Seeding

AWSim ships a bulk-seeder that fills services with realistic fake data. It bypasses SigV4 + the gateway and writes straight into each service's internal state, so a 10k-user / 1k-table seed completes in well under a second instead of taking the full per-request hit.

Three entry points sit on top of the same admin endpoints:

1. /seed UI page — service cards with count inputs and Run buttons. Quick + interactive.
2. /_awsim/seed/<service> admin endpoints — script directly with curl or your favourite client.
3. awsim seed --file seed.toml — reproducible scenarios for CI / dev fixtures.

UI

Open Admin → Seed data in the sidebar. Each service has its own card with sensible defaults:

Card	Inputs
Cognito users	Pool (dropdown of existing pools) + count
DynamoDB	Tables + items per table
S3	Buckets + objects per bucket + body bytes
Secrets Manager	Count
SQS	Queues + messages per queue

Cards show a status badge (idle / running / done / error) and a result line with what was actually created. Run buttons are disabled while the request is in flight so double-clicks don't kick off duplicate seeds.

Admin endpoints

All endpoints accept JSON POST bodies and return a counts summary. Defaults: account from --account-id, region from --region.

POST /_awsim/seed/cognito-users

{
  "pool_id":  "us-east-1_abcdef",
  "count":    10000,
  "prefix":   "seed-",          // optional, default "seed-"
  "password": "Seed-Pass-1234!" // optional, default meets typical pool policy
}

Each user gets a random name + email, status biased ~80% CONFIRMED / ~15% FORCE_CHANGE_PASSWORD / ~5% UNCONFIRMED, ~95% enabled, email_verified flipped on for ~90% of confirmed users. Capped at 100 000 / call.

Returns: { "created": N, "skipped": M, "pool_id": "..." }.

POST /_awsim/seed/dynamodb

{
  "tables":          5,
  "items_per_table": 1000,
  "prefix":          "seed",   // optional
  "account":         "...",    // optional
  "region":          "..."     // optional
}

Each table gets a single id (String) hash key and items_per_table items with random name, email, score, and active attributes. Items go through the same SQLite store the gateway path uses, so they're immediately queryable. Capped at 1 000 tables / 100 000 items per table.

Returns: { "tables_created": N, "items_created": M, "errors": [] }.

POST /_awsim/seed/s3

{
  "buckets":            5,
  "objects_per_bucket": 100,
  "body_bytes":         256,   // capped at 64 KiB
  "prefix":             "seed",
  "account":            "...",
  "region":             "..."
}

Each object body is body_bytes of deterministic ASCII so the seed is fast and repeatable; ETags match the body content. Capped at 500 buckets / 10 000 objects per bucket.

Returns: { "buckets_created": N, "objects_created": M }.

POST /_awsim/seed/secrets

{
  "count":   20,
  "prefix":  "seed",
  "account": "...",
  "region":  "..."
}

Each secret carries a JSON body shaped like a typical credential blob: { username, password, host, port }. Capped at 50 000 / call.

Returns: { "created": N }.

POST /_awsim/seed/sqs

{
  "queues":             5,
  "messages_per_queue": 50,
  "prefix":             "seed",
  "account":            "...",
  "region":             "..."
}

Each message body is a random fake sentence with the correct md5_of_body hash. Queues are created as standard (non-FIFO). URLs are built against awsim's own port so the seeded queues are immediately usable by SDK clients. Capped at 1 000 queues / 100 000 messages per queue.

Returns: { "queues_created": N, "messages_created": M }.

CLI: awsim seed

Use the CLI subcommand for reproducible scenarios — commit a seed.toml to your repo and any teammate (or CI job) can recreate the same fixture with one command:

awsim seed --file fixtures/dev.toml

# Override the endpoint:
awsim seed --file fixtures/dev.toml --endpoint http://staging-awsim.local:4566

# Or via env var:
AWSIM_ENDPOINT=http://staging-awsim.local:4566 awsim seed --file fixtures/dev.toml

Endpoint resolution: --endpoint flag → endpoint in the TOML file → AWSIM_ENDPOINT env var → http://localhost:4566.

Output is one ✓-prefixed line per service with the JSON response, plus a final summary line:

✓ Cognito users: {"created":1000,"skipped":0,"pool_id":"us-east-1_abc"}
✓ DynamoDB: {"tables_created":5,"items_created":5000,"errors":[]}
✓ S3: {"buckets_created":5,"objects_created":500}
✓ Secrets Manager: {"created":20}
✓ SQS: {"queues_created":5,"messages_created":250}
✓ Seed complete.

A failing service stops the run and surfaces the error message.

TOML shape

Each section is optional; omit a service to skip it. cognito_users is an array (one entry per pool); the other sections are single tables.

endpoint = "http://localhost:4566"   # optional

[[cognito_users]]
pool_id = "us-east-1_abcdef"
count   = 1000

# Multiple pools? Just add another [[cognito_users]] block.
[[cognito_users]]
pool_id = "us-east-1_xyz789"
count   = 500
prefix  = "load-"

[dynamodb]
tables          = 5
items_per_table = 1000

[s3]
buckets             = 5
objects_per_bucket  = 100
body_bytes          = 256

[secrets]
count = 20

[sqs]
queues             = 5
messages_per_queue = 50

Cleanup

The seeders use a seed- prefix by default for resource names so the synthetic data is easy to spot. Drop a named snapshot before seeding if you want to roll back cleanly:

curl -XPOST http://localhost:4566/_awsim/snapshots/pre-seed
# ...seed and use...
curl -XPOST http://localhost:4566/_awsim/snapshots/pre-seed/load

Or just restart with a fresh --data-dir (or no data-dir) to nuke everything.

When not to use seed

The seeders prioritise speed over realism — they don't fire Lambda triggers, don't generate stream records, and don't go through SigV4 / IAM. If you need any of that, write a small script that uses the AWS SDK against AWSim instead. The seed endpoints are meant for "I just need 10k rows in a table" workloads.