Describe the bug

The GET /v1/agents/{agent_id}/core-memory/blocks endpoint returns duplicate blocks when paginating.
The same block IDs are returned multiple times across pagination requests, resulting in the client receiving more blocks than actually exist.

Example

An agent created with 3 memory blocks (persona, system, human) returns:

• Total blocks returned: 6
• Unique block IDs: 3

Block ID counts

• block-ce930f61-e152-4e98-9a18-315b35581803 — 3 times (system)
• block-22ecddf7-93f5-4f6a-9807-85bf7372a80d — 2 times (human)
• block-6fec3b2e-002a-443f-8ef0-1469cb143d82 — 1 time (persona)

Server logs

The server shows multiple paginated requests being made:

GET /v1/agents/{id}/core-memory/blocks HTTP/1.1 200 OK
GET /v1/agents/{id}/core-memory/blocks?after=block-3553bfbb-... HTTP/1.1 200 OK
GET /v1/agents/{id}/core-memory/blocks?after=block-65e0062e-... HTTP/1.1 200 OK
GET /v1/agents/{id}/core-memory/blocks?after=block-6fec3b2e-... HTTP/1.1 200 OK

The pagination logic appears to not properly exclude already-returned blocks.

Note: Block updates still work correctly (all duplicates point to the same underlying block), so this is primarily a listing/display issue.

Please describe your setup

How are you running Letta?

• Docker

Environment

• OS: macOS (Darwin 24.5.0)
• Deployment: Docker Compose using letta/letta:latest
• Client: letta-client (latest)

Reproduction steps

from letta_client import Letta

client = Letta(base_url="http://localhost:8283")

# Create an agent with memory blocks
agent = client.agents.create(
    name="test-agent",
    model="openai/gpt-4o-mini",
    embedding="openai/text-embedding-ada-002",
    memory_blocks=[
        {"label": "persona", "value": "You are a helpful assistant."},
        {"label": "system", "value": "System instructions here."},
        {"label": "human", "value": "User context here."},
    ],
)

# List blocks - returns duplicates
blocks = list(client.agents.blocks.list(agent_id=agent.id))
print(f"Total blocks returned: {len(blocks)}")  # Returns 6, expected 3

# Verify duplicates
block_ids = [b.id for b in blocks]
unique_ids = set(block_ids)
print(f"Unique block IDs: {len(unique_ids)}")  # Returns 3

from collections import Counter
for bid, count in Counter(block_ids).items():
    if count > 1:
        print(f"Duplicate: {bid} appears {count} times")

Expected behavior

client.agents.blocks.list() should return each block exactly once.
For this example, the expected result is 3 total blocks.

Additional context

Model: openai/gpt-4o-mini

Embedding: openai/text-embedding-ada-002

Issue persists with freshly created agents

Deleting and recreating the agent does not resolve the issue