fshy

fshy / Netdisk and pan / Storage

Lark /

larksuite.com · Netdisk and pan, Storage · 2 records · JSON

Lark / 飞书 Drive via multi-tenant app credentials (the FeiSync pattern) Netdisk and pan

caveatemailno cardphoneRper-object-cap

open.larksuite.com/app · platform storage reached through app credentials · netdisk-pan/lark-drive-via-multi-tenant-app-credentials · JSON · compare

Free tier100 GB per Lark tenant. '任意扩容,只要添加 App ID 和 App Secret 就可以免费获得100G空间,单文件250M,简单说就是容量上限取决于你的号数量'
unit GiB; metered on stored-bytes
Pricenone stated
Per-object cap250 MB per file (250000000 bytes)
RetentionWhile the tenant persists. No retention clock published
EgressUnmetered in practice; there is no egress line item on the free tier
Identityemail a distinct email address or phone number per tenant. Lark has historically required an overseas IP to register outside the mainland, which is a feature here
Limits
  • per-object-cap 250 MB per file source
  • request-rate apps inside one tenant share that tenant's 100 GB, so a second app is not a second quota source
Bindingper-object-cap 100 GB per tenant multiplies cleanly and the traffic is free, so the only number that shapes a plan is 250 MB per file: the 20 TiB archive becomes 84,000 objects, and the 10 TiB working set 42,000.
Interfacerest
ToolingFeiSync v1.0.0 by fish2018, Tauri and Rust with a Vue front end, presented on linux.do on 2025-12-28: multi-tenant routing, one-way or two-way cron sync, resumable transfer and an open HTTP API. The underlying API is the Lark Drive open platform v1 at /drive/v1/files
Composes withThis is the Google service-account pattern reimplemented for Lark: create N free tenants, register a custom app in each, and FeiSync load-balances writes across them for 100 GB times N in one flat namespace. 250 MB per file is the binding number, not the bytes
ReliabilityTwo constraints bite: 250 MB per file caps the cold-archive case hard, and app-permission changes break the wrapper. Tenant creation needs a distinct identity each time
NotesThe most directly applicable of the app-credential patterns, because tenancy is free and self-service. The 250 MB file cap is what makes it a shard target rather than a store, and that is a job for a chunker rather than a blocker.
Evidence rungreported
Checked2026-10-21 (inherited); stale on 2026-11-04
Parsed numbers
  • $.limits[0].value limit:per-object-cap 250.0 MB
  • $.limits[1].value limit:request-rate 100.0 GB
Evidence (3)
datesourcewhat it proved
2026-10-21linux.do/t/topic/1371150'免费无限扩容,每个账号免费扩容100G,单文件250M' and '不能是同一个企业的不同应用,会共享100G'
2026-10-21linux.do/t/topic/360439'larksuite所谓飞书海外版,必须要海外IP才能注册' '海外ip,会成功'
2026-10-21linux.do/t/topic/588707'Lark 免费用户就有自定域名邮箱、100 GB 存储'
More fields (6)
{
 "access_mode": "registration-gated",
 "wrapper": "FeiSync v1.0.0 (fish2018)",
 "quota_per_tenant_bytes_derived": 107374182400,
 "per_file_cap_bytes_derived": 250000000,
 "overseas_ip_required": true,
 "aliases": [
  "Lark / 飞书 Drive via multi-tenant app credentials (FeiSync pattern)"
 ]
}

Lark (Feishu) Starter - free plan Storage

caveatemail?card?Rper-object-cap

www.larksuite.com · paas · storage/lark-starter-free-plan · JSON · compare

Free tier100 GB total storage per organization on Starter (Free); 250 MB single-file limit for Docs uploads on Starter
unit GB; metered on stored-bytes
Pricenone stated
Per-object cap250 MB per file (250000000 bytes)
Retentionnot stated
EgressDownloads served from Lark CDN (ByteDance edge) at no metered charge; download via OpenAPI media download returns a signed URL, no per-GB line item anywhere in the plan table
Identityemail (placeholder) not published
Limits
  • per-object-cap 250 MB per file source
250 MB per file for Docs on Starter (vs 50 GB on Pro); upload_all hard-caps at 20 MB by design, must use multipart; 5 QPS on upload_all; 4 MB fixed chunk size on multipart; upload transaction ID valid 1 day for resume
Bindingper-object-cap the per-object cap sets the shard size and therefore the object count; it bounds the total more tightly than any headline quota
Interfacerest, rclone, cli rest
ToolingOfficial OpenAPI v3 drive endpoints (upload_all for <=20MB, multipart upload of 4MB chunks above that) + first-party CLI larksuite/cli (17,762 stars, pushed 2026-10-21) with lark-cli drive +upload; no rclone backend
Composes with100 GB x N free organizations. Starter is self-serve per organization and the quota is per-org, not per-user, so N disposable orgs = N x 100 GB. Paid Basic (5 TB) shows the ceiling is only billing, not architecture.
ReliabilityRate limits are explicit (5 QPS on single upload) and multipart is resumable, but Starter orgs that lapse back from a paid plan are force-collapsed to Starter member and storage caps - data above 100 GB becomes unwritable, not deleted.
Notes100 GB total storage per organization on Starter (Free); 250 MB single-file limit for Docs uploads on Starter
Evidence rungreported One page read moves it to documented.
Checked2026-10-21 (inherited); stale on 2026-11-04
Parsed numbers
  • $.limits[0].value limit:per-object-cap 250.0 MB
Evidence (3)
datesourcewhat it proved
2026-10-21www.larksuite.com/hc/en-US/articles/497171699982-overview-oftable row: 'Total storage space | 100 GB | 5 TB | 15 TB | 15 TB + 30 GB x purchased seats'; 'Single file upload limit for Docs | 250 MB'
2026-10-21open.larksuite.com/document/uAjLw4CM/ukTMukTMukTM/reference/'please do not use this interface to upload files larger than 20MB... use the multipart upload interface', supports 5QPS
2026-10-21open.larksuite.com/open-apis/auth/v3/tenant_access_token/intnot published
Unknown or withheld values (7)
  • $.evidence[2].proved withheld
  • $.free_tier.notes process
  • $.identity_cost.detail defaulted
  • $.identity_cost.kind identity-defaulted
  • $.per_object_cap.notes not-extracted
  • $.retention.notes not-extracted
  • $.retention.stated not-extracted