MemoryStore limitations and quotas: A quick reference for regions, memory stores, Scope, and SDK versions - Tablestore

Regions

MemoryStore is available only in the China (Beijing) region.

Memory store limits

Item	Limit
Memory store name characters	Letters, digits, and underscores only
Memory store name length	Up to 32 characters
Memory store description length	Up to 1,024 characters

After you create a memory store, index initialization takes about one minute. Wait for initialization to finish before you write or search data.

AddMemories limits

Item	Limit
`messages` count	Up to 20
`messages` total content length	Up to 32,000 characters
`text` length	Up to 32,000 characters
`messageId` length	Up to 256 characters
`metadata` key count	Up to 16
`metadata` key length	Up to 64 characters
`metadata` value length	Up to 1,024 characters

Provide at least one of messages or text. When you write memories, the wildcard * is not allowed in the Scope.

SearchMemories defaults

Parameter	Default	Description
`topK`	`10`	Number of results to return. Valid values: `0~50`.
`enableRerank`	`true`	Whether to enable reranking.

When you search long-term memories, appId and tenantId are required. You can use the wildcard * for agentId and runId.

Memory consolidation (Dream) limits

CreateMemoryDreamTask limits

Item	Limit
`scopes` count	Required. Up to 20.
`maxSessions`	`0~100`
`maxMessages`	`0~20000`
`maxMemories`	`0~5000`
`expandedScopeLimit`	`0~1000`
`instructions` length	Up to 4,000 characters
`actionIds` (`ApplyMemoryDreamActions`)	Up to 100 per call

Enum values

Field	Values
`taskType`	`memory` (default) / `skill` / `profile`
`applyMode`	`proposal` (default) / `safe_auto`. Supported only when `taskType`=`memory`.
`scopeOutputMode`	`preserve_scope` (default) / `promote_scope`
`confidenceThresholds` keys	`add` / `update` / `merge`. Values range from `0~1`. `DELETE` does not support auto confidence thresholds.
Task status (task `status`)	`queued` / `running` / `planning` / `applying` / `completed` / `completed_with_failures` / `failed` / `cancelled`
Action type (`action`)	`ADD` / `UPDATE` / `DELETE` / `MERGE` / `NOOP` / `EMIT_SKILL` / `EMIT_PROFILE`
Action status (action `status`)	`proposed` / `applied` / `skipped` / `failed`

Async tasks and list pagination

API	Default limit	Maximum limit
`ListMemoryTasks`	50	100
`ListMemoryStoreScopes`	100	100
`ListMemoryDreamTasks`	50	100
`ListMemoryDreamActions`	100	100

Extraction task status (ListMemoryTasks.status or the task.status returned by GetMemoryTask): queued / running / completed / failed / needs_reconcile.

Scope rules

Operation	Scope requirement	Wildcard `` allowed*
Write memories	`appId` is required. Empty levels default to `__default__`.	No
Search long-term memories	`appId` and `tenantId` are required.	`agentId` and `runId` allow wildcards
Query short-term memories	All four Scope levels are required.	No
Get a single long-term memory	All four Scope levels are required.	No
Update a single long-term memory	All four Scope levels are required.	No
Delete a single long-term memory	All four Scope levels are required.	No
List long-term memories	Specify Scope by level.	Yes
Query request audits	Specify Scope by level.	Yes

Wildcards must follow hierarchical order. After you use * at a level, all subsequent levels must also use * or be left empty. For example, app-001/user-001/*/* is valid, but app-001/*/agent-001/* is not.

Async write visibility

AddMemories.sync defaults to false (asynchronous). When async write is used:

Raw messages are written first and become immediately available as short-term memories.
Long-term memory extraction runs in the background.
After extraction completes, long-term memories become searchable through SearchMemories.

To view extraction results immediately during testing, set sync to true. Even with synchronous writes, a brief index refresh delay occurs before long-term memories become searchable.

Short-term memory queries

ListMemoryStoreMessages retrieves raw conversation messages. All four Scope levels (appId, tenantId, agentId, and runId) are required. Wildcards are not supported.

Use cases:

View raw conversation messages.
Replay a specific conversation.
Troubleshoot long-term memory extraction issues.

SDK versions

SDK	Minimum version
Python SDK	`tablestore >= 6.4.7` (`6.4.5` supports basic memory APIs; `6.4.7` adds task tracking and memory consolidation APIs)
Node.js SDK	`tablestore >= 5.6.5`
CLI	`tablestore-agent-cli >= 0.1.8`

CLI paging

CLI memory list commands return a single page of results. To retrieve the next page, pass the nextToken from the response.

Example:

tablestore-agent-cli memory list-units \
  --store agent_memory \
  --app-id app-001 \
  --next-token <token>

CLI auto-provisioning

If ots_endpoint and ots_instance_name are not configured, the CLI automatically creates a managed Tablestore instance in the China (Beijing) region when you run the doctor command or perform an operation. Auto-provisioning takes some time, and the result is saved to the local configuration file.

To use a different instance, manually set the endpoint and instance name. The CLI then uses your explicit configuration.

Agent plugin notes

The Hermes and OpenClaw plugins search across agents and conversations by default (agentId=*, runId=*) within the current tenant. If your business requires memory isolation between agents or conversations, use the SDK to control the search Scope directly, or adjust the plugin configuration.