fix: move large context to user message in compression example

- Only USER messages are compressed, not system messages
- Update example to put context in user message
- Add clarifying comment about compression behavior

Author: CLEMENTINATOR

example/compression.py

@@ -5,8 +5,9 @@
 2. Set a custom compression rate
 3. Access compression metrics from the response
 
-Note: Compression works on INPUT tokens, so this example includes a large
-context document to demonstrate meaningful compression savings.
+IMPORTANT: Only USER messages are compressed. System messages are not compressed.
+This example includes a large context in the user message to demonstrate meaningful
+compression savings.
 """
 
 import os
@@ -24,20 +25,20 @@
 LARGE_CONTEXT = """
 The History and Impact of Artificial Intelligence
 
-Artificial intelligence (AI) has evolved from a theoretical concept to a 
-transformative technology that influences nearly every aspect of modern life. 
-The field began in earnest in the 1950s when pioneers like Alan Turing and 
+Artificial intelligence (AI) has evolved from a theoretical concept to a
+transformative technology that influences nearly every aspect of modern life.
+The field began in earnest in the 1950s when pioneers like Alan Turing and
 John McCarthy laid the groundwork for machine intelligence.
 
-Early developments focused on symbolic reasoning and expert systems. These 
-rule-based approaches dominated the field through the 1970s and 1980s, with 
-systems like MYCIN demonstrating practical applications in medical diagnosis. 
-However, these early systems were limited by their inability to learn from data 
+Early developments focused on symbolic reasoning and expert systems. These
+rule-based approaches dominated the field through the 1970s and 1980s, with
+systems like MYCIN demonstrating practical applications in medical diagnosis.
+However, these early systems were limited by their inability to learn from data
 and adapt to new situations.
 
-The resurgence of neural networks in the 1980s and 1990s, particularly with 
-backpropagation algorithms, opened new possibilities. Yet it wasn't until the 
-2010s, with the advent of deep learning and the availability of massive datasets 
+The resurgence of neural networks in the 1980s and 1990s, particularly with
+backpropagation algorithms, opened new possibilities. Yet it wasn't until the
+2010s, with the advent of deep learning and the availability of massive datasets
 and computational power, that AI truly began to revolutionize industries.
 
 Modern AI applications span numerous domains:
@@ -48,21 +49,21 @@
 - Finance leverages AI for fraud detection, algorithmic trading, and risk assessment
 - Transportation is being transformed by autonomous vehicles and traffic optimization
 
-The development of large language models like GPT, BERT, and others has 
-particularly accelerated progress in natural language understanding and generation. 
-These models, trained on vast amounts of text data, can perform a wide range of 
+The development of large language models like GPT, BERT, and others has
+particularly accelerated progress in natural language understanding and generation.
+These models, trained on vast amounts of text data, can perform a wide range of
 language tasks with remarkable proficiency.
 
-Despite remarkable progress, significant challenges remain. Issues of bias, 
-interpretability, safety, and ethical considerations continue to be areas of 
-active research and debate. The AI community is working to ensure that these 
-powerful technologies are developed and deployed responsibly, with consideration 
+Despite remarkable progress, significant challenges remain. Issues of bias,
+interpretability, safety, and ethical considerations continue to be areas of
+active research and debate. The AI community is working to ensure that these
+powerful technologies are developed and deployed responsibly, with consideration
 for their societal impact.
 
-Looking forward, AI is expected to continue advancing rapidly, with potential 
-breakthroughs in areas like artificial general intelligence, quantum machine 
-learning, and brain-computer interfaces. The integration of AI into daily life 
-will likely deepen, raising important questions about human-AI collaboration, 
+Looking forward, AI is expected to continue advancing rapidly, with potential
+breakthroughs in areas like artificial general intelligence, quantum machine
+learning, and brain-computer interfaces. The integration of AI into daily life
+will likely deepen, raising important questions about human-AI collaboration,
 workforce transformation, and the future of human cognition itself.
 """
 
@@ -72,20 +73,24 @@
 print()
 
 # Example: Request with compression enabled and large input
-print("Example: Large context with compression enabled")
+print("Example: Large user message with compression enabled")
 print("-" * 70)
 print(f"Input context length: {len(LARGE_CONTEXT)} characters")
 print()
 
+# NOTE: Only USER messages are compressed
+# Put the large context in the user message to demonstrate compression
+user_message = f"""Here is some context about AI:
+
+{LARGE_CONTEXT}
+
+Based on this context, summarize the key milestones in AI development in 3 bullet points."""
+
 response = edgee.send(
     model="gpt-4o",
     input={
         "messages": [
-            {"role": "system", "content": LARGE_CONTEXT},
-            {
-                "role": "user",
-                "content": "Based on the context above, summarize the key milestones in AI development in 3 bullet points.",
-            },
+            {"role": "user", "content": user_message},
         ],
         "enable_compression": True,
         "compression_rate": 0.5,