Memory Compression

Uncompressed memory hits context limits and increases cost. Compression is mandatory for agents running 50+ step tasks.

Memory compression is packing for a trip — bring essentials, leave the 'just in case' items that never get used.

Scenario

A 50KB API response is compressed to 'Endpoint returned 200 with 3 users: Alice, Bob, Carol (IDs 1,2,3)' before entering context.

Solution

In Agent Memory, apply Memory Compression to this scenario: A 50KB API response is compressed to 'Endpoint returned 200 with 3 users: Alice, Bob, Carol (IDs 1,2,3)' before entering context. Identify the inputs, run the technique, validate the output, and note one thing you would monitor in production.

Do this before moving to the next module — reading alone is not enough.

Open the Code Walkthrough below and run it locally. Change one parameter related to Memory Compression (e.g. model, temperature, top_k, or tool name), observe the difference in output, and write 2–3 sentences explaining what changed.

Highlighted lines show where Memory Compression happens in the code.

1# Memory Compression — minimal example2from openai import OpenAI3
4client = OpenAI()  # create API client5
6# Ask the model to explain this topic7response = client.chat.completions.create(  # core API call for Memory Compression8    model="gpt-4o-mini",9    messages=[10        {"role": "system", "content": "You explain memory compression clearly."},11        {"role": "user", "content": f"What is memory compression?"},12    ],13    temperature=0,14)15print(response.choices[0].message.content)  # show output for debugging

• Treating Memory Compression as a black box without evaluation
• Ignoring cost and latency in production
• Skipping error handling for memory compression

Quick recap — the most important points from this module.