The Scenario

Your chatbot has a 4000 token limit. Users complain that the bot "forgets" things mentioned 20 messages ago. Currently, you just truncate old messages.

The Problem

Truncation deletes information.

• User: "My name is Alice." (Message #1)
• ... 50 messages later ...
• User: "What is my name?"
• Bot: "I don't know." (Message #1 was deleted)

The Goal

Implement Summarization:

1. When history exceeds the limit, don't just delete.
2. Call mock_summarize(text) on the oldest messages.
3. Replace them with a single "System Summary" message.

Requirements:

• Target limit: 4000 tokens.
• If history > limit:
  • Identify oldest messages to drop.
  • Summarize them.
  • Insert summary at the start of history.
• Preserve System Prompt and Latest User Message.