Shared Memory

Agents without shared memory re-derive context every turn and contradict earlier conclusions. A shared store is the team's single source of truth.

Shared memory is the shared Google Doc everyone edits — not separate notebooks that get out of sync.

Scenario

Research agent writes `findings.acme_revenue` to shared memory; Writer agent reads it without re-searching; Supervisor reads both to check consistency.

Solution

In Multi-Agent Systems, apply Shared Memory to this scenario: Research agent writes `findings. 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 Shared Memory (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 Shared Memory happens in the code.

1class SharedMemory:  # define a data structure or component2    def __init__(self, redis):  # define a reusable function3        self.redis = redis4
5    def set(self, session_id, key, value):  # define a reusable function6        self.redis.hset(f"session:{session_id}", key, json.dumps(value))7
8    def get(self, session_id, key):  # define a reusable function9        return json.loads(self.redis.hget(f"session:{session_id}", key))  # return the result

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

Quick recap — the most important points from this module.