← Blog
3 min read

Agent Memory: Your Agent Remembers What It Already Read

Create sessions, mark briefs as read, and get filtered feeds that skip already-consumed content. No more duplicate processing.

The Problem: Duplicate Processing

Every time your agent queries the feed, it gets the same briefs it already processed. It re-reads them, re-summarizes them, burns tokens and credits on content it already consumed. There's no state between calls.

Agent Memory fixes this. Create a session, read briefs, mark them as read, and your next feed call automatically skips already-consumed content.

How It Works

Three steps:

1. Create a session

A named session tracks which briefs your agent has seen. Sessions persist across API calls.

2. Read the feed

Pass your session ID to any feed endpoint. First call returns everything. Subsequent calls skip briefs you've marked as read.

3. Mark as read

After processing, mark brief IDs as read. They won't appear in future feed calls for that session.

Python SDK

python
from veroq import PolarisClient

client = PolarisClient()

# Create a named session (or reuse an existing one)
session = client.create_session(name="daily-digest-agent")

# Get the feed — only unread briefs
feed = client.feed(category="ai", session_id=session.id)
print(f"New briefs: {len(feed.briefs)}")

# Process each brief
for brief in feed.briefs:
    print(f"[{brief.confidence:.0%}] {brief.headline}")
    # ... your agent logic here ...

# Mark them as read
brief_ids = [b.id for b in feed.briefs]
client.mark_as_read(session_id=session.id, brief_ids=brief_ids)

# Next call returns only NEW briefs
next_feed = client.feed(category="ai", session_id=session.id)
print(f"New briefs after marking: {len(next_feed.briefs)}")

Use Cases

Daily digest agents

Run your agent on a cron schedule. Each run picks up only the briefs published since the last run. No duplicates, no wasted tokens.

Monitoring agents

Agents that continuously monitor a topic can use sessions to ensure they only process each brief once, even across restarts.

Multi-agent workflows

Multiple agents can share a session to avoid duplicate work. If Agent A already processed a brief, Agent B won't see it. Or give each agent its own session for independent tracking.

What's Next