AutoGen integration
AutoGen prompt compression
AutoGen enables multi-agent conversations where agents chat with each other. The conversation history grows with every exchange. SuperCompress compresses the history before each agent responds.
AutoGen conversation bloat
After 10 exchanges between two AutoGen agents, the conversation history can exceed 5,000 tokens. Many of these are pleasantries, confirmations, and irrelevant tangents. Only the factual conclusions and task context matter for the next response.
Custom reply function
from autogen import AssistantAgent
from supercompress import Compressor
comp = Compressor()
class CompressedAssistantAgent(AssistantAgent):
def generate_reply(self, messages, sender=None, **kwargs):
# Compress conversation history
history = "\n".join(m["content"] for m in messages[:-1])
query = messages[-1]["content"] if messages else ""
if history and query:
result = comp.compress(history, query)
messages[-1]["content"] = result.compressed_text
return super().generate_reply(messages, sender, **kwargs)Frequently asked questions
Does compression work with AutoGen's group chat?
Yes. Each agent in the group chat gets compressed context before responding.
Will compression break AutoGen's code execution?
No. Only conversation history is compressed. Code execution results and tool outputs are preserved.
Try it yourself
Paste your long prompt into the playground, ask a question, and see what SuperCompress keeps and removes. Free, no signup needed.