CompressibleAgent from PR #443 - Missing in Current AutoGen Version, Need Guidance on Token Management for Multi-Agent Workflows

[vanshika2906]

Hi AutoGen Community,

I'm working on a complex RCA (Root Cause Analysis) workflow with 11 agents that have verbose system messages and need intelligent token management. I discovered PR #443 which introduced CompressibleAgent - exactly what I need for my use case, but it seems this functionality isn't available in the current autogen_agentchat architecture.
Versions Used -
autogen-agentchat==0.5.7
autogen-ext==0.5.7
autogen-core==0.5.7
pyautogen==0.9

My Use Case:
Multi-agent swarm workflow with 11 specialized agents (Intent, Hypothesis, Anomaly detection)
Agents have detailed system messages (75+ lines for some agents like Hypothesis Runner)
Need automatic token compression when thresholds are reached
Want to preserve recent message context while compressing older messages

Questions:

Why was CompressibleAgent not included in the current architecture? Was it deprecated for specific reasons?
Is there a recommended approach for token management in multi-agent workflows with verbose system messages?
Are there plans to reintroduce similar functionality in upcoming releases?