Skip to main content

Quickstart (AWS)

Wire memledger to Aurora PostgreSQL + Bedrock — the canonical AWS production path. You'll use Aurora for vector storage and Bedrock Titan for embeddings.

This is the AWS reference path

For getting started without AWS credentials, use the OSS quickstart (in-cluster Postgres + local fastembed) and graduate here later.

0. Prerequisites​

  • An AWS account with Bedrock model access for amazon.titan-embed-text-v2:0 (request access in the Bedrock console — typically minutes).
  • An Aurora PostgreSQL cluster (≥ 14) with the vector extension enabled. RDS Postgres also works.
  • AWS credentials available to your shell (AWS_PROFILE or AWS_ACCESS_KEY_ID / AWS_SECRET_ACCESS_KEY).

1. Install​

pip install "memledger[aws]"

The [aws] extra includes Aurora-via-pgvector storage and Bedrock embeddings. Add [aws,opensearch] to opt into OpenSearch k-NN as the vector store.

2. Enable pgvector on Aurora​

Run once against your Aurora cluster:

CREATE EXTENSION IF NOT EXISTS vector;

For Aurora IAM database authentication, set iam_auth: true in backend_config and memledger will mint short-lived tokens via boto3 at connection time:

ml = await Memledger.create(
    backend_name="pgvector",
    backend_config={
        "host": "<your-aurora-cluster-endpoint>",
        "port": 5432,
        "database": "memledger",
        "user": "memledger",
        "iam_auth": True,
        "region": "<your-region>",
    },
    embedding_config=EmbeddingConfig(
        provider="bedrock",
        model="amazon.titan-embed-text-v2:0",
        dimensions=1024,
    ),
)

The agent's IAM principal needs rds-db:connect on the cluster's resource ID + master user ARN. See Aurora PostgreSQL for the full IAM setup.

3. Write and search a memory​

import asyncio
import os
from memledger import Memledger
from memledger.models import EmbeddingConfig

AURORA_DSN = os.environ["MEMLEDGER_AURORA_DSN"]
# postgresql://USER:PASS@<your-aurora-cluster-endpoint>:5432/memledger

async def main():
    ml = await Memledger.create(
        backend_name="pgvector",
        connection_string=AURORA_DSN,
        embedding_config=EmbeddingConfig(
            provider="bedrock",
            model="amazon.titan-embed-text-v2:0",
            dimensions=1024,
        ),
    )

    await ml.add(
        content="EKS node group rotation: drain at 5% capacity, not 0%, to avoid pod eviction storms",
        namespace="/ops/eks/runbooks",
        confidence=0.9,
        created_by="sre-agent",
    )

    results = await ml.search(
        query="how to drain EKS nodes safely",
        namespace="/ops/eks/runbooks",
    )
    for r in results.records:
        print(r.confidence, r.created_by, "::", r.content)

    await ml.close()

asyncio.run(main())

The same chain demo from Quickstart (OSS) works unchanged here — only the backend config differs.

4. Optional: alternative AWS backends​

pip install "memledger[aws,opensearch]"     # OpenSearch k-NN as the vector store

See the AWS backend pages for setup details:

Where next​