max_chars=8000)
                                             if not content.startswith("❌") and not content.startswith("⏱️"):
-                                                grounding_context += f"\n**Primary Source {i}:** {content}\n"
+                                                grounding_context += f"\n**Primary Source {i} - {url}:**\n{content}\n"
                                         except:
                                             pass
                                 
@@ -563,7 +563,7 @@ class SpaceGenerator:
                                         try:
                                             content = fetch_url_content(url, max_chars=2500)
                                             if not content.startswith("❌") and not content.startswith("⏱️"):
-                                                grounding_context += f"\n**Secondary Source {i}:** {content}\n"
+                                                grounding_context += f"\n**Secondary Source {i} - {url}:**\n{content}\n"
                                         except:
                                             pass
                             
@@ -576,6 +576,7 @@ class SpaceGenerator:
                                 system_content += f"\n\nIMPORTANT: You must respond EXCLUSIVELY in {language}. All your responses should be written entirely in {language}, even when user input is in a different language, particularly English."
                             
                             if grounding_context:
+                                system_content += "\n\nIMPORTANT: When providing information from the reference sources below, please cite the specific URL(s) where the information can be found."
                                 system_content = f"{system_content}\n\n{grounding_context}"
                             
                             messages = [{"role": "system", "content": system_content}]

space_template.py

@@ -213,13 +213,13 @@ def fetch_url_content(url: str, max_length: int = 3000) -> str:
             if len(text) > max_length:
                 text = text[:max_length] + "... [truncated]"
             
-            return f"📄 Content from {{url}}:\\n{{text}}\\n"
+            return f"📄 **Content from:** {{url}}\\n\\n{{text}}\\n"
         
         elif any(ct in content_type for ct in ['text/plain', 'application/json']):
             text = response.text
             if len(text) > max_length:
                 text = text[:max_length] + "... [truncated]"
-            return f"📄 Content from {{url}}:\\n{{text}}\\n"
+            return f"📄 **Content from:** {{url}}\\n\\n{{text}}\\n"
         
         else:
             return f"⚠️ Unsupported content type at {{url}}: {{content_type}}"
@@ -313,7 +313,7 @@ def get_grounding_context() -> str:
                 _url_content_cache[url] = content
             
             if not content.startswith("❌") and not content.startswith("⏱️"):
-                context_parts.append(f"\\n**Primary Source {{i}}:** {{content}}")
+                context_parts.append(f"\\n**Primary Source {{i}} - {{url}}:**\\n{{content}}")
     
     # Process secondary sources (URLs 3+ with 2500 char limit)
     secondary_urls = urls[2:]
@@ -327,7 +327,7 @@ def get_grounding_context() -> str:
                 _url_content_cache[url] = content
             
             if not content.startswith("❌") and not content.startswith("⏱️"):
-                context_parts.append(f"\\n**Secondary Source {{i}}:** {{content}}")
+                context_parts.append(f"\\n**Secondary Source {{i}} - {{url}}:**\\n{{content}}")
     
     if len(context_parts) > 0:
         return "\\n".join(context_parts)
@@ -428,6 +428,7 @@ Get your API key at: https://openrouter.ai/keys"""
         system_content += f"\\n\\nIMPORTANT: You must respond EXCLUSIVELY in {{LANGUAGE}}. All your responses should be written entirely in {{LANGUAGE}}, even when user input is in a different language, particularly English."
     
     if grounding_context:
+        system_content += "\\n\\nIMPORTANT: When providing information from the reference sources below, please cite the specific URL(s) where the information can be found."
         system_content = f"{{system_content}}\\n\\n{{grounding_context}}"
     if file_context:
         system_content = f"{{system_content}}\\n\\n{{file_context}}"