Compare commits

...
99 Commits
Author SHA1 Message Date
yiling 7753721731 Release v0.0.8 2026-06-19 20:42:48 +08:00
Yiling-J ef86cc2713 Merge pull request #138 from shumaiOne/feat/upload-progress-xhr
feat(upload): track byte-level progress with XMLHttpRequest
2026-06-19 20:11:28 +08:00
yiling cc79be7e34 fix(upload): resolve filesToUpload mutation race condition
Capture the files list snapshot at the time the mutation is triggered via React Query's onMutate context, and retrieve it in onSuccess via the context parameter to prevent race conditions during rapid upload successions.
2026-06-19 20:05:02 +08:00
yiling a9c5a3d9fd fix(ui): resolve Cannot access 'displayItem' before initialization ReferenceError
Use item.id directly since it is available at the start of the component from props.
2026-06-19 19:49:30 +08:00
yiling 0d9545f0b5 feat(upload): track byte-level progress with XMLHttpRequest
- Replace fetch uploads with XMLHttpRequest to capture upload progress
- Track active upload tasks and file progress in Zustand store
- Show byte-level uploader progress in the sidebar list
- Replace plain 'Uploading' card status text with a circular percentage animation
- Replace plain document icon in file list view rows with a mini progress circle
2026-06-19 19:44:35 +08:00
Yiling-J 2925fb6bf6 Merge pull request #137 from shumaiOne/feature/switch-and-rating-filter
feat(webui): use Shadcn Switch in toggle field and Rating component in search filter panel
2026-06-19 15:58:41 +08:00
yiling 7ae34a4bc0 feat(webui): use Shadcn Switch in toggle field and Rating component in search filter panel 2026-06-19 15:49:54 +08:00
Yiling-J f645c8fefa Merge pull request #136 from shumaiOne/fix-ui-issues
fix(webui): resolve folder tree sync, search dialog reset, and sidebar actions issues
2026-06-19 15:17:46 +08:00
yiling 1835016a9b fix(webui): use poster field instead of thumbnail for video notifications 2026-06-19 13:34:14 +08:00
yiling 5e92660d1c fix(webui): render sprite image scrubber for video upload notifications 2026-06-19 13:25:00 +08:00
yiling 6cb3ca6874 fix(webui): resolve folder tree sync, search dialog reset, and sidebar actions issues
- Change query key invalidation of folders to the generic ['folders'] key to update the sidebar tree on create/rename/delete/move
- Separate search box from custom filters reset button in search dialog
- Remove unused plus icon from Assets sidebar header
- Add dropdown menu (rename, download, delete) to subfolder tree items in the sidebar
2026-06-19 12:55:30 +08:00
Yiling-J 5d14f4fff1 Merge pull request #135 from shumaiOne/fix/search-dialog-ui-improvements
style(webui): fix search dialog layout height, scroll areas, and restrict filter toggle click
2026-06-19 12:32:30 +08:00
yiling 009566d42f style(webui): add min-h-0 to search results scroll area to allow scrolling 2026-06-19 12:23:15 +08:00
yiling 50777d332b style(webui): fix search dialog layout height, scroll areas, and restrict filter toggle click 2026-06-19 12:11:19 +08:00
Yiling-J 852a9873dc Merge pull request #134 from shumaiOne/fix/search-has-any
fix(search): fix hasAny and hasNone operators on metadata filter
2026-06-19 11:37:35 +08:00
yiling aa265c5322 fix(search): fix hasAny and hasNone operators on metadata filter 2026-06-19 11:29:26 +08:00
Yiling-J 01ab254c21 Merge pull request #133 from shumaiOne/ui/default-right-sidebar-collapsed
feat(webui): default right sidebar to collapsed for new users
2026-06-19 11:03:27 +08:00
yiling ca5ca5e080 feat(share): default right sidebar to collapsed on public share list page and expanded on detail page 2026-06-19 10:58:38 +08:00
yiling 3f27439733 feat(share): show right sidebar toggle in public share topnav and make collapse state local 2026-06-19 10:53:01 +08:00
yiling 2a611f9b5a fix(share): sync layout options and support right sidebar collapse on public page 2026-06-19 10:44:22 +08:00
yiling 7137e00911 fix(share): resolve share link drag-and-drop error and public view render loop 2026-06-19 10:32:21 +08:00
yiling 59db97aedf fix(share): fix right sidebar layout overflow and button cutoff under topnav 2026-06-19 10:24:58 +08:00
yiling a76dbdfc22 feat(share): improve share config page sidebars, layout options, and sorting 2026-06-19 10:22:20 +08:00
yiling 61e2a3c53a feat(webui): add separate sidebar configs for share config pages 2026-06-19 09:57:29 +08:00
yiling 7707498916 feat(webui): default right sidebar to collapsed for new users 2026-06-19 09:50:02 +08:00
Yiling-J 91c8af86db Merge pull request #132 from shumaiOne/feat/demo-mode-refinement
feat(auth): refine demo mode support
2026-06-19 00:44:00 +08:00
Yiling-Jandgemini-code-assist[bot] 286d71133f Update packages/webui/routes/signup.tsx
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
2026-06-19 00:39:02 +08:00
yiling 53c132e41d feat(auth): refine demo mode support using SHUMAI_DEMO_MODE API 2026-06-19 00:36:23 +08:00
Yiling-J 5c285b172b Merge pull request #131 from shumaiOne/feat/demo-mode-support
feat(auth): add demo mode support
2026-06-19 00:22:26 +08:00
yiling f1d52ac44e feat(auth): add demo mode support using PUBLIC_DEMO_MODE 2026-06-19 00:17:14 +08:00
Yiling-J ec6736b452 Merge pull request #130 from shumaiOne/fix/field-colors-and-filter-selects
fix(webui): use system colors for fields and improve search filter select options
2026-06-18 23:27:48 +08:00
yiling 46c975836a fix(webui): render selectMulti 'is exactly' operator as single-select 2026-06-18 23:22:36 +08:00
yiling 5d7ea7d118 refactor(webui): use Popover instead of Combobox for search dialog MultiSelect 2026-06-18 23:18:17 +08:00
yiling 5065346e8a fix(webui): support fallback options for selectMulti fields in filter panel 2026-06-18 23:13:51 +08:00
yiling 60d469c91b fix(webui): use system colors for fields and improve search filter select options 2026-06-18 23:03:57 +08:00
Yiling-J f68c1c56e1 Merge pull request #129 from shumaiOne/fix/project-updated-at-asset-hook
fix(db): automatically update project updatedAt on asset modification
2026-06-18 22:44:20 +08:00
yiling 08cf90477a fix(db): inline project updatedAt touch on asset update 2026-06-18 22:39:23 +08:00
yiling e01c348fde fix(db): automatically update project updatedAt on asset modification 2026-06-18 21:53:16 +08:00
Yiling-J b0eb5b1eef Merge pull request #128 from shumaiOne/feat/video-search-jump-timestamp
feat(webui): support jumping to video timestamp from search results
2026-06-18 21:06:58 +08:00
yiling 3b7fc410b2 feat(webui): support jumping to video timestamp from search results 2026-06-18 20:35:50 +08:00
Yiling-J 4c7ff6f58f Merge pull request #127 from shumaiOne/feat/semantic-search-chunk-overlap
feat(agent): implement overlapping video chunks for semantic search
2026-06-18 20:02:26 +08:00
yiling a197c83580 fix(agent): prevent redundant final video chunks in embedding workflow 2026-06-18 19:57:01 +08:00
yiling 718d9d5b51 feat(transcode): optimize video chunks for embedding by forcing 1 FPS and dropping audio 2026-06-18 19:53:19 +08:00
yiling 212ac3bc0e docs(search): add detail about video chunking and configuration to semantic search docs 2026-06-18 19:46:50 +08:00
yiling eaf2c867b3 feat(agent): implement overlapping video chunks for semantic search and update docs 2026-06-18 19:45:58 +08:00
Yiling-J 570d95fa6f Merge pull request #126 from shumaiOne/feat/external-drag-drop
feat(webui): add external drag-and-drop file and folder upload support
2026-06-18 16:21:23 +08:00
yiling 32892a21c1 fix(webui): guard external drag and drop in share, public, and collection views 2026-06-18 16:13:40 +08:00
yiling 7819a2034e fix(webui): clear external drag state synchronously on folder drops 2026-06-18 15:58:34 +08:00
yiling e0f2698a66 feat(webui): add external drag-and-drop file and folder upload support 2026-06-18 15:39:08 +08:00
Yiling-J 72c50ffd73 Merge pull request #125 from shumaiOne/improve-embedding-workflow
feat(embedding): defer embedding workflow, use transcoded assets, and execute chunks in separate activities
2026-06-18 15:02:21 +08:00
yiling d84950d372 feat(transcode): set loglevel to warning for all ffmpeg commands 2026-06-18 11:57:38 +08:00
yiling 46815736e2 feat(embedding): decouple video chunk transcode and embedding generation 2026-06-18 11:54:30 +08:00
yiling 9d1e8378f7 fix(embedding): resolve transcoded MIME type and validate EMBEDDING_CHUNK_DURATION positive limits 2026-06-18 11:14:09 +08:00
yiling f399d2c469 feat(embedding): defer embedding workflow, use transcoded assets, and execute chunks in separate activities 2026-06-18 10:57:22 +08:00
Yiling-J 4af081f128 Merge pull request #124 from shumaiOne/feat/custom-server-port
feat(server): allow customizing server port via SHUMAI_SERVER_PORT
2026-06-18 10:32:36 +08:00
yiling ce457e1161 fix(s3): remove port from AWS_ENDPOINT_URL_S3 and sync with server port in local mode 2026-06-18 10:22:16 +08:00
yiling a69dd6a660 feat(server): allow customizing server port via SHUMAI_SERVER_PORT 2026-06-18 10:08:31 +08:00
Yiling-J 62d7061cb3 Merge pull request #123 from shumaiOne/fix/fps-transcode-failure
fix(transcode): handle high-precision FPS using rational fractions
2026-06-18 09:24:11 +08:00
yiling 1ad5720f42 fix(transcode): handle high-precision FPS using rational fractions 2026-06-17 22:55:01 +08:00
Yiling-J 0be74dd4ab Merge pull request #122 from shumaiOne/feat/read-only-mode
feat(api): implement read-only mode for reviewers
2026-06-17 22:35:25 +08:00
yiling f4276a19aa feat(api): allow search API in read-only mode 2026-06-17 22:30:09 +08:00
yiling d81fc275f4 fix(api): correct lint errors in authMiddleware tests 2026-06-17 22:19:44 +08:00
yiling ead91455d8 feat(api): implement read-only mode for reviewers controlled by REVIEWER_READ_ONLY env 2026-06-17 21:51:34 +08:00
yiling 438ea79535 Release v0.0.7 2026-06-17 17:02:21 +08:00
Yiling-J 3455cd0b13 Merge pull request #121 from shumaiOne/feat/api-protocol-enum
feat(api): use enum for API protocol in provider config
2026-06-17 16:48:52 +08:00
yiling 2aba1c898b feat(core): allow deleting builtin providers 2026-06-17 16:39:23 +08:00
yiling 6f91b9e809 feat(webui): allow deleting builtin providers and improve dialog layout 2026-06-17 16:31:51 +08:00
yiling 10760f5a57 fix(webui): finalize provider dialog layout for better scrolling 2026-06-17 16:26:10 +08:00
yiling c150f62b4c fix(webui): improve dialog layout for scrollable content 2026-06-17 16:15:04 +08:00
yiling 1284c1de3f fix(webui): make model list in provider dialog scrollable 2026-06-17 16:12:35 +08:00
yiling 0c89616a4e feat(api): use enum for API protocol in provider config 2026-06-17 16:07:48 +08:00
Yiling-J 96a682b414 Merge pull request #118 from shumaiOne/docs/improve-shumai-docs
docs: improve and restructure shumai documentation
2026-06-17 15:48:00 +08:00
Yiling-J 5f8067cbe8 Merge pull request #120 from shumaiOne/feat/replace-overflow-with-scrollarea
feat(webui): replace native overflow with shadcn ScrollArea in settings
2026-06-17 15:43:09 +08:00
yiling ae39a98e84 feat(webui): replace native overflow with shadcn ScrollArea in settings 2026-06-17 15:41:43 +08:00
Yiling-J 0944f1ad8e Merge pull request #119 from shumaiOne/feat/ui-system-colors
feat(webui): replace hardcoded colors with system tokens
2026-06-17 15:23:46 +08:00
yiling 812a8b3b88 feat(webui): replace hardcoded colors with system tokens in settings components 2026-06-17 15:22:46 +08:00
yiling 24a2f74035 update ui 2026-06-17 11:54:42 +08:00
yiling 73d02bf319 docs: convert screenshots to webp and update documentation 2026-06-17 11:37:48 +08:00
yiling 6e98d6ee4a docs: document collection creation from search dialog and dragging assets to share links 2026-06-16 16:34:43 +08:00
yiling af381ff99c docs: update notifications and metadata documentation to match UI/API configurations 2026-06-16 16:27:11 +08:00
yiling c9da7794d9 docs: explain team member visibility scopes and role overrides 2026-06-16 15:51:35 +08:00
yiling 1fb03cd852 docs: describe unified team and project members dialog 2026-06-16 15:45:59 +08:00
yiling 74fff87a2b docs: describe search filter dialog and drag reorder rules 2026-06-16 15:33:33 +08:00
yiling 3af5940ddf docs: update version stack creation workflows 2026-06-16 15:21:10 +08:00
yiling 3bb44041d7 docs: rename upload to asset management and add delete and clipboard paste docs 2026-06-16 15:13:26 +08:00
yiling 29b7124f48 docs: move temporal deployment guide to separate advanced deployment section 2026-06-16 14:56:38 +08:00
yiling 718f50ef10 docs: add temporal deploy guide and link from quickstart 2026-06-16 14:53:03 +08:00
yiling 006774042a docs: refine documentation to be user-oriented and remove database jargon 2026-06-16 14:44:42 +08:00
yiling 2b058bde40 docs: remove legacy introduction and docker docs 2026-06-16 14:39:04 +08:00
yiling fc6ad20329 docs: improve and restructure shumai documentation 2026-06-16 12:15:53 +08:00
Yiling-J 7e5735f0cc Merge pull request #116 from shumaiOne/feat/concurrency-control
feat(workflow): configure transcode and agent concurrency via environment variables
2026-06-16 10:46:45 +08:00
yiling 6e74ede9fe fix(workflow): validate concurrency limits and fallback to defaults on invalid or non-positive values 2026-06-16 10:40:39 +08:00
yiling b8d876d28e feat(workflow): allow configuring transcode and agent concurrency via environment variables 2026-06-16 10:24:45 +08:00
Yiling-J 95984f19b1 Merge pull request #115 from shumaiOne/refactor/transcode-service-and-api-flattening
refactor(transcode): relocate TranscodeService to core and flatten API package
2026-06-16 09:58:25 +08:00
yiling 5111fcd6a5 refactor(transcode): address code review feedback on imports and dataFormatNames typing 2026-06-16 09:53:28 +08:00
yiling f16a8a7069 refactor(transcode): relocate TranscodeService to core and flatten API package 2026-06-16 09:47:27 +08:00
Yiling-J 7f6e41708a Merge pull request #114 from shumaiOne/fix/video-player-carousel-change
fix(webui): reload video player and reset state when switching assets
2026-06-16 09:35:23 +08:00
yiling 6fa69d3310 fix(webui): use fileData?.id as reset dependency to prevent loading race conditions 2026-06-16 09:25:40 +08:00
yiling 7c5f462ba5 fix(webui): reload video player and reset state when switching assets 2026-06-16 09:21:56 +08:00
177 changed files with 5146 additions and 1601 deletions
+2 -1
View File
@@ -244,7 +244,7 @@ The `bunfig.toml` is configured to run `setup-tests.ts` automatically as a prelo
### API Tests
- Located in `packages/api/src/api/**/*.test.ts`.
- Located in `packages/api/src/**/*.test.ts`.
- Mock the Service layer using `vi.spyOn(service, 'method')` or `vi.mock('@shumai/core/src/team/team')`.
- Verify that the API calls the Service methods correctly.
@@ -396,6 +396,7 @@ To prevent Temporal from indefinitely retrying fatal, expected business validati
## Radix UI / Shadcn UI Patterns
- **Modal Overlays**: When triggering a dialog (e.g., `Dialog`, `AlertDialog`) from inside a `DropdownMenu`, you must set `modal={false}` on the `<DropdownMenu>` component. Failing to do so will cause the UI to freeze due to conflicting focus management between the two modal components.
- **Scrollable Content**: Always use the Shadcn `ScrollArea` component (`@/ui/components/ui/scroll-area`) for scrollable content containers to ensure consistent custom scrollbar UI across the application.
## Logging
+5 -4
View File
@@ -44,7 +44,7 @@ The easiest way to run Shumai is using Docker Compose. You do not need to clone
curl -o docker-compose.yaml https://raw.githubusercontent.com/shumaiOne/shumai/main/docker-compose/local/docker-compose.yaml
```
2. Configure environment variables (if deploying remotely):
If you are deploying Shumai to a remote server (e.g., AWS ECS, EC2, VPS), you must edit `docker-compose.yaml`, set `AWS_ENDPOINT_URL_S3` to your server's public IP or domain name (e.g., `http://12.345.567.789:3000`).
If you are deploying Shumai to a remote server (e.g., AWS ECS, EC2, VPS), you must edit `docker-compose.yaml`, set `AWS_ENDPOINT_URL_S3` to your server's public IP or domain name (e.g., `http://12.345.567.789`).
3. Start the services:
```bash
docker compose up -d
@@ -102,11 +102,12 @@ Create a `.env` file in this folder and add the following configuration (which a
DATABASE_URL=postgresql://shumai:shumai_password@localhost:5432/shumai_db?schema=public
BETTER_AUTH_SECRET=ySxs7DxzHDZBbeeHNPEwBuspYwipBqz5Gk5XdBjNhWw=
STORAGE_BACKEND=local
AWS_ENDPOINT_URL_S3=http://localhost:3000
SHUMAI_SERVER_PORT=3000
AWS_ENDPOINT_URL_S3=http://localhost
```
> [!NOTE]
> If you are deploying on a remote server, make sure to change `AWS_ENDPOINT_URL_S3` from `http://localhost:3000` to your server's public IP address or domain name.
> If you are deploying on a remote server, make sure to change `AWS_ENDPOINT_URL_S3` from `http://localhost` to your server's public IP address or domain name. `SHUMAI_SERVER_PORT` defaults to `3000` if not set.
#### Step 4: Run Shumai
@@ -189,7 +190,7 @@ DATABASE_URL=postgresql://shumai:shumai_password@<host-ip>:5432/shumai_db?schema
WORKFLOW_EXECUTOR=temporal
TEMPORAL_ADDRESS=<host-ip>:7233
STORAGE_BACKEND=s3
AWS_ENDPOINT_URL_S3=http://<host-ip>:3000 # Or your public S3/R2 endpoint URL
AWS_ENDPOINT_URL_S3=http://<host-ip> # Or your public S3/R2 endpoint URL
S3_REGION=us-east-1
S3_BUCKET=shumai
S3_ACCESS_KEY_ID=your-access-key-id
+1 -1
View File
@@ -1,6 +1,6 @@
{
"name": "@shumai/agent-app",
"version": "0.0.6",
"version": "0.0.8",
"private": true,
"type": "module",
"bin": {
+1 -1
View File
@@ -1,6 +1,6 @@
{
"name": "@shumai/transcode-app",
"version": "0.0.6",
"version": "0.0.8",
"private": true,
"type": "module",
"bin": {
+1 -1
View File
@@ -1,6 +1,6 @@
{
"name": "@shumai/web-app",
"version": "0.0.6",
"version": "0.0.8",
"private": true,
"type": "module",
"dependencies": {
+4 -2
View File
@@ -44,10 +44,12 @@ if (process.env.WORKFLOW_EXECUTOR === 'temporal') {
const isProd = process.env.NODE_ENV === 'production'
const port = process.env.SHUMAI_SERVER_PORT ? parseInt(process.env.SHUMAI_SERVER_PORT) : 3000
const server = Bun.serve(
isProd
? {
port: 3000,
port,
maxRequestBodySize: process.env.MAX_REQUEST_BODY_SIZE
? parseInt(process.env.MAX_REQUEST_BODY_SIZE)
: 1024 * 1024 * 1024 * 10, // Default 10GB
@@ -88,7 +90,7 @@ const server = Bun.serve(
},
}
: {
port: 3000,
port,
maxRequestBodySize: process.env.MAX_REQUEST_BODY_SIZE
? parseInt(process.env.MAX_REQUEST_BODY_SIZE)
: 1024 * 1024 * 1024 * 10, // Default 10GB
+12 -12
View File
@@ -32,7 +32,7 @@
},
"apps/agent": {
"name": "@shumai/agent-app",
"version": "0.1.0",
"version": "0.0.6",
"bin": {
"shumai-agent": "./src/index.ts",
},
@@ -44,7 +44,7 @@
},
"apps/transcode": {
"name": "@shumai/transcode-app",
"version": "0.1.0",
"version": "0.0.6",
"bin": {
"shumai-transcode": "./src/index.ts",
},
@@ -56,7 +56,7 @@
},
"apps/web": {
"name": "@shumai/web-app",
"version": "0.1.0",
"version": "0.0.6",
"dependencies": {
"@shumai/agent": "workspace:*",
"@shumai/api": "workspace:*",
@@ -71,7 +71,7 @@
},
"packages/agent": {
"name": "@shumai/agent",
"version": "0.1.0",
"version": "0.0.6",
"dependencies": {
"@anthropic-ai/sandbox-runtime": "0.0.52",
"@earendil-works/pi-agent-core": "0.78.0",
@@ -95,7 +95,7 @@
},
"packages/api": {
"name": "@shumai/api",
"version": "0.1.0",
"version": "0.0.6",
"dependencies": {
"@hono/zod-validator": "0.7.6",
"@shumai/core": "workspace:*",
@@ -109,7 +109,7 @@
},
"packages/core": {
"name": "@shumai/core",
"version": "0.1.0",
"version": "0.0.6",
"dependencies": {
"@shumai/db": "workspace:*",
"@shumai/dtos": "workspace:*",
@@ -123,6 +123,7 @@
"hono": "4.12.23",
"jittered-fractional-indexing": "1.0.1",
"pino": "10.3.1",
"sharp": "0.34.5",
"tweetnacl": "1.0.3",
"ulid": "3.0.2",
"yaml": "2.9.0",
@@ -135,7 +136,7 @@
},
"packages/db": {
"name": "@shumai/db",
"version": "0.1.0",
"version": "0.0.6",
"dependencies": {
"@earendil-works/pi-agent-core": "0.78.0",
"@prisma/adapter-pg": "7.8.0",
@@ -151,14 +152,14 @@
},
"packages/dtos": {
"name": "@shumai/dtos",
"version": "0.1.0",
"version": "0.0.6",
"dependencies": {
"zod": "4.4.3",
},
},
"packages/transcode": {
"name": "@shumai/transcode",
"version": "0.1.0",
"version": "0.0.6",
"dependencies": {
"@shumai/core": "workspace:*",
"@shumai/db": "workspace:*",
@@ -167,13 +168,12 @@
"@temporalio/activity": "1.17.2",
"@temporalio/worker": "1.17.2",
"@temporalio/workflow": "1.17.2",
"sharp": "0.34.5",
"ulid": "3.0.2",
},
},
"packages/webui": {
"name": "@shumai/webui",
"version": "0.1.0",
"version": "0.0.6",
"dependencies": {
"@base-ui/react": "1.5.0",
"@dnd-kit/abstract": "0.4.0",
@@ -273,7 +273,7 @@
},
"packages/workflow-core": {
"name": "@shumai/workflow-core",
"version": "0.1.0",
"version": "0.0.6",
"dependencies": {
"@shumai/core": "workspace:*",
"@shumai/db": "workspace:*",
BIN
View File
Binary file not shown.

After

Width:  |  Height:  |  Size: 39 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 93 KiB

+83
View File
@@ -0,0 +1,83 @@
---
title: Environment Variables
description: Comprehensive configuration reference for Shumai.
---
Shumai is configured via environment variables. You can define these variables in a `.env` file at the root of your application workspace, or pass them directly to your Docker containers.
---
## Database Config
| Variable | Description | Example / Default |
| :------------- | :----------------------------------------------------------------------------------------- | :------------------------------------------------------- |
| `DATABASE_URL` | PostgreSQL connection URL. Must target a database with the `pgvector` extension installed. | `postgresql://user:pass@localhost:5432/db?schema=public` |
---
## Server Config
| Variable | Description | Example / Default |
| :---------------------- | :------------------------------------------------------------------------------------------------ | :---------------- |
| `SHUMAI_SERVER_PORT` | The port on which the Shumai web server will listen. | `3000` (default) |
| `MAX_REQUEST_BODY_SIZE` | Maximum allowed size for request bodies (e.g., for large file uploads). Default is 10GB in bytes. | `10737418240` |
---
## Authentication Config
| Variable | Description | Example / Default |
| :------------------- | :--------------------------------------------------------------------- | :--------------------------------------------------------------- |
| `BETTER_AUTH_SECRET` | A secure, random 32-byte secret used to sign session cookies and JWTs. | Generate with `openssl rand -base64 32` |
| `BETTER_AUTH_URL` | The public URL of your Hono backend API. Used for redirect flows. | `http://localhost:3000` (respects `SHUMAI_SERVER_PORT` fallback) |
---
## Storage Config
Shumai supports storing files locally or on S3-compatible endpoints.
| Variable | Description | Example / Default |
| :------------------------- | :------------------------------------------------------------------------------------------------ | :--------------------------- |
| `STORAGE_BACKEND` | Storage type. Use `local` for local file system storage, or `s3` for S3-compatible cloud storage. | `local` (default) or `s3` |
| `AWS_ENDPOINT_URL_S3` | Custom endpoint for S3 storage. Must change to public domain/IP if deploying remotely. | `http://localhost` (default) |
| `S3_BUCKET` | S3 bucket name. | `shumai` (default) |
| `S3_ACCESS_KEY_ID` | Access key credentials for S3. | `minioadmin` (default) |
| `S3_SECRET_ACCESS_KEY` | Secret key credentials for S3. | `minioadmin` (default) |
| `S3_REGION` | S3 storage region. | `auto` (default) |
| `S3_USE_SSL` | Set to `true` to use HTTPS for S3 endpoints. | `false` (default) |
| `PRESIGNED_URL_EXPIRES_IN` | Number of hours for which generated presigned download/upload links are valid. | `5` (default) |
---
## Workflow Config
Configures whether background transcoding and AI agents are executed locally or offloaded to a Temporal cluster.
| Variable | Description | Example / Default |
| :------------------------- | :------------------------------------------------------------------------------------------------- | :------------------------------ |
| `WORKFLOW_EXECUTOR` | Engine choice. Use `local` for in-process database polling, or `temporal` for distributed workers. | `local` (default) or `temporal` |
| `TEMPORAL_ADDRESS` | Host and port of the Temporal cluster gRPC service. | `localhost:7233` (default) |
| `EMBEDDING_CHUNK_DURATION` | The duration in seconds for each video chunk when generating video embeddings. | `60.0` (default) |
| `EMBEDDING_CHUNK_OVERLAP` | The duration in seconds to overlap video chunks when generating video embeddings. | `5.0` (default) |
---
## Logging Config
| Variable | Description | Example / Default |
| :----------- | :--------------------------------------------------------------------------------------------------- | :---------------- |
| `LOG_LEVEL` | Pino logging level. Options: `debug`, `info`, `warn`, `error`. | `info` (default) |
| `LOG_FORMAT` | Log format style. Use `json` for production logging, or `pretty` for human-readable terminal output. | `json` (default) |
---
## LLM API Key
API keys to AI providers on startup.
| Variable | Description | Example / Default |
| :------------------ | :---------------------------------- | :---------------- |
| `GEMINI_API_KEY` | API key for Google Gemini provider. | `AIzaSy...` |
| `OPENAI_API_KEY` | API key for OpenAI provider. | `sk-proj-...` |
| `ANTHROPIC_API_KEY` | API key for Anthropic provider. | `sk-ant-...` |
+82
View File
@@ -0,0 +1,82 @@
---
title: Deploy with Temporal
description: Configure a high-scale production deployment using Temporal and S3-compatible storage.
---
For production environments or workloads with high-volume video processing and heavy AI agent usage, running background tasks on a single server can cause bottlenecks. Shumai resolves this by offloading task queue management, retries, and background executions to **Temporal**.
This allows you to decouple workloads and run specialized worker nodes on separate, dedicated machines.
---
## How it Works
When configured for Temporal, the core Shumai web application does not process background jobs itself. Instead, it publishes tasks to a Temporal Cluster queue.
```mermaid
graph TD
UserUpload[User Uploads Video] --> WebApp[Shumai Web UI / API]
WebApp -->|Register Task| Temporal[Temporal Cluster]
Temporal -->|Queue: Transcoding| TranscodeWorker[Transcode Worker Machine]
Temporal -->|Queue: AI Agents| AgentWorker[Agent Worker Machine]
```
- **Transcoding Workers**: Can be run on GPU-accelerated or high-CPU compute instances.
- **AI Agent Workers**: Can be run on isolated, sandboxed hosts to safely execute automation scripts.
> [!IMPORTANT]
> **Storage Requirement**: Because the core application and the background workers run on separate machines in a distributed setup, **you must configure an S3-compatible storage backend** (such as AWS S3, Cloudflare R2, or MinIO). Local storage (`STORAGE_BACKEND=local`) is not supported in this mode because the workers and the web application do not share a common filesystem.
---
## Step 1: Deploy the Core Stack
On your primary host machine, spin up the database, Temporal cluster services, and the main Shumai web application:
1. **Download the Temporal Stack Configuration**:
```bash
curl -o docker-compose.yaml https://raw.githubusercontent.com/shumaiOne/shumai/main/docker-compose/temporal/docker-compose.yaml
```
2. **Configure S3 Settings**:
Open the `docker-compose.yaml` file and set your S3 credentials and endpoint environment variables under the `shumai` service.
3. **Start the Services**:
```bash
docker compose up -d
```
This exposes the web application on port `3000`, the database on `5432`, the Temporal gRPC service on `7233`, and the Temporal management dashboard on `8080`.
---
## Step 2: Configure & Run Workers
On your specialized worker machines, you will run the dedicated background worker services. Each worker requires a `.env` file referencing the database, S3 storage, and the Temporal cluster:
```env
DATABASE_URL=postgresql://shumai:shumai_password@<primary-host-ip>:5432/shumai_db?schema=public
WORKFLOW_EXECUTOR=temporal
TEMPORAL_ADDRESS=<primary-host-ip>:7233
STORAGE_BACKEND=s3
AWS_ENDPOINT_URL_S3=https://your-s3-endpoint.com
S3_BUCKET=shumai
S3_REGION=us-east-1
S3_ACCESS_KEY_ID=your-access-key-id
S3_SECRET_ACCESS_KEY=your-secret-access-key
```
### Media Transcoding Worker
Run this on your media processing nodes (requires system installations of `ffmpeg` and `ffprobe`):
```bash
npm install -g @shumai-one/shumai-transcode
shumai-transcode
```
### AI Agent Worker
Run this on your sandboxed agent execution nodes (requires sandbox dependencies like `bubblewrap` or `sandbox-exec`):
```bash
npm install -g @shumai-one/shumai-agent
shumai-agent
```
+37 -1
View File
@@ -16,7 +16,43 @@
"groups": [
{
"group": "Getting Started",
"pages": ["introduction", "getting-started/docker"]
"pages": ["getting-started/overview", "getting-started/quickstart"]
},
{
"group": "Using Shumai",
"pages": [
"using-shumai/asset-management",
"using-shumai/comments",
"using-shumai/version-stack",
"using-shumai/sort-and-search",
"using-shumai/semantic-search",
"using-shumai/team-project-management",
"using-shumai/notifications",
"using-shumai/metadata",
"using-shumai/collections",
"using-shumai/share-links"
]
},
{
"group": "Shumai Agent Settings",
"pages": [
"using-shumai/agent-providers",
"using-shumai/agents",
"using-shumai/agent-skills",
"using-shumai/agent-sandbox"
]
},
{
"group": "Media Transcoding Settings",
"pages": ["using-shumai/transcode-settings"]
},
{
"group": "Configuration",
"pages": ["configuration/env-variables"]
},
{
"group": "Temporal Deployment",
"pages": ["configuration/temporal"]
}
]
}
Binary file not shown.

After

Width:  |  Height:  |  Size: 87 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 94 KiB

-98
View File
@@ -1,98 +0,0 @@
---
title: Get Started with Docker
description: Learn how to spin up the Shumai stack using Docker Compose.
---
We support two Docker Compose setups depending on your environment and use case:
1. **Local Pre-built Stack**: A lightweight setup to run the core API and frontend using pre-built images.
2. **Temporal Stack**: A production-like environment with full background worker orchestration (Agent and Transcode workers) and Temporal services, built from your local source code.
---
## Prerequisites
Before starting, ensure you have the following installed on your machine:
- **Docker**: [Install Docker Desktop](https://docs.docker.com/get-docker/) or Docker Engine.
- **Docker Compose**: Included in Docker Desktop, or [installed separately](https://docs.docker.com/compose/install/).
- **Git**: To clone and manage the repository.
---
## 1. Local Pre-built Stack
This setup runs a PostgreSQL database (with `pgvector`) and the Shumai application container.
### Step 1: Navigate to the local docker-compose folder
From the root of your Shumai workspace, run:
```bash
cd docker-compose/local
```
### Step 2: Start the services
Spin up the containers in detached mode:
```bash
docker compose up -d
```
### Step 3: Access the application
Once all containers are running and healthy:
- Open your browser and navigate to **`http://localhost:3000`** to access the Shumai Web UI.
### Step 4: Tear down
To stop and remove the containers, run:
```bash
docker compose down
```
---
## 2. Temporal Stack (Recommended for Development)
This setup builds the containers from your local workspace source code and starts the API, workers, and Temporal scheduler.
### Step 1: Navigate to the temporal docker-compose folder
From the root of your Shumai workspace, run:
```bash
cd docker-compose/temporal
```
### Step 2: Start the services
Build the images and run the full stack:
```bash
docker compose up -d --build
```
### Step 3: Access the services
- **Web UI**: Access the Shumai application at **`http://localhost:3000`**.
- **Temporal Dashboard**: View active workflows and orchestration logs at **`http://localhost:8080`**.
### Step 4: Tear down
To stop the services and clean up volumes, run:
```bash
docker compose down -v
```
---
## Troubleshooting
### Port Conflicts
If you receive port conflict errors, ensure that port `3000` (Web UI), `5432` (PostgreSQL), `7233` (Temporal), or `8080` (Temporal UI) are not currently being used by other processes on your machine.
+37
View File
@@ -0,0 +1,37 @@
---
title: Overview
description: Learn about Shumai, a modern collaboration and media review platform.
---
## What is Shumai?
**Shumai** is a self-hosted media collaboration platform designed for creative teams, video editors, designers, and clients. It provides a centralized workspace where you can upload, organize, and review assets (video, audio, images, and documents) in real-time. With built-in AI capabilities and background transcoding, Shumai helps streamline asset organization and feedback workflows.
![Shumai Dashboard](/screenshot.webp)
---
## Core Features
- **Secure Storage**: Host your creative assets locally on your server or connect directly to any S3-compatible cloud storage provider (like AWS S3, Cloudflare R2, or MinIO).
- **Media Commenting & Drawing**: Speed up review cycles with frame-level comments and drawing tools that let you sketch directly onto video frames or images.
- **Version Control**: Stack multiple versions of the same file into a single entry to keep your workspace organized and track revision history.
- **AI-Powered Tags & Search**: Automatically tag your files on upload using Google Gemini, and find assets based on visual or conceptual meaning.
- **Custom Automations**: Build tailored background workflows using custom scripts to automate repetitive team actions.
- **Proxy Transcoding**: Automatically convert heavy camera and design files in the background into web-friendly streaming formats for smooth playback.
---
## Deployment Scales
Depending on your team size and production volume, Shumai can be configured to run in one of two modes:
### 1. Single-Server Deployment (Default)
Ideal for small teams, local development, or low-volume workflows. In this setup, the main web server, database, and background tasks (like transcoding and AI tagging) all run on a single machine.
### 2. High-Scale Distributed Deployment
Recommended for production environments with high-volume video uploads. In this setup, background tasks are managed via a centralized queue (Temporal). This allows you to offload heavy workloads to specialized worker machines (for example, putting video transcoding tasks on high-CPU/GPU machines and AI agents on separate sandboxed environments).
_Note: S3-compatible cloud storage is required for distributed deployments so that all workers can access the uploaded files._
+106
View File
@@ -0,0 +1,106 @@
---
title: Quickstart
description: Get Shumai up and running using Docker, NPM, or from source.
---
We support three installation options. For most administrators, **Docker Compose** is the fastest and easiest way to deploy Shumai.
---
## Option 1: Docker Compose (Recommended)
Docker Compose automatically configures and spins up all required services (the web application, a database, and local storage).
1. **Download the Configuration**:
```bash
curl -o docker-compose.yaml https://raw.githubusercontent.com/shumaiOne/shumai/main/docker-compose/local/docker-compose.yaml
```
2. **Start the Services**:
```bash
docker compose up -d
```
3. **Access the Dashboard**:
Open your browser and navigate to **`http://localhost:3000`** to access the Shumai Web UI.
### Production Deployments (Temporal)
If you are deploying Shumai for a production team with high-volume video uploads, we recommend setting up a distributed deployment using **Temporal** and an S3-compatible storage backend. This offloads resource-heavy video transcoding and AI agents to separate, dedicated worker nodes to keep your main dashboard fast and responsive. For setup instructions, see the [Deploy with Temporal](/configuration/temporal) guide.
---
## Option 2: Install via NPM
You can run Shumai as a global service using the Node Package Manager (NPM).
### System Prerequisites
To handle media processing and security isolation, the host machine must have the following system utilities installed:
- **FFmpeg & FFprobe**: Required for video transcoding and metadata extraction.
- _macOS_: `brew install ffmpeg`
- _Linux (Ubuntu/Debian)_: `sudo apt-get install ffmpeg`
- **Agent Sandbox Utilities**: Used to run custom automation scripts securely.
- _macOS_: `brew install ripgrep`
- _Linux (Ubuntu/Debian)_: `sudo apt-get install bubblewrap socat ripgrep`
### Step-by-Step Setup
1. **Start a Database**:
Shumai requires a PostgreSQL database with the `pgvector` extension. Start a database container:
```bash
docker run --name shumai_postgres \
-e POSTGRES_USER=shumai \
-e POSTGRES_PASSWORD=shumai_password \
-e POSTGRES_DB=shumai_db \
-p 5432:5432 \
-d pgvector/pgvector:pg18
```
2. **Install Shumai**:
```bash
npm install -g @shumai-one/shumai
```
3. **Configure Environment Variables**:
Create a new directory, navigate into it, and create a `.env` file:
```bash
mkdir shumai && cd shumai
```
Create a `.env` file with the following variables:
```env
DATABASE_URL=postgresql://shumai:shumai_password@localhost:5432/shumai_db?schema=public
BETTER_AUTH_SECRET=your-random-32-character-secret-key
STORAGE_BACKEND=local
AWS_ENDPOINT_URL_S3=http://localhost
```
4. **Start Shumai**:
```bash
shumai
```
Open your browser to `http://localhost:3000`.
---
## Option 3: Run from Source (For Developers)
If you wish to run Shumai locally to customize the platform or contribute:
1. **Clone the Repository**:
```bash
git clone https://github.com/shumaiOne/shumai.git
cd shumai
```
2. **Install Dependencies**:
```bash
bun install
```
3. **Start a Database**:
Launch a local database container (as described in the NPM section above).
4. **Run Migrations**:
Set up your `.env` file at the root, then apply database migrations:
```bash
bun run db:migrate
```
5. **Start the Dev Server**:
```bash
bun run dev
```
This spins up the web server on port `3000` with hot-reloading active.
-26
View File
@@ -1,26 +0,0 @@
---
title: Introduction
description: Welcome to Shumai, a modern AI agent and media processing platform.
---
## What is Shumai?
**Shumai** is a highly decoupled, modular platform built with a layered monorepo architecture. It integrates a React-based frontend, a fast Hono API backend, business logic services, and background workflow engines (supporting both local and Temporal-based execution).
### Key Features
- **Decoupled Monorepo Structure**: Managed via Bun Workspaces, ensuring each domain or layer is its own self-contained package.
- **Robust API & Core Service Layers**: Built on Bun and Hono, featuring strict end-to-end type safety with Zod schemas and DTOs.
- **Workflow & Activity Engine**: An orchestration engine capable of polling-based execution locally and enterprise-grade execution via Temporal.
- **Modern UI**: A responsive, clean React interface powered by Tailwind CSS and Shadcn UI.
- **Database Resilience**: Prisma ORM with PostgreSQL and `pgvector` for vector similarity and search capabilities.
## Explore the Documentation
Start by setting up Shumai on your local machine using our step-by-step guides.
<CardGroup cols={2}>
<Card title="Get Started with Docker" icon="docker" href="/getting-started/docker">
Learn how to build and spin up the Shumai stack quickly using Docker Compose.
</Card>
</CardGroup>
BIN
View File
Binary file not shown.

After

Width:  |  Height:  |  Size: 23 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 51 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 40 KiB

BIN
View File
Binary file not shown.

After

Width:  |  Height:  |  Size: 52 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 65 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 70 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 45 KiB

BIN
View File
Binary file not shown.

After

Width:  |  Height:  |  Size: 39 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 49 KiB

Binary file not shown.

After

Width:  |  Height:  |  Size: 32 KiB

+49
View File
@@ -0,0 +1,49 @@
---
title: AI Providers
description: Connect Shumai to OpenAI, Anthropic, Gemini, or custom LLM gateways.
---
When creating agents for chat or metadata autofill, you must select an underlying AI model. The available models and providers are configured globally within your team settings.
> 🔒 **Access Control:** Only **Team Owner** can manage AI providers. To update these settings, click your avatar, navigate to **Settings**, and select **Providers** from the left sidebar.
---
## Built-In Providers
Shumai ships with pre-configured configurations for major LLM providers. Because Shumai uses [Pi](https://github.com/earendil-works/pi) as its core agent framework, the supported provider and model configurations are automatically synchronized with Pi's capabilities.
Team Owner can freely edit, delete, or add new providers. All configured provider and model settings are injected directly into the Pi framework at runtime.
---
## Configuration Reference
### Provider Settings
When configuring a provider, you will need to specify the following fields:
| Field | Description | Example / Notes |
| :---------------- | :------------------------------------------------------------------------- | :------------------------------------------------------------------------------------------------------------------------------- |
| **Provider Name** | A unique, human-readable label for the provider. | `Ollama-Local`, `Enterprise-Gateway` |
| **Base URL** | The target API endpoint URL. | `http://localhost:11434/v1` |
| **API Protocol** | The streaming and payload implementation format. | Most OpenAI-compatible gateways use `openai-completions`. |
| **API Key** | The authentication token. Leave blank for unauthenticated or local setups. | Accepts a literal value (`sk-...`) or an environment variable name (`MY_API_KEY`) to load secrets securely from the host system. |
### Model Settings
Each provider can host multiple models. You can configure the following attributes per model:
| Field | Description | Example / Notes |
| :-------------------- | :---------------------------------------------------------- | :------------------------------------------------- |
| **Model ID** | The exact identifier string expected by the upstream API. | `gpt-4o`, `claude-3-5-sonnet-20240620` |
| **Display Name** | An optional, friendly label shown to users in dropdowns. | `GPT-4 Omni`, `Claude 3.5` |
| **Reasoning Support** | Toggle to enable extended thinking protocols. | Use for models with native reasoning capabilities. |
| **Context Window** | The maximum total token capacity (input + output). | |
| **Max Output Tokens** | The hard ceiling for generated response tokens per request. | |
| **Input Cost** | Cost tracking specified per million input tokens. | Optional; used to calculate usage metrics. |
| **Output Cost** | Cost tracking specified per million output tokens. | Optional; used to calculate usage metrics. |
---
![AI Providers List](../providers.webp)
+45
View File
@@ -0,0 +1,45 @@
---
title: Agent Sandbox
description: Secure execution environment and network domain restrictions for AI agent scripts.
---
Running automated scripts poses security risks. Shumai resolves this by executing all agent skills inside a secure, isolated **Sandbox** environment.
---
## Sandbox Security
The sandbox restricts the execution process from accessing the host filesystem, system processes, and internal networks:
- **Isolation**: Scripts run in a virtualized workspace container that prevents them from modifying critical system files or reading other project data.
- **Network Bridging**: External network calls from inside the sandbox are strictly monitored.
### File System Restrictions
To protect your workspace environments and sensitive data, the sandbox enforces strict file-level access controls:
- **Read Access**: The sandbox is allowed to read files within the workspace, with a strict exception for sensitive environment and certificate files. It cannot read: `['.env', '.env.*', '*.pem', '*.key']`.
- **Write Access**: The sandbox is heavily restricted and may only write to the `.pi` folder located in the current working directory (cwd). Even within this allowed directory, the sandbox is explicitly blocked from writing or creating sensitive files matching: `['.env', '.env.*', '*.pem', '*.key']`.
---
## Domain Whitelisting
By default, sandboxed scripts cannot make network requests to the internet. Team Owners must explicitly authorize external APIs by managing domain access:
### Allowed Domains
A list of domains that sandboxed scripts are permitted to access.
- Default whitelisted domains include standard package registries (like npm or python package hosts) and repository hosts (like `github.com`).
- If a script needs to call an external API (e.g. `api.slack.com`), the administrator must add it to the allowed list.
### Pending Domains
When an agent script attempts to reach a domain that is not currently authorized:
1. The network connection is blocked immediately.
2. The target domain is logged and added to the **Pending Domains** list.
3. Admins can review this list and click **Approve** to authorize the domain, or **Reject** to keep it blocked.
![Agent Sandbox Settings](../sandbox.webp)
+54
View File
@@ -0,0 +1,54 @@
---
title: Agent Skills
description: Extend your agents' capabilities with custom automation tools.
---
By default, AI models can only process text and image. In Shumai, you can extend both your **Chat Agents** and **Autofill Agents'** capabilities by registering **Skills**—custom scripts that the agent can run when solving a task.
---
## How Skills Work
Skills are actions that you define and make available to your AI agents.
- **Descriptions for the AI**: When you register a skill, the AI agent reads its metadata to understand exactly when and how it should use the tool to solve a problem.
- **Execution & Local Caching**: When an agent decides it needs to use a skill, Shumai handles the execution. Newly added skill data is synced to the database first. Upon execution, the agent downloads and caches the skill locally from the database for fast, reliable access.
---
## Adding a Skill
To install a new skill into Shumai:
1. Navigate to the **Settings** page and open the **Skills** tab.
2. Click the **Add Skill** button.
3. In the dialog box, choose your installation method:
- **Upload**: Select and upload a `.zip` file containing your skill.
- **GitHub**: Provide a valid GitHub repository path.
> **Important:** Regardless of the installation method, the root folder of the skill must contain a `SKILL.md` file.
### Updating an Existing Skill
To update a skill, simply follow the steps above and install the new version using the exact same skill name. Shumai will automatically replace the older version.
---
## Configuring Environment Variables
If your custom skill requires environment variables (e.g., API keys, external tokens) to function:
1. Click on the installed **Skill Card** in the Skills tab to open the configuration dialog.
2. Add the required environment variables.
3. _(Optional)_ Set a default value for the variable. If no default value is provided, the skill will dynamically load the required value from the system environment at runtime.
---
## Enabling Skills for Agents
**All skills are disabled by default.** Merely installing a skill does not give agents access to it. To allow an agent to use an installed skill:
1. Navigate to your specific **Agent Settings**.
2. Explicitly enable the skills you want to authorize for that particular agent.
![Skills List View](../skills.webp)
+51
View File
@@ -0,0 +1,51 @@
---
title: AI Agents
description: Create and configure specialized AI agents for chat, metadata autofill, and vector embedding.
---
Shumai uses specialized AI Agents to perform background automation, chat workflows, and semantic indexing. **Only Team Owner** can create, manage, and change agent settings.
---
## Agent Types
Shumai supports three types of agents:
### 1. Chat Agent
An interactive, conversational bot that appears in your project workspace.
- You can create as many chat agents as you want.
- You can use "@" to mention chat agents directly in file comments.
- The agent is context-aware: it can read files, search the project folder, inspect metadata, and read comment threads.
- You can add "Skills" to a chat agent, enabling it to write files, run scripts, or interact with external APIs on your behalf.
### 2. Autofill Agent
A background agent that monitors file uploads. You can create **only one** Autofill Agent.
- Once created and enabled, Shumai automatically triggers the agent when a new file is uploaded, provided the file type is supported (image or video).
- It analyzes the file's content (multimodal images, text descriptions) and populates any metadata fields flagged for "AI Autofill".
### 3. Embedding Agent
A specialized utility agent that processes assets to generate search indexes. You can create **only one** Embedding Agent.
- Once created and enabled, Shumai automatically runs embedding generation for new file uploads if the file type is supported (image or video).
- These search indexes are saved to power Shumai's natural language semantic search capabilities.
- **Note:** The Embedding Agent is hardcoded to use `google-embedding-2` as the LLM model. Because of this, you cannot choose a provider or model. Make sure you set the `GEMINI_API_KEY` environment variable to make it work.
- **Chunk Duration:** You can configure the duration of chunks used for long video indexing via the `EMBEDDING_CHUNK_DURATION` environment variable (defaults to 60 seconds).
---
## Configuring an Agent
To edit or create an agent, navigate to **Team Settings** > **Agents** (Note: Only Team Owners can change these settings):
- **Avatar**: Choose an avatar for the agent by picking from the predefined avatars.
- **Enabled / Disabled Toggle**: Stop or start the agent. If the embedding agent is disabled, semantic search indexing is paused.
- **System Prompt / Soul**: The core instructions that define the agent's behavior, guidelines, formatting requirements, and operational limits.
- **Model Provider**: Select which registered provider and model the agent should use for execution. _(Note: This option is disabled for Embedding Agents)._
- **Assigned Skills**: Select which custom skills this agent is permitted to execute.
![Agent Config Panel](../agents.webp)
+54
View File
@@ -0,0 +1,54 @@
---
title: Asset Management
description: Learn how to upload, organize, delete, and recover assets in your workspace.
---
Shumai makes it easy to bring creative media into your workspace, organize them in folders, and manage their lifecycle from upload to deletion.
---
## Uploading Assets
You can upload files and folders into Shumai using three methods:
1. **Drag and Drop**: Drag files or entire folder directories directly from your computer's file explorer and drop them anywhere on the Shumai dashboard interface.
2. **Upload Buttons**: Click the **Upload File** or **Upload Folder** buttons in the toolbar.
3. **Copy and Paste**: Copy an image or file from another website, application, or your clipboard, and paste it directly (`Ctrl+V` or `Cmd+V`) into the Shumai browser window to upload it instantly.
![Asset Management - Drag and Drop Upload](../dragdrop.webp)
---
## Folder Organization
When you upload an entire folder, Shumai replicates your folder hierarchy exactly inside the project.
- All subfolders are created automatically.
- Nested files remain in their respective subfolders.
- There is no limit to the depth of nested folders you can upload.
---
## Upload and Processing States
Once an upload starts, your assets progress through the following states:
- **Uploading**: The file is transferring to your storage server. The UI displays an active progress percentage.
- **Processing**: The file has successfully uploaded. Shumai is now extracting media parameters (such as resolution, video length, and audio codecs) and creating a web-friendly streaming preview (proxy).
- **Ready**: Processing is complete. The asset is fully visible, playable, and ready for comments, metadata tagging, or search.
---
## Deleting and Restoring Assets
If you no longer need an asset, you can delete it to clean up your workspace:
### The Trash Bin (Recently Deleted)
- When you delete a file or folder, it is moved to the **Recently Deleted** tab in your project settings.
- Deleted assets are kept in the Trash for **30 days**. During this window, you can select any asset and click **Recover** to restore it to its original location.
- If you want to free up space immediately, you can open the Recently Deleted tab and click **Empty Trash** to permanently purge the files.
### Permanent Removal
After **30 days** in the Trash, assets are automatically and permanently purged. Once purged, the records are removed from the database, and the files are deleted from your storage backend (local file system or S3 bucket) and cannot be recovered.
+55
View File
@@ -0,0 +1,55 @@
---
title: Collections
description: Build dynamic virtual folders using metadata filters and search criteria.
---
Instead of copying files into static folders, Shumai uses **Collections**—dynamic, query-based virtual views. A collection automatically groups files that match specific criteria in real-time.
---
## What is a Collection?
A **Collection** does not duplicate files on disk. Instead, it runs a pre-defined filter query against your project assets database. Whenever a new file is uploaded or metadata is updated to match the collection's query, the file automatically appears inside the collection.
---
## Creating a Collection
You can create collections in two ways:
- **From the Sidebar**: Click the `+` icon next to **Collections** in the project sidebar to open the query builder, configure your search rules, and save it.
- **From the Search Dialog**: When searching and filtering files within the Search Dialog, you can save your active search criteria by clicking the **Save as collection** button at the bottom of the dialog. This will create a dynamic collection that mirrors your search filters.
---
## Setting Up Collection Filters
When creating or editing a collection, you specify a name and set up a filter query in the builder interface:
### 1. Source Folder
Specify the starting folder directory. The collection will scan this folder and all its nested subfolders recursively.
### 2. Operators and Conditions
- **Logical Operators**: Combine conditions using **AND** (all conditions must be met) or **OR** (at least one condition must be met).
- **Fields**: Filter on standard fields (name, type, size) or custom metadata fields (such as "Status" or "Approval").
- **Comparison Rules**:
- _Equals / Does Not Equal_: Match exact values.
- _Contains / Does Not Contain_: Match parts of words.
- _Greater Than / Less Than_: Set up numeric or date ranges (e.g., date uploaded is after yesterday).
- _Is Empty / Is Not Empty_: Check if fields have values.
### 3. Collection Default Sorting
Define a custom sorting rule specific to the collection (e.g., sort by the custom field "Publication Date" ascending) to override the default project sort order.
---
## Example Use Cases
- **Review Dashboard**: A collection named "Pending Review" that displays assets where `Status` equals `In Review`.
- **Daily Deliverables**: A collection filtering files uploaded in the last 24 hours.
- **Approved Assets**: A collection filtering assets matching `Tag` contains `marketing` AND `Approved` is `Yes`.
![Collection Creation](../collection.webp)
+51
View File
@@ -0,0 +1,51 @@
---
title: Comments & Annotations
description: Collaborate on media assets using drawings and timestamped comments.
---
Shumai provides a review workspace designed for video, audio, and image feedback. Instead of generic text comments, you can drop timestamped feedback and sketch directly onto the media.
---
## Timestamped Comments
When reviewing video or audio files, comments can be anchored to specific seconds on the timeline:
1. Play or scrub the video player to the exact frame or timestamp you want to review.
2. Type your feedback in the comment box.
3. Click **Submit**. Your comment will be saved with a timestamp marker shown on the video timeline.
4. Collaborators can click your comment card in the sidebar to jump the video player to the exact moment you commented on.
![Timestamped Comments](../timestamp-comment.webp)
---
## Draw Annotations (Canvas)
To give clear visual feedback, you can draw directly onto image or video frames:
1. Pause the video on the frame you wish to annotate.
2. Select a drawing tool (brush, rectangle, circle, arrow) from the annotation toolbar overlaying the player.
3. Draw directly on the video screen.
4. Type your comment and click **Submit**. Your drawing is saved along with your comment.
5. When other users select your comment, your drawings will overlay on the player screen automatically at that frame.
![Draw Annotations](../draw_annotation.webp)
---
## Comment Threads (Replies)
Keep discussions organized by replying directly to feedback:
- Click **Reply** on any comment.
- Replies are nested under the parent comment card, making it easy to track resolution conversations.
---
## Attachments
You can attach reference documents, style guides, or other assets to your comments:
- Click the **Paperclip** icon inside the comment input box.
- Choose a file to upload. It will appear as a downloadable link directly on your comment card.
+68
View File
@@ -0,0 +1,68 @@
---
title: Asset Metadata
description: Customize your assets with dynamic, scoped metadata fields and AI-driven autofill features.
---
Standard file systems limit you to basic attributes like file name and modification date. Shumai lets you create structured metadata fields tailored to your creative pipeline—from tracking review statuses to capturing technical specs automatically.
---
## Metadata Scopes
Fields are grouped and displayed in the sidebar inspector by their scope:
- **Essentials (System Scope)**: Core fields built into Shumai that appear across all teams and projects. Many of these (like resolution, duration, and codec) are automatically extracted during file upload.
- **Team Attributes (Team Scope)**: Custom fields defined at the team level, which are inherited by all projects under that team. Excellent for enforcing company-wide metadata guidelines.
- **Custom Fields (Project Scope)**: Unique metadata fields defined specifically for a single project workspace.
---
## Custom Field Types
When building or editing your metadata schema, you can choose from these custom types:
- **Text**: A single-line input field for simple strings, codes, or labels.
- **Long Text**: A paragraph input area for detailed descriptions, summaries, or notes.
- **Dropdown (Select)**: A single-choice dropdown menu where options can be styled with custom background colors (e.g., Green for Approved, Yellow for In Progress).
- **Tags (Multi-Select)**: A multi-choice tag field allowing you to assign multiple custom colored labels to a single asset.
- **Star Rating**: An interactive star rating scale (supports configurations from 2 to 6 stars) for scoring and ranking.
- **Number**: A numeric field supporting custom decimal precision (from 0 to 3 decimal places) for values like frame rates, counts, or scores.
- **Checkbox (Toggle)**: A simple Yes/No switch for binary indicators (e.g., "VFX Required").
- **Date**: A calendar date selector with support for custom display formats (Local, ISO, USA, Euro, Friendly) and optional time tracking.
![Asset Metadata Fields](../metafields.webp)
---
## Default Built-in Fields
Every project comes pre-configured with several essential fields populated automatically during upload or status changes:
- **Technical Specs**: File Size, Format, File Type (Video, Audio, Image, Document, File), Resolution Width/Height, Video/Audio Codecs, Duration, Frame Rate, and Audio Sample Rates.
- **Collaborative Fields**:
- **Status**: A dropdown select field containing `Needs Review` (Red), `In Progress` (Yellow), and `Approved` (Green).
- **Rating**: A 5-star rating scale for quick review evaluations.
- **Keywords**: A tags (multi-select) field to categorize assets.
---
## Field Settings & Controls
### Read-Only Fields
Marking a custom field as **Read-Only** prevents team members from editing its value directly. This is ideal for fields meant to be populated programmatically or by automated integrations.
### AI Autofill
When you enable **AI Autofill** on a metadata field, Shumai automatically analyzes uploaded assets using your team's default Gemini AI model. The AI will populate the field with extracted insights:
- For text/long text: AI generates summaries, descriptions, or shot logs.
- For select/multi-select: AI automatically categorizes themes or applies matching tags.
- For numbers: AI can output confidence scores or other metrics.
### Visibility & Ordering
In the metadata inspector:
- **Visibility**: Toggle fields on or off to clean up the sidebar and only show what is relevant to your current review process.
- **Ordering**: Drag and drop fields in the builder to arrange their display order within the Custom Fields section.
+46
View File
@@ -0,0 +1,46 @@
---
title: Notifications
description: Stay updated on comment activity, uploads, and project updates.
---
Shumai includes a built-in notification system to keep you informed of team activities, file statuses, and review comments. You can fully customize which events trigger notifications, check updates from your inbox, or toggle notifications for specific projects.
---
## Notification Settings
You can customize your notification preferences for each team to avoid unnecessary noise. To manage your notification events:
1. Click on **Team Settings** in the sidebar.
2. Select the **Notifications** tab.
3. Toggle the switches to enable or disable notifications for specific events:
### Comments & Replies
- **General Comments**: Notify me when someone comments on an asset I collaborate on.
- **Comment Replies**: Notify me when someone replies to my comment threads.
- **@Mentions**: Notify me when someone tags me (using `@username`) in any comment.
### Assets & Statuses
- **Your Uploads**: Notify me when files I upload finish transcoding and are ready for review.
- **Other Uploads**: Notify me when other team members upload new assets.
- **Status Updates**: Notify me when someone changes an asset's status (e.g., from "In Progress" to "Approved").
---
## Project-Level Notifications
If you want to mute a specific project entirely (except for direct `@mentions` which are always delivered):
1. Navigate to the **Project** dashboard.
2. Open **Project Settings**.
3. Toggle **Enable Notifications** on or off.
---
## The Notification Inbox
- The **Bell** icon in the navigation bar displays a red badge when you have unread notifications.
- Click the bell to open your notification feed panel.
- Click any notification card to mark it as read and jump directly to the relevant file and comment thread.
+41
View File
@@ -0,0 +1,41 @@
---
title: Semantic Search
description: Find files based on visual and textual meaning using AI vector similarity.
---
Traditional search relies on matching exact characters (e.g., searching "dog" will not find files named "golden_retriever.jpg"). Shumai resolves this with **Semantic Search**, which indexes the conceptual meaning of your files.
---
## How Semantic Search Works
Semantic search translates text queries and file contents into conceptual profiles. Similar ideas and visuals are mapped near each other.
1. **Index Generation**: When you upload an asset, Shumai's background agent scans the file's visual elements, text contents, or descriptions, translating them into a concept profile.
- **Video Processing**: Shumai uses **gemini-embedding-2** for video indexing.
- **Chunking**: For long videos, Shumai breaks them into chunks to ensure high-quality indexing. By default, chunks are **60 seconds** long with a **5-second overlap** to ensure no context is lost at the boundaries.
- **Sampling**: The model processes a maximum of 32 frames per chunk: short chunks (≤32s) are sampled at 1 fps, while longer chunks are uniformly sampled to 32 frames.
- **Optimization**: To optimize processing, Shumai automatically transcodes each chunk to **1 FPS** and **removes the audio track** before sending it to the embedding agent.
- **Configuration**: Administrators can customize these settings using the `EMBEDDING_CHUNK_DURATION` and `EMBEDDING_CHUNK_OVERLAP` environment variables.
2. **Concept Matching**: When you search with the "Semantic" toggle enabled, Shumai matches the conceptual meaning of your query against your assets.
3. **Result**: Searching "sunset at the beach" will successfully find a video file of ocean waves at dusk, even if the filename is `IMG_4029.mp4`.
### Understanding Search Results
Semantic search works differently than traditional keyword filtering:
- **Sort vs. Filter**: Semantic search acts more like a **sort** than a strict filter. While the most relevant results will appear at the top, less related content may still be present further down the list, ranked lower. If you need to exclude specific types of files, use the **Filter** panel in conjunction with semantic search.
- **Video Chunks**: Because Shumai indexes videos in chunks, a single long video may appear multiple times in your search results. Each result corresponds to a specific segment of the video where the match was found.
- **Match Timestamps**: When a search result comes from a video chunk, Shumai displays the matching start and end time. Clicking on these results will open the video and automatically jump to the relevant timestamp.
---
## Activating Semantic Search
To use semantic search:
1. Verify that your administrator has enabled the **Embedding Agent** under **Team Settings** > **Agents**.
2. Once active, Shumai will automatically begin indexing new uploads.
3. Toggle the **Semantic** button next to the search input in the file explorer to search by conceptual meaning.
![Semantic Search Toggle](../semantic-search.webp)
+57
View File
@@ -0,0 +1,57 @@
---
title: Share Links
description: Share files and folders securely with external clients and stakeholders.
---
Shumai allows you to collaborate with external stakeholders who do not have accounts in your workspace. Using **Share Links**, you can grant view and comment access to specific folders or assets.
---
## Creating a Share Link
1. Select the files or folders you want to share, then right-click and choose **Create New Share Link**.
2. Click the newly created **Share Link** in the project sidebar to open its settings page.
3. Configure the share link in the right sidebar. You can:
- Add password protection
- Set an expiration date
- Configure which metadata fields are visible to recipients
4. Click **Copy Link** at the bottom of the sidebar, then send the link to your client or stakeholder.
![Share Link Configuration](../sharelink.webp)
---
## Adding Assets to an Existing Share Link
Once a share link is created, you can easily add more files or folders to it without generating a new link:
1. Select the files or folders you want to add from your file browser.
2. Drag and drop them directly onto the share link item in the project sidebar.
3. The share link will immediately update to include the newly added assets.
---
## Security and Settings
### Expiration Date
Set an optional date and time after which the link becomes inactive. Once the date passes, the link returns a `404 Not Found` page to anyone attempting to access it.
### Password Protection
Secure sensitive work by adding a password. External users visiting the link will be prompted to enter the correct password before they can view the assets.
### Default Sort Order
Set a default sorting rule for the shared page (e.g. sorting by name or version) to ensure the client views the files in the intended order.
### Column Visibility
To protect proprietary information, you can configure exactly which metadata columns are visible to external users. For example:
- **Visible**: `Asset Name`, `Resolution`, `Download Link`.
- **Hidden**: `Internal Budget Code`, `Assignee Notes`, `Cost`.
### Disabling a Link
If you need to revoke access immediately, navigate to the **Share Links** tab under Project settings and toggle the **Disabled** switch. You can re-enable the link at any time.
+39
View File
@@ -0,0 +1,39 @@
---
title: Sort & Search
description: Find assets using the search dialog and organize them with custom sort orders.
---
As your projects grow, finding the right media files quickly is critical. Shumai provides a powerful search dialog and flexible sorting controls.
---
## Searching Assets
Searching is performed inside a dedicated **Search Dialog** that you can open by clicking the search icon in the toolbar.
![Search Dialog](../search-dialog.webp)
### Search Dialog Features
- **Keyword Matches**: Type a keyword in the text input to instantly filter file and folder names.
- **Advanced Filtering**: Click to expand the filters panel and build multi-conditional queries (e.g. searching for files where the custom field "Status" equals "In Progress" AND the file type is "Video").
- **Semantic Search Toggle**: If enabled by your team administrator, toggle "Semantic" on to search by conceptual meaning (e.g. searching "sunny beach" to find clips of oceans even if the filenames do not contain those words).
- **Save as Collection**: Easily convert your active search criteria into a dynamic folder. Click **Save as Collection** inside the dialog to create the collection and navigate to it instantly.
- **Infinite Scroll Results**: Search results load dynamically in real-time as you type or adjust filters.
---
## Sorting and Reordering Assets
You can change the sorting order of the asset list using the sort dropdown or by clicking metadata column headers:
- **Custom Sort (Default)**: By default, the project file explorer is set to **Custom** sorting. In this mode, you can manually drag and drop files and folders to reorder them exactly as you wish.
- **Standard Sort Options**: You can also sort files by:
- _Newest Upload_: Displays recently uploaded assets first.
- _Name_: Sorts alphabetically (A-Z or Z-A).
- _Size_: Sorts by file size.
- _Type_: Groups folders together at the top, followed by version stacks and individual files.
- _Custom Columns_: Sorts by dynamic metadata fields (e.g., sorting by "Due Date" or "Priority").
> [!NOTE]
> **Drag and Drop Reordering Limit**: Drag-and-drop manual reordering is **only** available when the sort order is set to **Custom**. If you select any other sort option (such as Name or Size), manual reordering is disabled.
@@ -0,0 +1,58 @@
---
title: Team & Project Management
description: Organize your workspaces, define roles, and manage members.
---
Shumai uses a two-tier organizational hierarchy—**Teams** and **Projects**—to manage file access and collaboration.
---
## Member Visibility & Scopes
Members added to a team belong to one of two permission scopes:
- **Team Scope**: Team-scoped members can view all projects inside the team by default.
- **Project Scope**: Project-scoped members can only see and access the specific projects they have been explicitly granted permission to join.
---
## Member Roles
Shumai uses permission roles to control what actions team and project members can perform:
- **Owner**: Full administrative control. Team Owners can modify team settings, skills, and AI providers. Project Owners can edit project configurations, manage members, and generate invites.
- **Editor**: Can upload files, edit custom metadata, create folders, and add comments or canvas annotations.
- **Reviewer**: Read-only access. Can view media assets, toggle versions, and leave text comments or canvas drawings, but cannot upload/delete files or modify metadata.
---
## Project Role Overrides
Administrators can add existing team members to specific projects to override their permissions:
- Project-level roles take higher priority than team-level roles within that project.
- _Example_: If Peter is a team member with an **Editor** role, an administrator can add Peter to _Project A_ with a **Reviewer** role. Peter will act as a Reviewer inside Project A, but maintain his Editor role across the rest of the team's projects.
---
## Managing Members
Member management is handled using a unified **Members Dialog** popup:
### 1. Managing Team Members
- **Access**: Open your **Team Dashboard** and click the members list or avatar group in the header.
- **Settings**:
- Change member roles between Editor and Reviewer.
- Remove members from the team.
- **Invite New Members**: Select a role (Editor or Reviewer) and click **Generate Link**. Copy the signup URL to invite new users to the team.
### 2. Managing Project Members
- **Access**: Open your project **File Browser** and click the members avatars or button in the top toolbar.
- **Settings**:
- Change project member roles or remove them from the project.
- **Add Team Members**: The dialog displays a list of active team members who are not yet added to this project. Select their project role (which will override their team role inside this project) and click **Add** to add them instantly.
- **Invite External Collaborators**: Select a role (Editor or Reviewer) and click **Generate Link** to invite new users directly into this specific project.
![Members Dialog](../members.webp)
+36
View File
@@ -0,0 +1,36 @@
---
title: Media Transcoding Settings
description: Configure video transcode strategies and preview proxy generation.
---
High-resolution camera formats and raw media files cannot be rendered natively in web browsers. Shumai transcodes these assets into web-friendly streaming formats in the background.
Regardless of the selected transcoding strategy, Shumai's pipeline will **always** generate at least one lightweight `180p` proxy version of the raw file to ensure fast scrubbing, thumbnail generation, and instant previews.
---
## Video Transcode Strategies
Team Owners can select how the transcoding engine handles uploaded files. Shumai evaluates the original file's dimensions against its standard target resolutions (2160p, 1080p, 720p, 540p, and 360p) to determine the output.
### 1. Best Match
- **Behavior**: Shumai checks the longest dimension (width or height) of the uploaded file. It then finds the highest standard resolution that is equal to or smaller than the original file, preventing unnecessary upscaling. If the original video is exceptionally small, it defaults to a minimum target of `360p`.
- **Output**: Generates exactly two web-optimized files: the universal `180p` proxy and the single calculated "best match" resolution.
- **Use Case**: This is the recommended option. It ensures web compatibility without wasting processing time and storage space on redundant lower-resolution versions.
### 2. Transcode All
- **Behavior**: Shumai calculates the "best match" resolution exactly as it does in the previous strategy. However, instead of stopping at one version, it processes the video into the best match _and_ every standard resolution smaller than it.
- **Output**: Generates a full cascade of files. For example, if the best match for an upload is `1080p`, Shumai will generate the `180p` proxy, `1080p`, `720p`, `540p`, and `360p` versions.
- **Use Case**: Ideal for environments requiring adaptive bitrate streaming, ensuring consistent playback performance across all client devices regardless of varying network conditions.
![Transcode Settings](../transcode.webp)
---
## Processing Architecture
Transcoding is resource-intensive. Depending on your scaling needs, your administrator can run the transcode engine on the main server (Single-Server mode) or delegate tasks to specialized transcoding worker machines (Temporal Worker mode).
Note: In all modes, the processing machine must have the standard system media utilities (`ffmpeg` and `ffprobe`) installed.
+44
View File
@@ -0,0 +1,44 @@
---
title: File Version Stack
description: Group multiple revisions of an asset under a single version-controlled stack.
---
When collaborating on creative assets, files naturally go through multiple iterations (e.g. `draft_v1.mp4`, `draft_v2.mp4`, `draft_final.mp4`). To prevent workspace clutter, Shumai allows you to bundle these revisions into a **Version Stack**.
---
## What is a Version Stack?
A **Version Stack** groups multiple revisions of the same file into a single entry. Instead of cluttering your folder with several similar files, Shumai collapses them into one item with a version badge (e.g., "V3") and a dropdown list of older versions.
![Version Stack Selector](../versionstack.webp)
---
## Creating & Adding to a Version Stack
You can create a version stack or add new versions to an existing stack using two interactive methods:
- **Drag and Drop**:
- _To create a stack_: Drag one standalone file directly over another standalone file in your file explorer.
- _To add a version_: Drag a standalone file directly onto an existing version stack.
- **Context Menu (Right-Click)**:
- Right-click any standalone file or existing version stack.
- Choose **Create New Version** from the menu.
- Select the revision file from your computer to upload it directly into the stack.
---
## Managing Versions
### Active Version
The file at the top of the stack is the **Active Version**. This is the version displayed on your project dashboard, search results, and metadata panels.
### Reordering and Reverting Versions
You can promote a past version or reorder your stack:
- Open the version stack details sidebar.
- Drag and drop the versions in the list to change their order. The version you drag to the top automatically becomes the active version.
- To remove a file from the stack, drag it out or select **Unstack**. This moves the file back to the parent folder as a standalone file.
Binary file not shown.

After

Width:  |  Height:  |  Size: 83 KiB

+1 -1
View File
@@ -1,6 +1,6 @@
{
"name": "shumai",
"version": "0.0.6",
"version": "0.0.8",
"private": true,
"type": "module",
"repository": {
+1 -1
View File
@@ -1,6 +1,6 @@
{
"name": "@shumai/agent",
"version": "0.0.6",
"version": "0.0.8",
"private": true,
"type": "module",
"main": "./src/index.ts",
+1 -1
View File
@@ -306,7 +306,7 @@ describe('Agent Database Activities Integration', () => {
teamId: team.id,
name: 'openai',
config: {
api: 'openai',
api: 'openai-completions',
apiKey: 'mock-key',
},
},
+72 -30
View File
@@ -1,16 +1,16 @@
import { describe, it, expect, vi, beforeEach } from 'vitest'
import {
generateEmbeddingActivity,
generateTextEmbeddingActivity,
extractAiMetadataActivity,
getEmbeddingContextActivity,
saveAssetEmbeddingsActivity,
generateImageEmbeddingActivity,
generateVideoChunkEmbeddingActivity,
} from './ai'
import { s3Service } from '@shumai/core/src/s3/s3'
import { transcodeService } from '@shumai/transcode'
import { transcodeService } from '@shumai/core'
import { prisma, AssetType, AssetStatus } from '@shumai/db'
import { setupTestDbHooks } from '@shumai/db/test'
import * as fs from 'fs'
vi.mock('fs', async (importOriginal) => {
// eslint-disable-next-line @typescript-eslint/no-explicit-any -- importOriginal returns the raw fs module, typed dynamically as any here
@@ -38,7 +38,7 @@ vi.mock('@shumai/core/src/s3/s3', () => ({
},
}))
vi.mock('@shumai/transcode', () => ({
vi.mock('@shumai/core/src/transcode/transcode', () => ({
transcodeService: {
extractVideoFrames: vi.fn(),
},
@@ -93,19 +93,13 @@ describe('AI Activities Unit Tests', () => {
contentType: 'image/png',
} as unknown as { buffer: Buffer; contentType: string })
const context = {
agent: { config: { provider: 'google', model: 'gemini' } },
asset: { id: 'a1', mediaType: 'image/png', storageKey: { key: 'test.png' } },
}
const res = await generateEmbeddingActivity({
const res = await generateImageEmbeddingActivity({
teamId: 't1',
assetId: 'a1',
context,
assetKey: 'test.png',
mediaType: 'image/png',
})
expect(res.embeddings.length).toBe(1)
expect(res.embeddings[0].embedding).toEqual([0.1, 0.2, 0.3])
expect(res.embedding).toEqual([0.1, 0.2, 0.3])
expect(s3Service.getObject).toHaveBeenCalledWith('shumai', 'test.png')
})
@@ -115,25 +109,16 @@ describe('AI Activities Unit Tests', () => {
contentType: 'video/mp4',
} as unknown as { buffer: Buffer; contentType: string })
const context = {
agent: { config: { provider: 'google', model: 'gemini' } },
asset: { id: 'a2', mediaType: 'video/mp4', storageKey: { key: 'test.mp4' } },
}
const res = await generateEmbeddingActivity({
const res = await generateVideoChunkEmbeddingActivity({
teamId: 't1',
assetId: 'a2',
context,
chunkKey: 'files/a1/tmp-embedding-chunks/chunk-0-60.mp4',
})
// Duration is 120s, chunks are 60s, so it should generate 2 chunks (0-60, 60-120)
expect(res.embeddings.length).toBe(2)
expect(res.embeddings[0]).toEqual({ embedding: [0.1, 0.2, 0.3], startTime: 0, endTime: 60 })
expect(res.embeddings[1]).toEqual({ embedding: [0.1, 0.2, 0.3], startTime: 60, endTime: 120 })
expect(vi.mocked(fs.writeFileSync)).toHaveBeenCalled()
expect(vi.mocked(fs.readFileSync)).toHaveBeenCalled()
expect(vi.mocked(fs.unlinkSync)).toHaveBeenCalled()
expect(res.embedding).toEqual([0.1, 0.2, 0.3])
expect(s3Service.getObject).toHaveBeenCalledWith(
'shumai',
'files/a1/tmp-embedding-chunks/chunk-0-60.mp4',
)
})
it('should generate text embedding successfully', async () => {
@@ -291,6 +276,63 @@ describe('AI Database Activities Integration', () => {
expect(ctx.asset.id).toBe(asset.id)
expect(ctx.asset.mediaType).toBe('video/mp4')
})
it('should parse valid positive chunk duration and fallback on invalid or negative ones', async () => {
// Create embedding agent if not already present
const existingAgent = await prisma.agent.findUnique({ where: { id: user.id } })
if (!existingAgent) {
await prisma.agent.create({
data: {
id: user.id,
teamId: team.id,
type: 'embedding',
enabled: true,
config: {
provider: 'openai',
model: 'gpt-4',
},
},
})
}
// Test default/fallback
delete process.env.EMBEDDING_CHUNK_DURATION
let ctx = await getEmbeddingContextActivity({
teamId: team.id,
assetId: asset.id,
})
expect(ctx.chunkDuration).toBe(60.0)
expect(ctx.chunkOverlap).toBe(5.0)
// Test valid positive
process.env.EMBEDDING_CHUNK_DURATION = '45.5'
process.env.EMBEDDING_CHUNK_OVERLAP = '10.0'
ctx = await getEmbeddingContextActivity({
teamId: team.id,
assetId: asset.id,
})
expect(ctx.chunkDuration).toBe(45.5)
expect(ctx.chunkOverlap).toBe(10.0)
// Test negative/invalid for overlap
process.env.EMBEDDING_CHUNK_OVERLAP = '-5.0'
ctx = await getEmbeddingContextActivity({
teamId: team.id,
assetId: asset.id,
})
expect(ctx.chunkOverlap).toBe(5.0)
process.env.EMBEDDING_CHUNK_OVERLAP = 'invalid'
ctx = await getEmbeddingContextActivity({
teamId: team.id,
assetId: asset.id,
})
expect(ctx.chunkOverlap).toBe(5.0)
// Cleanup env
delete process.env.EMBEDDING_CHUNK_DURATION
delete process.env.EMBEDDING_CHUNK_OVERLAP
})
})
describe('saveAssetEmbeddingsActivity', () => {
+100 -1
View File
@@ -1,6 +1,6 @@
import { GoogleGenAI } from '@google/genai'
import { s3Service } from '@shumai/core/src/s3/s3'
import { transcodeService } from '@shumai/transcode'
import { transcodeService } from '@shumai/core'
import { execFile } from 'child_process'
import * as fs from 'fs'
import * as os from 'os'
@@ -140,6 +140,8 @@ export async function generateEmbeddingActivity(params: GenerateEmbeddingParams)
const chunkTmp = path.join(os.tmpdir(), `video-chunk-${Date.now()}.mp4`)
await execFileAsync('ffmpeg', [
'-y',
'-loglevel',
'warning',
'-i',
tmpFile,
'-ss',
@@ -309,9 +311,18 @@ export async function getEmbeddingContextActivity(params: { teamId: string; asse
throw ApplicationFailure.create({ message: 'asset has no media type', nonRetryable: true })
}
const parsedDuration = parseFloat(process.env.EMBEDDING_CHUNK_DURATION || '')
const chunkDuration =
Number.isFinite(parsedDuration) && parsedDuration > 0 ? parsedDuration : 60.0
const parsedOverlap = parseFloat(process.env.EMBEDDING_CHUNK_OVERLAP || '')
const chunkOverlap = Number.isFinite(parsedOverlap) && parsedOverlap >= 0 ? parsedOverlap : 5.0
return {
agent,
asset,
chunkDuration,
chunkOverlap,
}
}
@@ -338,3 +349,91 @@ export async function saveAssetEmbeddingsActivity(params: {
}
}
}
export interface GenerateImageEmbeddingParams {
teamId: string
assetKey: string
mediaType: string
}
export async function generateImageEmbeddingActivity(
params: GenerateImageEmbeddingParams,
): Promise<{
embedding: number[]
usage: Usage
}> {
const apiKey = process.env.GEMINI_API_KEY
if (!apiKey) {
throw ApplicationFailure.create({
message: 'GEMINI_API_KEY environment variable is not configured',
nonRetryable: true,
})
}
const resolvedApiKey = process.env[apiKey] || apiKey
const ai = new GoogleGenAI({ apiKey: resolvedApiKey })
const usage: Usage = {
model: 'gemini-embedding-2',
inputTokens: 0,
outputTokens: 0,
}
const { buffer: data } = await s3Service.getObject(
process.env.S3_BUCKET || 'shumai',
params.assetKey,
)
const embVec = await generateMultimodalEmbedding(ai, data, params.mediaType)
return {
embedding: embVec,
usage,
}
}
export interface GenerateVideoChunkEmbeddingParams {
teamId: string
chunkKey: string
}
export async function generateVideoChunkEmbeddingActivity(
params: GenerateVideoChunkEmbeddingParams,
): Promise<{
embedding: number[]
usage: Usage
}> {
const apiKey = process.env.GEMINI_API_KEY
if (!apiKey) {
throw ApplicationFailure.create({
message: 'GEMINI_API_KEY environment variable is not configured',
nonRetryable: true,
})
}
const resolvedApiKey = process.env[apiKey] || apiKey
const ai = new GoogleGenAI({ apiKey: resolvedApiKey })
const usage: Usage = {
model: 'gemini-embedding-2',
inputTokens: 0,
outputTokens: 0,
}
try {
const { buffer: data } = await s3Service.getObject(
process.env.S3_BUCKET || 'shumai',
params.chunkKey,
)
const embVec = await generateMultimodalEmbedding(ai, data, 'video/mp4')
return {
embedding: embVec,
usage,
}
} catch (err) {
throw ApplicationFailure.create({
message: `Failed to generate video chunk embedding for key ${params.chunkKey}: ${err instanceof Error ? err.message : String(err)}`,
nonRetryable: false,
})
}
}
+2
View File
@@ -380,6 +380,8 @@ export {
generateTextEmbeddingActivity,
type ExtractAiMetadataParams,
extractAiMetadataActivity,
generateImageEmbeddingActivity,
generateVideoChunkEmbeddingActivity,
} from './activities/ai'
export * from './workflows/agent-autofill'
export * from './workflows/agent-chat'
@@ -29,8 +29,11 @@ describe('Agent Embedding Workflow', () => {
getEmbeddingContextActivity: Object.assign(vi.fn(), {
_activityName: 'getEmbeddingContextActivity',
}),
generateEmbeddingActivity: Object.assign(vi.fn(), {
_activityName: 'generateEmbeddingActivity',
generateImageEmbeddingActivity: Object.assign(vi.fn(), {
_activityName: 'generateImageEmbeddingActivity',
}),
generateVideoChunkEmbeddingActivity: Object.assign(vi.fn(), {
_activityName: 'generateVideoChunkEmbeddingActivity',
}),
saveAssetEmbeddingsActivity: Object.assign(vi.fn(), {
_activityName: 'saveAssetEmbeddingsActivity',
@@ -41,20 +44,58 @@ describe('Agent Embedding Workflow', () => {
getAgentWorkerQueueActivity: Object.assign(vi.fn(), {
_activityName: 'getAgentWorkerQueueActivity',
}),
getTranscodeWorkerQueueActivity: Object.assign(vi.fn(), {
_activityName: 'getTranscodeWorkerQueueActivity',
}),
downloadMediaToTmpActivity: Object.assign(vi.fn(), {
_activityName: 'downloadMediaToTmpActivity',
}),
cleanupTmpDirActivity: Object.assign(vi.fn(), {
_activityName: 'cleanupTmpDirActivity',
}),
transcodeVideoChunkActivity: Object.assign(vi.fn(), {
_activityName: 'transcodeVideoChunkActivity',
}),
deleteS3ObjectActivity: Object.assign(vi.fn(), {
_activityName: 'deleteS3ObjectActivity',
}),
}
mockActivities.getAgentWorkerQueueActivity.mockResolvedValue('agent_queue')
mockActivities.getTranscodeWorkerQueueActivity.mockResolvedValue('transcode_queue')
mockActivities.updateTaskStatusActivity.mockResolvedValue({})
mockActivities.createCommentActivity.mockResolvedValue({ id: 'comment-placeholder-id' })
mockActivities.getEmbeddingContextActivity.mockResolvedValue({
agent: { id: 'b1' },
asset: { id: 'a1' },
dbProvider: { name: 'google' },
asset: {
id: 'a1',
mediaType: 'image/png',
storageKey: { key: 'test.png' },
},
chunkDuration: 60.0,
chunkOverlap: 5.0,
})
mockActivities.generateEmbeddingActivity.mockResolvedValue({
embeddings: [{ embedding: [0.1, 0.2] }],
mockActivities.generateImageEmbeddingActivity.mockResolvedValue({
embedding: [0.1, 0.2],
usage: { inputTokens: 5, outputTokens: 5, model: 'gpt' },
})
mockActivities.generateVideoChunkEmbeddingActivity.mockResolvedValue({
embedding: [0.3, 0.4],
usage: { inputTokens: 3, outputTokens: 3, model: 'gpt' },
})
mockActivities.downloadMediaToTmpActivity.mockResolvedValue({
filePath: '/tmp/test.mp4',
tmpDir: '/tmp/test-dir',
})
mockActivities.transcodeVideoChunkActivity.mockImplementation(
async (params: { assetId: string; startTime: number; endTime: number }) => {
return {
chunkKey: `files/${params.assetId}/tmp-embedding-chunks/chunk-${params.startTime}-${params.endTime}.mp4`,
}
},
)
mockActivities.deleteS3ObjectActivity.mockResolvedValue(undefined)
mockActivities.cleanupTmpDirActivity.mockResolvedValue(undefined)
// eslint-disable-next-line @typescript-eslint/no-explicit-any -- mockActivities contains vi.fn mock functions which are cast to expected activity proxy types
vi.mocked(workflowUtils.getActivities).mockReturnValue(mockActivities as any)
@@ -64,7 +105,7 @@ describe('Agent Embedding Workflow', () => {
})
})
it('should run agent embedding workflow successfully', async () => {
it('should run agent embedding workflow successfully for image', async () => {
const task = await prisma.workflowTask.create({
data: {
type: 'ai_embedding',
@@ -103,15 +144,11 @@ describe('Agent Embedding Workflow', () => {
assetId: 'a1',
})
// Verify embedding generation
expect(mockActivities.generateEmbeddingActivity).toHaveBeenCalledWith({
// Verify image embedding generation
expect(mockActivities.generateImageEmbeddingActivity).toHaveBeenCalledWith({
teamId: 't1',
assetId: 'a1',
context: {
agent: { id: 'b1' },
asset: { id: 'a1' },
dbProvider: { name: 'google' },
},
assetKey: 'test.png',
mediaType: 'image/png',
})
// Verify embeddings saved
@@ -125,7 +162,7 @@ describe('Agent Embedding Workflow', () => {
taskId: task.id,
inputTokens: 5,
outputTokens: 5,
model: 'gpt',
model: 'gemini-embedding-2',
})
// Verify placeholder comment updated
@@ -141,6 +178,142 @@ describe('Agent Embedding Workflow', () => {
})
})
it('should run agent embedding workflow successfully for image with transcoded format', async () => {
mockActivities.getEmbeddingContextActivity.mockResolvedValue({
agent: { id: 'b1' },
asset: {
id: 'a1',
mediaType: 'image/png',
storageKey: { key: 'test.png' },
media: {
imageTranscodes: [{ key: 'test-transcoded.webp', isRaw: false, format: 'webp' }],
},
},
chunkDuration: 60.0,
chunkOverlap: 5.0,
})
const task = await prisma.workflowTask.create({
data: {
type: 'ai_embedding',
status: 'pending',
assetId: 'a1',
teamId: 't1',
},
})
await agentEmbeddingMedia(task)
expect(mockActivities.generateImageEmbeddingActivity).toHaveBeenCalledWith({
teamId: 't1',
assetKey: 'test-transcoded.webp',
mediaType: 'image/webp',
})
})
it('should run agent embedding workflow successfully for video in chunks', async () => {
mockActivities.getEmbeddingContextActivity.mockResolvedValue({
agent: { id: 'b1' },
asset: {
id: 'a1',
mediaType: 'video/mp4',
storageKey: { key: 'test.mp4' },
media: {
duration: 150.0,
videoTranscodes: [{ key: 'test-transcoded.mp4', isRaw: false }],
},
},
chunkDuration: 60.0,
chunkOverlap: 5.0,
})
const task = await prisma.workflowTask.create({
data: {
type: 'ai_embedding',
status: 'pending',
assetId: 'a1',
teamId: 't1',
payload: {
projectId: 'p1',
agent: { sessionId: 's1', agentId: 'agent-1' },
},
},
})
await agentEmbeddingMedia(task)
// Verify video transcode queue discovery & download
expect(mockActivities.getTranscodeWorkerQueueActivity).toHaveBeenCalled()
expect(mockActivities.downloadMediaToTmpActivity).toHaveBeenCalledWith({
assetKey: 'test-transcoded.mp4',
})
// Verify video chunk slicing activity called 3 times on transcode queue
expect(mockActivities.transcodeVideoChunkActivity).toHaveBeenCalledTimes(3)
expect(mockActivities.transcodeVideoChunkActivity).toHaveBeenNthCalledWith(1, {
assetId: 'a1',
filePath: '/tmp/test.mp4',
startTime: 0,
endTime: 60,
})
expect(mockActivities.transcodeVideoChunkActivity).toHaveBeenNthCalledWith(2, {
assetId: 'a1',
filePath: '/tmp/test.mp4',
startTime: 55,
endTime: 115,
})
expect(mockActivities.transcodeVideoChunkActivity).toHaveBeenNthCalledWith(3, {
assetId: 'a1',
filePath: '/tmp/test.mp4',
startTime: 110,
endTime: 150,
})
// Verify video chunk embedding generation (3 chunks) on agent queue
expect(mockActivities.generateVideoChunkEmbeddingActivity).toHaveBeenCalledTimes(3)
expect(mockActivities.generateVideoChunkEmbeddingActivity).toHaveBeenNthCalledWith(1, {
teamId: 't1',
chunkKey: 'files/a1/tmp-embedding-chunks/chunk-0-60.mp4',
})
expect(mockActivities.generateVideoChunkEmbeddingActivity).toHaveBeenNthCalledWith(2, {
teamId: 't1',
chunkKey: 'files/a1/tmp-embedding-chunks/chunk-55-115.mp4',
})
expect(mockActivities.generateVideoChunkEmbeddingActivity).toHaveBeenNthCalledWith(3, {
teamId: 't1',
chunkKey: 'files/a1/tmp-embedding-chunks/chunk-110-150.mp4',
})
// Verify deletion of temporary chunks from S3
expect(mockActivities.deleteS3ObjectActivity).toHaveBeenCalledTimes(3)
expect(mockActivities.deleteS3ObjectActivity).toHaveBeenNthCalledWith(1, {
key: 'files/a1/tmp-embedding-chunks/chunk-0-60.mp4',
})
// Verify temp directory cleanup on transcode queue
expect(mockActivities.cleanupTmpDirActivity).toHaveBeenCalledWith({
tmpDir: '/tmp/test-dir',
})
// Verify usage update (3 chunks * 3 tokens = 9)
expect(mockActivities.updateTaskUsageActivity).toHaveBeenCalledWith({
taskId: task.id,
inputTokens: 9,
outputTokens: 9,
model: 'gemini-embedding-2',
})
// Verify embeddings saved
expect(mockActivities.saveAssetEmbeddingsActivity).toHaveBeenCalledWith({
assetId: 'a1',
embeddings: [
{ embedding: [0.3, 0.4], startTime: 0, endTime: 60 },
{ embedding: [0.3, 0.4], startTime: 55, endTime: 115 },
{ embedding: [0.3, 0.4], startTime: 110, endTime: 150 },
],
})
})
it('should throw if task has no teamId', async () => {
const task = await prisma.workflowTask.create({
data: {
@@ -160,10 +333,19 @@ describe('Agent Embedding Workflow', () => {
})
})
it('should skip saving embeddings if none are returned', async () => {
mockActivities.generateEmbeddingActivity.mockResolvedValue({
embeddings: [],
usage: null,
it('should skip saving embeddings if none are returned (video duration 0)', async () => {
mockActivities.getEmbeddingContextActivity.mockResolvedValue({
agent: { id: 'b1' },
asset: {
id: 'a1',
mediaType: 'video/mp4',
storageKey: { key: 'test.mp4' },
media: {
duration: 0.0,
},
},
chunkDuration: 60.0,
chunkOverlap: 5.0,
})
const task = await prisma.workflowTask.create({
@@ -178,7 +360,6 @@ describe('Agent Embedding Workflow', () => {
await agentEmbeddingMedia(task)
expect(mockActivities.saveAssetEmbeddingsActivity).not.toHaveBeenCalled()
expect(mockActivities.updateTaskUsageActivity).not.toHaveBeenCalled()
expect(mockActivities.updateTaskStatusActivity).toHaveBeenCalledWith({
taskId: task.id,
status: 'completed',
@@ -186,7 +367,9 @@ describe('Agent Embedding Workflow', () => {
})
it('should handle failures, update placeholder with error, and set status to failed', async () => {
mockActivities.generateEmbeddingActivity.mockRejectedValue(new Error('AI Service Unavailable'))
mockActivities.generateImageEmbeddingActivity.mockRejectedValue(
new Error('AI Service Unavailable'),
)
const task = await prisma.workflowTask.create({
data: {
@@ -212,4 +395,40 @@ describe('Agent Embedding Workflow', () => {
output: { error: 'AI Service Unavailable' },
})
})
it('should not create redundant chunks when video exactly matches chunk duration', async () => {
mockActivities.getEmbeddingContextActivity.mockResolvedValue({
agent: { id: 'b1' },
asset: {
id: 'a1',
mediaType: 'video/mp4',
storageKey: { key: 'test.mp4' },
media: {
duration: 60.0,
},
},
chunkDuration: 60.0,
chunkOverlap: 5.0,
})
const task = await prisma.workflowTask.create({
data: {
type: 'ai_embedding',
status: 'pending',
assetId: 'a1',
teamId: 't1',
},
})
await agentEmbeddingMedia(task)
// Verify only 1 chunk was created instead of 2 (0-60 and 55-60)
expect(mockActivities.transcodeVideoChunkActivity).toHaveBeenCalledTimes(1)
expect(mockActivities.transcodeVideoChunkActivity).toHaveBeenCalledWith(
expect.objectContaining({
startTime: 0,
endTime: 60,
}),
)
})
})
+173 -19
View File
@@ -1,25 +1,48 @@
import { ApplicationFailure } from '@temporalio/workflow'
import type { WorkflowTask } from '@shumai/db'
import { getActivities, executeActivity, TaskQueueAgent } from '@shumai/workflow-core'
import {
getActivities,
executeActivity,
TaskQueueAgent,
TaskQueueTranscode,
} from '@shumai/workflow-core'
export interface GeneratedEmbedding {
embedding: number[]
startTime?: number
endTime?: number
}
export async function agentEmbeddingMedia(task: WorkflowTask): Promise<void> {
const {
updateTaskStatusActivity,
getEmbeddingContextActivity,
generateEmbeddingActivity,
generateImageEmbeddingActivity,
generateVideoChunkEmbeddingActivity,
saveAssetEmbeddingsActivity,
updateTaskUsageActivity,
createCommentActivity,
updateCommentActivity,
getAgentWorkerQueueActivity,
getTranscodeWorkerQueueActivity,
downloadMediaToTmpActivity,
cleanupTmpDirActivity,
transcodeVideoChunkActivity,
deleteS3ObjectActivity,
} = getActivities()
let placeholderCommentId: string | undefined
let agentWorkerQueue = ''
let transcodeWorkerQueue = ''
let tmpDir: string | undefined
try {
// 0. Discover queue
// 0. Discover queues
agentWorkerQueue = await executeActivity(TaskQueueAgent, getAgentWorkerQueueActivity)
transcodeWorkerQueue = await executeActivity(
TaskQueueTranscode,
getTranscodeWorkerQueueActivity,
)
// Update status to processing
await executeActivity(agentWorkerQueue, updateTaskStatusActivity, {
@@ -47,30 +70,153 @@ export async function agentEmbeddingMedia(task: WorkflowTask): Promise<void> {
assetId: task.assetId,
})
// Call the activity to generate embeddings
const result = await executeActivity(agentWorkerQueue, generateEmbeddingActivity, {
teamId: task.teamId,
assetId: task.assetId,
context,
})
// eslint-disable-next-line @typescript-eslint/no-explicit-any
const { asset, chunkDuration, chunkOverlap } = context as any
const isImage = asset.mediaType?.startsWith('image/')
const isVideo = asset.mediaType?.startsWith('video/')
if (!isImage && !isVideo) {
throw ApplicationFailure.create({
message: `unsupported media type for embeddings: ${asset.mediaType}`,
nonRetryable: true,
})
}
// Resolve the S3 key to use for embedding (preferring transcoded version)
let embeddingKey = asset.storageKey?.key
let embeddingMediaType = asset.mediaType
if (asset.media) {
// eslint-disable-next-line @typescript-eslint/no-explicit-any
const media = asset.media as any
if (isVideo) {
if (media.videoTranscodes && media.videoTranscodes.length > 0) {
// Find first non-raw transcode or fallback to first transcode
const transcode =
media.videoTranscodes.find((t: PrismaJson.VideoTranscode) => !t.isRaw) ||
media.videoTranscodes[0]
if (transcode?.key) {
embeddingKey = transcode.key
}
}
} else if (isImage) {
if (media.imageTranscodes && media.imageTranscodes.length > 0) {
// Find first non-raw transcode or fallback to first transcode
const transcode =
media.imageTranscodes.find((t: PrismaJson.ImageTranscode) => !t.isRaw) ||
media.imageTranscodes[0]
if (transcode?.key) {
embeddingKey = transcode.key
if (transcode.format) {
embeddingMediaType = `image/${transcode.format}`
}
}
}
}
}
if (!embeddingKey) {
throw ApplicationFailure.create({ message: 'asset has no key', nonRetryable: true })
}
const results: GeneratedEmbedding[] = []
const usage = {
model: 'gemini-embedding-2',
inputTokens: 0,
outputTokens: 0,
}
if (isImage) {
const result = await executeActivity(agentWorkerQueue, generateImageEmbeddingActivity, {
teamId: task.teamId,
assetKey: embeddingKey,
mediaType: embeddingMediaType,
})
results.push({ embedding: result.embedding })
if (result.usage) {
usage.inputTokens += result.usage.inputTokens || 0
usage.outputTokens += result.usage.outputTokens || 0
}
} else if (isVideo) {
// 1. Download full video to transcode worker temp space
const download = await executeActivity(transcodeWorkerQueue, downloadMediaToTmpActivity, {
assetKey: embeddingKey,
})
const { filePath } = download
tmpDir = download.tmpDir
// Get video duration
// eslint-disable-next-line @typescript-eslint/no-explicit-any
const duration = (asset.media as any)?.duration || 0
const limit = chunkDuration || 60.0
const overlap = chunkOverlap ?? 5.0
const step = Math.max(1, limit - overlap)
for (let start = 0.0; start < duration; start += step) {
let end = start + limit
if (end >= duration) {
end = duration
}
// Step A: Slice chunk on transcode worker and upload to S3
const chunkRes = await executeActivity(transcodeWorkerQueue, transcodeVideoChunkActivity, {
assetId: asset.id,
filePath,
startTime: start,
endTime: end,
})
const { chunkKey } = chunkRes
try {
// Step B: Generate embedding on agent worker
const chunkResult = await executeActivity(
agentWorkerQueue,
generateVideoChunkEmbeddingActivity,
{
teamId: task.teamId,
chunkKey,
},
)
results.push({
embedding: chunkResult.embedding,
startTime: start,
endTime: end,
})
if (chunkResult.usage) {
usage.inputTokens += chunkResult.usage.inputTokens || 0
usage.outputTokens += chunkResult.usage.outputTokens || 0
}
} finally {
// Step C: Delete temporary chunk from S3
await executeActivity(transcodeWorkerQueue, deleteS3ObjectActivity, {
key: chunkKey,
})
}
// If this chunk reached the end of the video, stop to avoid redundant overlapping chunks
if (end >= duration) {
break
}
}
}
// Save computed embeddings
if (result.embeddings.length > 0) {
if (results.length > 0) {
await executeActivity(agentWorkerQueue, saveAssetEmbeddingsActivity, {
assetId: task.assetId,
embeddings: result.embeddings,
embeddings: results,
})
}
// Update Usage
if (result.usage) {
await executeActivity(agentWorkerQueue, updateTaskUsageActivity, {
taskId: task.id,
inputTokens: result.usage.inputTokens,
outputTokens: result.usage.outputTokens,
model: result.usage.model,
})
}
await executeActivity(agentWorkerQueue, updateTaskUsageActivity, {
taskId: task.id,
inputTokens: usage.inputTokens,
outputTokens: usage.outputTokens,
model: usage.model,
})
// 7. Update Placeholder Comment
if (placeholderCommentId) {
@@ -109,6 +255,14 @@ export async function agentEmbeddingMedia(task: WorkflowTask): Promise<void> {
})
}
throw err
} finally {
if (tmpDir && transcodeWorkerQueue) {
try {
await executeActivity(transcodeWorkerQueue, cleanupTmpDirActivity, { tmpDir })
} catch (cleanupErr) {
console.error('Failed to cleanup transcode worker tmp dir:', cleanupErr)
}
}
}
}
+1 -1
View File
@@ -1,6 +1,6 @@
{
"name": "@shumai/api",
"version": "0.0.6",
"version": "0.0.8",
"private": true,
"type": "module",
"main": "./src/index.ts",
@@ -17,7 +17,8 @@ describe('authn api', () => {
it('GET /signup-info', async () => {
mockGetSignupInfo.mockResolvedValue({
userCount: 5,
initialized: true,
demoMode: false,
})
const app = new Hono().route('/', authnRoute)
@@ -25,7 +26,8 @@ describe('authn api', () => {
expect(res.status).toBe(200)
expect(await res.json()).toEqual({
userCount: 5,
initialized: true,
demoMode: false,
})
})
})
@@ -17,7 +17,8 @@ const route = app
.get('/signup-info', async (c) => {
const info = await teamService.getSignupInfo()
return c.json({
userCount: info.userCount,
initialized: info.initialized,
demoMode: info.demoMode,
})
})
.get('/me', async (c) => {
@@ -6,7 +6,7 @@ import { assetService } from '@shumai/core/src/asset/asset'
import { metadataService } from '@shumai/core/src/metadata/metadata'
import { notificationService } from '@shumai/core/src/notification/notification'
import { s3Service } from '@shumai/core/src/s3/s3'
import { transcodeService } from '@shumai/transcode/src/transcode'
import { transcodeService } from '@shumai/core'
import { authMiddleware } from './middleware/auth'
import { authzService, ResourceType, Permission } from '@shumai/core/src/authz/authz'
@@ -17,7 +17,7 @@ import {
uploadFileRequestSchema,
getDownloadLinksRequestSchema,
} from '@shumai/dtos'
import { transcodeService } from '@shumai/transcode/src/transcode'
import { transcodeService } from '@shumai/core'
import fs from 'fs'
import { Hono } from 'hono'
import os from 'os'
+20 -20
View File
@@ -1,25 +1,25 @@
import type { Prisma } from '@shumai/db'
import { Hono } from 'hono'
import agentRoute from './api/agent'
import attachmentRoute from './api/attachment'
import authnRoute from './api/authn'
import collectionRoute from './api/collection'
import fileRoute from './api/file'
import folderRoute from './api/folder'
import inviteRoute from './api/invite'
import metadataRoute from './api/metadata'
import { authMiddleware } from './api/middleware/auth'
import notificationRoute from './api/notification'
import projectRoute from './api/project'
import providerRoute from './api/provider'
import publicInviteRoute from './api/public-invite'
import publicShareRoute from './api/public-share'
import s3Route from './api/s3'
import shareRoute from './api/share'
import skillRoute from './api/skill'
import teamRoute from './api/team'
import uploadRoute from './api/upload'
import versionStackRoute from './api/versionStack'
import agentRoute from './agent'
import attachmentRoute from './attachment'
import authnRoute from './authn'
import collectionRoute from './collection'
import fileRoute from './file'
import folderRoute from './folder'
import inviteRoute from './invite'
import metadataRoute from './metadata'
import { authMiddleware } from './middleware/auth'
import notificationRoute from './notification'
import projectRoute from './project'
import providerRoute from './provider'
import publicInviteRoute from './public-invite'
import publicShareRoute from './public-share'
import s3Route from './s3'
import shareRoute from './share'
import skillRoute from './skill'
import teamRoute from './team'
import uploadRoute from './upload'
import versionStackRoute from './versionStack'
type User = Prisma.UserGetPayload<Record<string, never>>
+122
View File
@@ -0,0 +1,122 @@
import { describe, expect, it, vi, beforeEach } from 'vitest'
import { Hono } from 'hono'
import { authMiddleware } from './auth'
import { auth } from '@shumai/core/src/auth/auth'
import { prisma } from '@shumai/db'
vi.mock('@shumai/core/src/auth/auth', () => ({
auth: {
api: {
getSession: vi.fn(),
},
},
}))
vi.mock('@shumai/db', () => ({
prisma: {
user: {
findUnique: vi.fn(),
},
teamMember: {
findFirst: vi.fn(),
},
},
}))
describe('authMiddleware', () => {
const app = new Hono()
.use('*', authMiddleware)
.get('/test', (c) => c.text('ok'))
.post('/test', (c) => c.text('ok'))
.post('/test/search', (c) => c.text('ok'))
beforeEach(() => {
vi.resetAllMocks()
delete process.env.SHUMAI_DEMO_MODE
})
it('allows access when authenticated', async () => {
// eslint-disable-next-line @typescript-eslint/no-explicit-any
vi.mocked(auth.api.getSession).mockResolvedValue({ user: { id: 'user1' } } as any)
// eslint-disable-next-line @typescript-eslint/no-explicit-any
vi.mocked(prisma.user.findUnique).mockResolvedValue({ id: 'user1' } as any)
const res = await app.request('/test')
expect(res.status).toBe(200)
})
it('rejects when not authenticated', async () => {
vi.mocked(auth.api.getSession).mockResolvedValue(null)
const res = await app.request('/test')
expect(res.status).toBe(401)
})
describe('read-only mode', () => {
beforeEach(() => {
process.env.SHUMAI_DEMO_MODE = '1'
})
it('allows GET for reviewer', async () => {
// eslint-disable-next-line @typescript-eslint/no-explicit-any
vi.mocked(auth.api.getSession).mockResolvedValue({ user: { id: 'user1' } } as any)
// eslint-disable-next-line @typescript-eslint/no-explicit-any
vi.mocked(prisma.user.findUnique).mockResolvedValue({ id: 'user1' } as any)
const res = await app.request('/test')
expect(res.status).toBe(200)
})
it('rejects POST for reviewer', async () => {
// eslint-disable-next-line @typescript-eslint/no-explicit-any
vi.mocked(auth.api.getSession).mockResolvedValue({ user: { id: 'user1' } } as any)
// eslint-disable-next-line @typescript-eslint/no-explicit-any
vi.mocked(prisma.user.findUnique).mockResolvedValue({ id: 'user1' } as any)
vi.mocked(prisma.teamMember.findFirst).mockResolvedValue(null) // No owner/editor role
const res = await app.request('/test', { method: 'POST' })
expect(res.status).toBe(403)
expect(await res.json()).toEqual({ error: 'System is in read-only mode' })
})
it('allows POST /search for reviewer', async () => {
// eslint-disable-next-line @typescript-eslint/no-explicit-any
vi.mocked(auth.api.getSession).mockResolvedValue({ user: { id: 'user1' } } as any)
// eslint-disable-next-line @typescript-eslint/no-explicit-any
vi.mocked(prisma.user.findUnique).mockResolvedValue({ id: 'user1' } as any)
const res = await app.request('/test/search', { method: 'POST' })
expect(res.status).toBe(200)
})
it('allows POST for owner', async () => {
// eslint-disable-next-line @typescript-eslint/no-explicit-any
vi.mocked(auth.api.getSession).mockResolvedValue({ user: { id: 'user1' } } as any)
// eslint-disable-next-line @typescript-eslint/no-explicit-any
vi.mocked(prisma.user.findUnique).mockResolvedValue({ id: 'user1' } as any)
vi.mocked(prisma.teamMember.findFirst).mockResolvedValue({
id: 'member1',
role: 'owner',
// eslint-disable-next-line @typescript-eslint/no-explicit-any
} as any)
const res = await app.request('/test', { method: 'POST' })
expect(res.status).toBe(200)
})
it('allows POST for editor', async () => {
// eslint-disable-next-line @typescript-eslint/no-explicit-any
vi.mocked(auth.api.getSession).mockResolvedValue({ user: { id: 'user1' } } as any)
// eslint-disable-next-line @typescript-eslint/no-explicit-any
vi.mocked(prisma.user.findUnique).mockResolvedValue({ id: 'user1' } as any)
vi.mocked(prisma.teamMember.findFirst).mockResolvedValue({
id: 'member1',
role: 'editor',
// eslint-disable-next-line @typescript-eslint/no-explicit-any
} as any)
const res = await app.request('/test', { method: 'POST' })
expect(res.status).toBe(200)
})
})
})
@@ -28,6 +28,24 @@ export const authMiddleware = createMiddleware<{
}
c.set('user', user)
if (process.env.SHUMAI_DEMO_MODE === '1') {
const method = c.req.method.toUpperCase()
const path = c.req.path
if (['POST', 'PUT', 'PATCH', 'DELETE'].includes(method) && !path.endsWith('/search')) {
const member = await prisma.teamMember.findFirst({
where: {
userId: user.id,
role: { in: ['owner', 'editor'] },
},
})
if (!member) {
return c.json({ error: 'System is in read-only mode' }, 403)
}
}
}
await next()
} catch {
return c.json({ error: 'Authentication failed' }, 401)
@@ -48,6 +48,8 @@ const route = app
hasPassword: shareLink.hasPassword,
rootFolderId: shareLink.rootFolderId,
projectId: shareLink.projectId,
viewMode: shareLink.viewMode,
defaultSortOrder: shareLink.defaultSortOrder,
})
} catch (err) {
return c.json({ error: (err as Error).message }, 404)
+2 -1
View File
@@ -1,6 +1,6 @@
{
"name": "@shumai/core",
"version": "0.0.6",
"version": "0.0.8",
"private": true,
"type": "module",
"main": "./src/index.ts",
@@ -18,6 +18,7 @@
"hono": "4.12.23",
"jittered-fractional-indexing": "1.0.1",
"pino": "10.3.1",
"sharp": "0.34.5",
"tweetnacl": "1.0.3",
"ulid": "3.0.2",
"yaml": "2.9.0",
+7 -3
View File
@@ -15,9 +15,13 @@ export const auth = betterAuth({
},
secret: process.env.BETTER_AUTH_SECRET,
// eslint-disable-next-line @typescript-eslint/naming-convention
baseURL: process.env.BETTER_AUTH_URL || 'http://localhost:3000',
baseURL:
process.env.BETTER_AUTH_URL || `http://localhost:${process.env.SHUMAI_SERVER_PORT || '3000'}`,
basePath: '/api/auth',
trustedOrigins: ['http://localhost:3000', 'http://127.0.0.1:3000'],
trustedOrigins: [
`http://localhost:${process.env.SHUMAI_SERVER_PORT || '3000'}`,
`http://127.0.0.1:${process.env.SHUMAI_SERVER_PORT || '3000'}`,
],
advanced: {
disableCSRFCheck: true,
},
@@ -30,7 +34,7 @@ export const auth = betterAuth({
const info = await teamService.getSignupInfo()
if (info.userCount > 0) {
if (info.initialized) {
if (!inviteCode) {
console.log('[Better Auth Hook] Public signup is disabled')
return ctx.json(
+1 -1
View File
@@ -315,7 +315,7 @@ describe('AuthzService', () => {
// Provider
const provider = await prisma.provider.create({
data: { name: 'Prov 1', teamId: team.id, config: { api: 'openai' } },
data: { name: 'Prov 1', teamId: team.id, config: { api: 'openai-completions' } },
})
// Invite
+1
View File
@@ -13,6 +13,7 @@ export * from './search/search'
export * from './share/share'
export * from './skill/skill'
export * from './team/team'
export * from './transcode/transcode'
export * from './upload/upload'
export * from './user-metadata/user-metadata'
export * from './user/avatar'
@@ -362,4 +362,70 @@ describe('NotificationService', () => {
count = await notificationService.getUnreadCount(tm.id, u1.id)
expect(count).toBe(0)
})
it('toNotificationInfo maps video sprite and poster urls', async () => {
const tm = await prisma.team.create({ data: { name: 'Team 1' } })
const asset = await prisma.asset.create({
data: {
name: 'test-video.mp4',
type: 'file',
mediaType: 'video/mp4',
status: 'processed',
media: {
duration: 10,
filesize: 1024,
fps: 24,
frames: 240,
videoPreview: { width: 640, height: 360 },
videoTranscodes: [],
imageTranscodes: [],
sprite: { key: 'sprites/test-video.jpg', tileX: 10, tileY: 10, frames: 100 },
poster: { key: 'posters/test-video.jpg' },
thumbnail: {
key: 'thumbnails/test-video.jpg',
width: 120,
height: 90,
format: 'jpg',
quality: 80,
},
metadata: {
originalWidth: 1280,
originalHeight: 720,
duration: 10,
bitRate: 1000,
frameRate: 24,
hasAudio: false,
format: {},
},
finishedAt: new Date().toISOString(),
mimeType: 'video/mp4',
original: null,
},
},
})
const n = await prisma.notification.create({
data: {
teamId: tm.id,
type: NotificationType.comment_created,
assetId: asset.id,
},
})
const svc = notificationService
const info = await svc['toNotificationInfo'](
await prisma.notification.findUniqueOrThrow({
where: { id: n.id },
include: { creator: true, team: true, project: true, asset: true, user: true },
}),
null,
)
expect(info.asset?.preview).toBeDefined()
expect(info.asset?.thumbnailUrl).toBeDefined()
expect(info.asset?.thumbnailUrl).toContain('posters/test-video.jpg')
expect(info.asset?.originalWidth).toBe(1280)
expect(info.asset?.originalHeight).toBe(720)
})
})
@@ -371,6 +371,9 @@ export class NotificationService {
if (n.asset) {
const a = n.asset
let previewUrl = ''
let thumbnailUrl = ''
let originalWidth = 0
let originalHeight = 0
const media = a.media as PrismaJson.MediaInfo | null
@@ -379,8 +382,16 @@ export class NotificationService {
try {
if (media.sprite) {
previewUrl = await s3Service.presign(bucket, media.sprite.key, 'GET')
if (media.poster) {
thumbnailUrl = await s3Service.presign(bucket, media.poster.key || '', 'GET')
}
if (media.metadata) {
originalWidth = media.metadata.originalWidth
originalHeight = media.metadata.originalHeight
}
} else if (media.thumbnail) {
previewUrl = await s3Service.presign(bucket, media.thumbnail.key || '', 'GET')
thumbnailUrl = previewUrl
}
} catch (e) {
console.error('Failed to generate presigned url for notification asset preview:', e)
@@ -392,6 +403,9 @@ export class NotificationService {
name: a.name,
mediaType: a.mediaType || undefined,
preview: previewUrl || undefined,
thumbnailUrl: thumbnailUrl || undefined,
originalWidth: originalWidth || undefined,
originalHeight: originalHeight || undefined,
}
}
+4 -4
View File
@@ -38,7 +38,7 @@ describe('ProviderService', () => {
expect(providers[0].name).toBe(firstInGenerated)
})
it('should prevent deletion of builtin providers', async () => {
it('should allow deletion of builtin providers', async () => {
const team = await teamService.ensureDefaultTeam()
await providerService.initBuiltinProviders(team.id)
@@ -46,9 +46,9 @@ describe('ProviderService', () => {
const builtin = providers.find((p) => p.isBuiltin)
expect(builtin).toBeDefined()
await expect(providerService.delete(team.id, builtin!.id)).rejects.toThrow(
'Cannot delete builtin provider',
)
await providerService.delete(team.id, builtin!.id)
const afterDelete = await providerService.listByTeam(team.id)
expect(afterDelete.find((p) => p.id === builtin!.id)).toBeUndefined()
})
it('should allow editing builtin providers', async () => {
-4
View File
@@ -143,10 +143,6 @@ export class ProviderService {
throw new Error('Provider not found')
}
if (provider.isBuiltin) {
throw new Error('Cannot delete builtin provider')
}
return this.prismaClient.provider.delete({
where: { id, teamId },
})

Some files were not shown because too many files have changed in this diff Show More