* chore(deps): upgrade bun to 1.4.2 and update packages
- Upgrade Bun to 1.4.2 in Dockerfile, CI workflows, and release workflows.
- Update dependencies across workspace with Bun 1.4.2.
- Preserve @earendil-works/pi-agent-core at 0.82.1.
- Retain stable compatible versions for database adapters and pdfjs-dist.
- Add explicit Buffer type annotation in transcode service.
* chore(deps): upgrade prisma to 7.10.0, pg to 8.23.0, and pdfjs-dist to 6.3.289
* chore(deps): revert dotenv-expand to 13.0.0
* feat(agent): add built-in image and video generation tools and settings
* fix(settings): refine media generation layout and expose media tools in agent and quota forms
* feat(media-generation): add alibaba, bytedance, minimax, prodia, and azure providers
* feat(agent): support s3 key references for media generation and s3KeyOnly in read_asset
* feat(asset): record and display agent provenance on assets and audit logs
* style(webui): apply 30% translucent frosted glass background to bot badges
* style(webui): match bot icon badge style to thumbnail media overlay (bg-black/60 text-white)
* style(webui): use "AI" text badge with pdf overlay style instead of bot icon
* feat(provider): optimize provider and model configuration management
* style(provider): use default tabs variant in provider settings dialog
* style(provider): make tabs list full width
* feat(provider): make model card row clickable and remove pencil edit icon
* feat(settings): persist active settings tab in url hash
* feat(agent): add search box and alphabetical sorting for provider and model selectors
* refactor(providers): remove frontend pagination and display all models in scrollable list
* fix(providers): address code review findings for model editing, i18n, and api protocol propagation
* feat(agent): consolidate media tools into read_asset and add S3 key support to download_asset
* fix(agent): refine parameter descriptions and remove download URL from read_asset text content
* fix(agent): resolve owning asset and enforce RBAC permissions when downloading by S3 key
* feat(agent): improve timestamp precision and proxy selection for video screenshots
* feat(agent): support dynamic annotationId resolution across chat and comments
* fix(agent): scope annotation resolution to asset and format has_markup consistently
* feat(chat): support markups, attachments, and timestamps in agent 1-to-1 chat
* feat(chat): improve message card layout with inline badges and integrated assets
* fix(chat): enable image attachment thumbnails in 1-to-1 chat cards
* fix(chat): address code review feedback on input state, mentions, i18n, and session deletion
* fix(chat): replace ring with border color on selected message card
* fix(quota): exclude agents from quota rules and usage records
* refactor(quota): remove agent_skill_call_count resource type and related logic
* chore(db): replace manual migration with prisma-generated migration
* feat(quota): add database-backed resource quota system
* feat(webui): add resource quota management in team settings (#319)
* feat(webui): add resource quota settings in team management
* refactor(webui): update quotas list to full row layout and restrict skill/mcp to select list
* feat(quota): refactor quota system to 2-level rules and records
* fix(webui): fix overflow and scroll area in quota usage records dialog
* fix(webui): ensure quota records scroll area has explicit max height
* feat(webui): improve quota cards interaction and restrict editable fields in edit mode
* fix(quota): serialize initial records, validate merged updates, and reset on dimension changes (#320)
* feat(quota): replace network quota with agent tools quota (#321)
* feat(webui): show remaining time until next refresh in quota usage records
* fix(webui): rename bash quota resource to Agent Bash Calls
* refactor(quota): remove unused reserved field from quota records
* refactor(quota): remove quota rule cache (#322)
* feat(authz): enforce project-aware roles for agent/skill/mcp permissions
Introduce a shared resolveEffectiveRole helper that mirrors AuthzService
precedence rules (projectMember.role override -> team role for team-scoped
members -> restricted), and route all agent/skill/mcp runtime permission
checks through it. Previously these checks used the raw teamMember.role,
which let project-only members execute team-level agents/skills/MCP tools
and ignored per-project role overrides.
Changes:
- authz: add resolveEffectiveRole(teamId, projectId, userId, db) and
refactor hasPermission to use it (behavior-preserving)
- chat/asset/agent-activity: gate agent usage by the effective project role
- read_skill: enforce skill permission via effective role (projectId
threaded from the agent session)
- mcp-service: checkPermission and buildAgentTools filter by effective
role; un-permitted servers' tools are no longer advertised to the model
- agent activity: only advertise skills the user's effective role permits
- listProjectMembers: filter included agent bots by the requester's
effective project role
- new GET /projects/:projectId/chat-agents API returning chat agents
scoped to the project and the user's effective role; WebUI chat selector
now uses it (settings keeps the team-level API)
- tests: effective-role resolver, chat/mention/MCP/skill project-role
matrices, listProjectChatAgents, listProjectMembers filter, API route
* fix(agent): forward projectId to getAgentChatContextActivity in agentChat workflow
The agentChat workflow called getAgentChatContextActivity without passing
projectId from the task payload, so project-scoped members (scope:
'project') resolved a null effective role in the activity and were denied
with 'Insufficient role to use agent' even though the chat selector and
chat.ts both authorized them. Forward payload.projectId so the activity
enforces the user's effective project role, and assert the forwarding in
the workflow unit test (fails without the fix).
* feat(agent): add permission controls for chat agents
* refactor(agent): extract shared agent permission helpers
Consolidate the duplicated role-hierarchy logic into
packages/core/src/agent/permissions.ts:
- getRoleLevel(): numeric level for a team role (fail closed at 0)
- getAgentRequiredLevel(): required level for an agent permission
(defaults to reviewer/1)
- getAllowedAgentRoles(): inclusive role set for filtering
The typed Record over TeamMemberRole makes the mapping exhaustive,
so adding a new role fails compilation until all sites are updated.
Used by ChatService.startOrContinueChat, AssetService comment
mentions, getAgentChatContextActivity, AgentService.listAgents, and
TeamService.getTeamMembers, replacing four inline copies of the
hierarchy and the allowed-roles conditionals.
Also fix the API permission-denial test to mock HTTPException(403)
and assert a 403 response instead of a generic Error/500, matching
real authzService behavior.
* test(agent): fix flaky comment DAG tests by spacing ULID id generation
ULID ids have 1ms timestamp granularity; back-to-back comment creates
frequently land in the same millisecond, in which case Prisma ULIDs sort
by their random suffix instead of creation order. The DAG spine in
initializeAgentSessionActivity chains top-level comments by id, so a
shuffled order produces paths that omit preceding comments, making the
path-to-root assertions flaky (~80% of full-file runs locally).
Add a 2ms sleep between consecutive comment creates in the affected
tests, guaranteeing strictly increasing ULID timestamps. A 2ms delay is
sufficient because the next create always fires at least 2ms after the
previous ULID was generated. Production is unaffected: real comments
arrive seconds apart via separate requests.
* feat(agent): add nested agents.md support across backend and webui
* fix(webui): update agentsmd route paths and unnest folder agentsmd route
* fix(agent): flush agents.md autosave on close and address review feedback
- Flush pending debounced AGENTS.md autosave on close/unmount so edits
made within 750ms of navigating away are no longer silently lost
- Serialize autosaves through a promise chain so an older in-flight
request can never overwrite newer content on the server
- Internationalize the hardcoded "Failed to save" and
"Loading AGENTS.md..." strings (en + zh)
- Validate that AGENTS.md is only stored on folder/root assets in
AssetService, preventing bogus virtual paths for file assets
- Cap updateAgentsMd content at 100k chars via Zod validation
- Remove unused getAgentsMd/updateAgentsMd response DTO schemas
- Drop unused teamId prop from AgentsMdEditor
- Add tests: non-folder asset rejection, oversized content 400,
and agent-chat workflow assetId passthrough
* refactor(db): isolate AGENTS.md content in dedicated 1:1 AssetAgentMd table
* style(webui): unify top bar button hover styles and add wand icon to agentsmd button
* style(webui): update agentsmd button state colors and remove presence dot
* fix(webui): fix close button overflow and stabilize debounced autosave in agentsmd editor
* fix(webui): update debounce to 1s and await pending save before closing agentsmd editor
* feat(webui): add terminal AGENTS.md breadcrumb on agentsmd editor page
* fix(webui): use ScrollArea for markdown editor scrolling
* feat(webui): redesign select and multi-select metadata fields with search box and dynamic option creation
* fix(webui): stabilize select popover placement and preserve field visibility in store
* refactor(webui): enforce 2-row collapsed height and whitespace-pre-wrap on LongTextField
* feat(agent): support dynamic newOption in select and selectMulti metadata fields
* fix(metadata): address code review feedback on selectMulti item validation and WebUI option deduplication
* feat(agent): type-safe CREATION_CONTEXT metadata on create_file/create_version
Build the create_file and create_version tool metadata parameter from the
project's CREATION_CONTEXT fields at harness creation time, instead of the
schemaless key-value map that required the agent to call list_autofill_fields
first.
- Thread projectId from the chat workflow into createAgentSession and build
a strict, all-required (nullable) metadata schema via fieldsToTypeBoxSchema
(additionalProperties: false, option IDs enforced via enum).
- Remove the list_autofill_fields tool and its activity case.
- Reject metadata keys that are not CREATION_CONTEXT fields of the target
project with a clear non-retryable error telling the agent to omit metadata.
- Skip null metadata values when persisting (create_file/create_version and
the autofill workflow).
- Switch the touched schemas from @sinclair/typebox 0.34 to typebox v1 to
match the harness validator.
* refactor(agent): migrate remaining tools from @sinclair/typebox 0.34 to typebox v1
All agent tool schemas now use the same typebox v1 package as the harness
validator; drop the @sinclair/typebox 0.34 dependency from @shumai/agent.
* test(agent): reproduce null coercion in harness tool validation
The pi-ai harness (<=0.82.1) coerces null values in [string, null] unions to
'' during tool argument validation for typebox v1 schemas. A valid create_file
call with path: null + data: {...} therefore arrives at execute with path: ''
and fails with 'Provide exactly one of path or data'. Null metadata values
(the 'unknown' sentinel) are likewise turned into ''.
This test reproduces both failures. It is expected to fail until
@earendil-works/pi-ai is upgraded to >=0.84.0 (#7328), which preserves values
that already match a union arm.
* fix(agent): patch pi-ai 0.82.1 to preserve null in nullable union tool args
Apply the upstream fix for earendil-works/pi#7328 (PR #7373) via a bun
patch instead of upgrading pi-ai. coerceWithUnionSchema now returns the
original value when it already validates against any union member,
instead of coercing null in [string, null] unions into "".
Fixes the create_file false "Provide exactly one of path or data" error
when the model passes path: null with data. Reproduced by
packages/agent/src/reproduce-tool-null-coercion.test.ts, which now passes.
* fix(docker): copy bun patch dir before frozen-lockfile install
The pi-ai patchedDependencies entry makes bun install --frozen-lockfile
fail in the builder stage because patches/ was not present in the image
yet. Copy it alongside package.json/bun.lock so the patch applies.
* refactor(agent): simplify metadata guidance in system prompt and error message
Shorten the create_file/create_version metadata instructions in the system
prompt and reword the invalid-metadata-key error to not reference
CREATION_CONTEXT internals. Update the matching test assertion.
* feat(metadata): replace aiAutofill boolean with autofillSource enum
* fix(agent): validate creation-context metadata and make writes transactional
Address code review feedback on the autofillSource migration:
- Update the agent system prompt to reference the 'metadata' parameter of
create_file/create_version instead of the removed 'context' parameter,
which was being silently discarded by the activity
- Write metadata inside the caller's transaction via the new
metadataService.updateAssetMetadataInTx, keeping metadata rows atomic
with asset creation instead of opening a separate non-atomic transaction
- Reject metadata keys that are not defined CREATION_CONTEXT fields of the
project with a non-retryable ApplicationFailure before creating the asset
- Remove dead agentContext handling from autofillAiActivity and its tests,
and drop the unused ai_autofill i18n keys (paraglide recompiled)
- Derive the DTO autofillSource enum from the AutofillSource const instead
of a hardcoded z.enum to prevent drift
* fix(agent): resolve version stack name and size in list_assets
* fix(agent): resolve version stack to latest version in download_asset, analyze_image, and read_pdf_pages
* feat(mcp): add MCP server support with proxy tool and OAuth flow
Add URL-based MCP (Model Context Protocol) server support to the backend:
- DB: McpServer / AgentMcpServer / McpServerCredential models with
typed JSON columns (auth config, cached tools, stored OAuth tokens,
pending auth state) and a migration.
- Core (@shumai/core/src/mcp): DB-backed credential store, OAuth
client provider (authorization_code + client_credentials) with a
Hono-hosted redirect callback, connection lifecycle manager
(Streamable HTTP with SSE fallback, lazy connect, 401 -> needs_auth),
JSON Schema -> AgentTool parameter conversion, in-memory tool registry
with ranked search, the single 'mcp' proxy AgentTool (search /
describe / list / call / connect / status / auth-start), per-server
opt-in direct tools ({server}_{tool}), and McpService (CRUD,
discovery, OAuth, per-agent assignment via AgentMcpServer, and
permission gating on tool execution mirroring skills).
- Agent: chat agents receive the 'mcp' proxy tool (plus direct tools
for opted-in servers) built from their assigned, enabled servers.
- API: /teams/:teamId/mcp/servers CRUD, tools refresh/list, auth
start/complete/status/remove, test, permission, and the public
OAuth callback route; Admin authz via new ResourceType.McpServer;
audit logging.
- DTOs: zod schemas + types for servers, tools, auth, and per-agent
mcpServerIds.
- Tests: DB store, schema conversion, registry/search, service
integration against a real in-process MCP server (v2
@modelcontextprotocol/server), permission gating, rename
invalidation, direct tools, per-agent assignment, API route tests,
and workflow E2E (local + temporal) for proxy and direct modes.
* fix(mcp): persist OAuth discovery snapshot and stamp SEP-2352 issuer
The callback leg of the OAuth flow failed with
AuthorizationServerMismatchError ('discoveryState was not available on
the callback leg') because McpDbStore.savePendingAuth overwrote the
entire pendingAuth JSON column, clobbering the discoverySnapshot that
the provider's saveDiscoveryState() had stored inside it.
- McpDbStore.savePendingAuth now preserves an in-flight discovery
snapshot so the callback leg can satisfy the SEP-2352 authorization-
server issuer check.
- McpOauthProvider now round-trips the issuer stamp derived from the
discovery snapshot: saveClientInformation/saveTokens persist it,
clientInformation()/tokens() return it, and tokens() backfills a
missing stamp for pre-existing credentials (removes the
'stored OAuth credential has no issuer stamp' warning).
- Regression tests: savePendingAuth keeps the snapshot; the provider
stamps/round-trips the issuer on both legs.
* fix(mcp): backfill SEP-2352 issuer stamp on stale OAuth client info
clientInformation() returned stored client info without an issuer field
when it had been dynamically registered before the stamping fix, so the
SDK logged 'stored OAuth credential has no issuer stamp' on every read.
Now the stored branch backfills the issuer from the discovery snapshot
(persisting it) exactly like the tokens() backfill.
* feat(webui): add MCP server configuration and agent assignment UI (#290)
* feat(webui): add MCP server configuration and agent assignment UI
* fix(webui): auto-detect auth for unconfigured MCP servers and show connect button
* fix(mcp): auto-generate state in mcp-oauth-provider when missing in DB
* fix(mcp): persist OAuth discovery snapshot and stamp SEP-2352 issuer
The callback leg of the OAuth flow failed with
AuthorizationServerMismatchError ('discoveryState was not available on
the callback leg') because McpDbStore.savePendingAuth overwrote the
entire pendingAuth JSON column, clobbering the discoverySnapshot that
the provider's saveDiscoveryState() had stored inside it.
- McpDbStore.savePendingAuth now preserves an in-flight discovery
snapshot so the callback leg can satisfy the SEP-2352 authorization-
server issuer check.
- McpOauthProvider now round-trips the issuer stamp derived from the
discovery snapshot: saveClientInformation/saveTokens persist it,
clientInformation()/tokens() return it, and tokens() backfills a
missing stamp for pre-existing credentials (removes the
'stored OAuth credential has no issuer stamp' warning).
- Regression tests: savePendingAuth keeps the snapshot; the provider
stamps/round-trips the issuer on both legs.
* fix(mcp): backfill SEP-2352 issuer stamp on stale OAuth client info
clientInformation() returned stored client info without an issuer field
when it had been dynamically registered before the stamping fix, so the
SDK logged 'stored OAuth credential has no issuer stamp' on every read.
Now the stored branch backfills the issuer from the discovery snapshot
(persisting it) exactly like the tokens() backfill.
* fix(mcp): attach McpOauthProvider for auto-detected OAuth servers and discover tools on auth completion
* fix(webui): fix flex min-h-0 and max-h scrolling on MCP tools dialog
* fix(webui): set explicit flex height on McpToolsDialog for active scrolling
* feat(webui): make MCP server cards clickable to edit and widen server config dialog
* feat(mcp): agent-level server enablement and auto-detected server metadata (#291)
* feat(mcp): agent-level server enablement and auto-detected server metadata
- Remove the server-level `enabled` column from mcp_servers; enable/disable
now happens per-agent via the AgentMcpServer join table (mirrors skills).
- Drop the unique (teamId, name) constraint so duplicate server names are allowed.
- Add a `description` column, auto-filled from client.getServerVersion() on
connect; the server name is auto-filled from the reported title (fallback name)
with a URL-derived placeholder until first connect.
- Remove manual name input: the create dialog is URL-only, and the edit dialog
shows the auto-detected name/description read-only.
- Replace the include/exclude raw-text inputs with the tool list in the edit
dialog, managing only excludeTools; remove includeTools support and the
separate McpToolsDialog component.
- Sanitize server-reported names into safe direct-tool prefixes (spaces/punctuation
become underscores) and use the same sanitization in the proxy resolver.
* fix(mcp): give the edit-dialog tool list a definite height
The Shadcn ScrollArea viewport is h-full, which requires a definite height
on the root. Using max-h-64 left the viewport at auto height, so the root
clipped the list at 256px without engaging its own scrollbar — tools past
the cutoff were unreachable. Use a fixed h-64 container (same pattern as
move-copy-dialog) so the inner scroll area scrolls independently of the
dialog's outer scroll.
* refactor(mcp): fixed header/footer with scrollable content in server config dialog
Mirror the skill config dialog layout: the edit dialog now keeps the header
and footer fixed and scrolls only the content area (ScrollArea flex-1
min-h-0 inside an h-[85vh] flex column). The tools list no longer has its
own nested ScrollArea — it flows inside the content-level scroll, so all
tools are reachable without a nested-scroll conflict. The create dialog
stays compact (URL-only) like the skill add dialog.
Verified headlessly that max-height-based ScrollArea roots do not engage
scrolling (viewport height resolves to auto), which is why the previous
nested list clipped; the flex-based definite-height pattern scrolls.
* feat(mcp): make the endpoint URL immutable after creation
Editing a server's URL effectively points it at a different server, and the
create flow is URL-only anyway — so the edit dialog no longer allows changing
it (delete + re-add covers endpoint changes). The URL is displayed read-only
next to the auto-detected name/description, and `url` was removed from the
update DTO and service; auth/transport changes still invalidate credentials
and cached tools.
* refactor(mcp): match MCP server cards to the skill card layout
The settings list cards now share the skill card structure: row 1 is the
server name (with direct-mode/transport badges) plus the three-dot menu,
row 2 is the description (falling back to 'No description provided') with
the endpoint URL and last error as small mono lines, and row 3 holds the
connection-status badge and updated date on the left with the permission
selector and the connect button at the right end.
* fix(mcp): stop card menu clicks from opening the edit dialog
The three-dot menu is portaled, so its item clicks bubble through React's
synthetic events to the card's onClick, which opened the edit dialog before
the delete confirmation appeared. Add e.stopPropagation() to every menu
item (matching the skill card pattern).
* fix(mcp): stop Refresh Tools from submitting/closing the config dialog
The Refresh Tools button lived inside the <form> and, like all buttons
without an explicit type, defaulted to type="submit" — clicking it ran the
form submit (save + close) instead of just refreshing. Set type="button".
Also per UX: while refreshing, the button is disabled with a spinning icon,
the per-tool enable/disable switches are disabled, and the success toast is
removed — the button reverting to its idle state is the success signal
(keep the error toast for failures).
* test(mcp): provide explicit redirectUri in McpOauthProvider test helper
* fix(mcp): scope tool registry to agent-assigned servers
The MCP tool registry was a process-global singleton keyed by server id,
so a chat agent could search, describe, and even call tools belonging to
MCP servers that were not assigned to it (e.g. disabled or unassigned
servers), as long as their metadata had been discovered at some point in
the process. This mirrors the disabled-skill bug: the read path never
checked per-agent enablement.
Replace the singleton with a fresh McpToolRegistry built per
buildAgentTools call and warmed from the persisted McpServer.tools DB
cache. Because the registry only ever contains the agent's assigned
servers, unassigned/disabled servers can neither be suggested, searched,
described, nor executed. Live discovery (lazy connect / connect action)
writes into the per-agent registry via ensureConnected(serverId,
registry?).
Adds test-first coverage: suggestions, search, describe and tool-call
leaks across an unassigned same-team server, staleness after
unassignment, and direct-tools scoping.
* feat(mcp): capture and surface server instructions + server-level refresh (#293)
* feat(mcp): capture and surface server instructions
- Add McpServer.instructions column, populated from getInstructions() at
connect time (legacy initialize + modern discover results)
- Surface instructions at three levels like pi-mcp-adapter: truncated
snippet in the mcp proxy tool description, preview in server listings,
and full text via mcp({ instructions: "name" })
- Fix misleading mcp({ server, connect }) usage line to mcp({ connect })
- Share one renderServerListing helper across list and connect modes
- Expose instructions in McpServerInfo DTO and WebUI config card
- Test server supports instructions; service + API tests updated
* refactor(mcp): server-level refresh endpoint, drop agent connect mode
- Replace POST /mcp/servers/:id/tools/refresh with POST /mcp/servers/:id/refresh
returning the refreshed McpServerInfo (server-level, not tools-level)
- Add McpService.refreshServer(): force-close the connection, re-discover and
persist tools, instructions, name/description and status; used by the refresh
endpoint, Test Connection, and post-OAuth completion (replacing discoverTools)
- Remove discoverTools entirely - refreshServer is the single discovery entry
point (testServer and completeAuth route through it)
- Remove refreshTools from UpdateMcpServerRequest/PATCH (dead param)
- Remove mcp({ connect: "name" }) from the agent proxy tool: lazy tool calls
still auto-connect, so agents are unaffected
- WebUI: remove the dialog 'Refresh Tools' button and create-mode background
auto-refresh; add a server-level Refresh button at the left end of the edit
dialog footer; i18n mcp_refresh_server (+ success message)
* feat(mcp): add in-flight request tracking and idle connection cleanup (#294)
* feat(mcp): add in-flight request tracking and idle connection cleanup
* feat(mcp): add background idle cleanup timer to McpService
* feat(mcp): update default background idle cleanup check interval to 30s
* feat(webui): improve MCP server card footer, instructions, and refresh dialog freeze (#295)
* feat(webui): improve MCP server list card, instructions dialog, and refresh freeze UI
* feat(webui): refine MCP card action menu items
* docs(webui): update direct tools mode description in i18n messages
* style(webui): refine Direct Tools Mode layout in MCP server config dialog
* feat(webui): remove tool search box from MCP server config dialog
* fix(mcp): make in-flight idle tracking test deterministic
The 'tracks in-flight requests and safely prunes idle connections' test
relied on a wall-clock sleep (10ms) to guess when the tool handler had
started executing on the server. On slow/loaded CI runners the HTTP round
trip through the real MCP client/server stack can exceed that window, so
resolveSlowTool() fired the no-op placeholder, the gated handler never
completed, and callPromise hung until the 5000ms test timeout.
The handler now signals via a deferred promise when it actually starts,
and the test awaits that signal (racing it against the call promise to
surface early failures) before asserting in-flight state and releasing
the gate. Also gives the test an explicit 10s timeout.
Disabling a skill in the agent config dialog only removed it from the
agent_skills join table, but the chat flow still advertised ALL team
skills to the model in the <available_skills> system prompt section and
the read_skill tool never checked per-agent enablement.
- getAgentChatContextActivity / getAgentAutofillContextActivity now only
return skills explicitly enabled for the agent (agent_skills), and
expose enabledSkillIds in the execution context.
- createReadSkillTool accepts enabledSkillIds and rejects read_skill
calls for skills not enabled for the agent, even if the model knows
the skill id from earlier conversation history.
- Added unit tests for context filtering (chat + autofill), tool-level
enforcement, and the zero-enabled-skills case.
* feat(webui): remove original-file display fallbacks, use proxies only
WebUI now never uses the raw original file for display. All viewers
(image, video, compare, pdf) render exclusively from transcoded proxies;
when no proxy exists the viewer shows an unavailable placeholder instead
of falling back to media.original.
- getBestTranscode: drop the isRaw fallback candidates
- video-viewer / compare-video-pane: remove the synthetic 'Original'
resolution entry and the original.downloadUrl hasMedia gate
- image-viewer / compare-image-pane: remove original.downloadUrl default
- pdf-viewer: require pdfTranscode.url (no original fallback)
- download menus (video-control-bar, breadcrumb-nav): use
media.original.key for the 'Original' download option instead of the
isRaw transcode marker entry
Backend:
- transcode-video workflow no longer pushes the isRaw:true marker entry
into videoTranscodes (it pointed at the original file key)
- remove media.original.downloadUrl from the DTO, asset service mapping,
MediaInfo DB type, and transcode workflows/activities (it was only
consumed by the removed display fallbacks; downloads already use key
via the download-url endpoints)
Tests/fixtures updated; docs updated to state the proxy-only display
policy.
* refactor: remove isRaw field via data migration, clean legacy media JSON
Add a Prisma data migration that cleans legacy media JSONB on both
assets and watermark_files:
- drop the raw-original marker entries (isRaw: true) from
videoTranscodes / imageTranscodes — they pointed at the original file
key and were only used by the removed display fallbacks / download
menus, which now use media.original.key directly
- strip the isRaw field from remaining transcode entries
- drop original.downloadUrl (removed from the AssetInfo DTO)
The migration is generated via prisma migrate dev --create-only and
applied automatically by migrate deploy in production. It is guarded by
a test (packages/db/src/clean-media-migration.test.ts) that seeds legacy
media JSON, runs the migration SQL, and asserts the transform on both
tables.
With the data cleaned, remove the isRaw field entirely from the codebase:
- DTO schemas + ImageTranscode/VideoTranscode interfaces
- PrismaJson MediaInfo transcode types
- watermark activity isRaw skip (and its test), watermark proxy mapping,
agent-embedding transcode selection, webui transcode filters, and all
test fixtures
AGENTS.md: clarify that the no-manual-migration rule applies to DDL;
data migrations (pure SQL transforms, no schema change) are allowed
inside a --create-only generated migration and must be tested.
Add an optional 'context' parameter to the create_file and create_version
agent tools so the chat agent can pass generation/source context (e.g.
'Generated using gemini') to the AI metadata autofill workflow.
- Add Asset.autofillContext column, persisted in the same transaction as
asset creation when context is provided
- Add new list_autofill_fields agent tool exposing autofillable metadata
fields (name, type, description, options) per project
- Autofill workflow reads the stored context and includes it in the
autofill AI prompt
- Add short chat system-prompt guidance to use list_autofill_fields and
pass <=50 word context when creating files/versions
Add an optional 'data' field ({name, content}) to the create_file agent tool
so agents can create files without a local file (e.g. when bash is not
available). The existing 'path' upload mode is preserved; exactly one of
'path' or 'data' must be provided.
Extract mime detection helpers (getFileMimeType, readFileMimeType) into a
new file-mime module, keeping utils/mime.ts free of fs/path imports so it
stays compatible with the Temporal workflow bundle (agent-autofill imports
getProxyType from it). Refactor create_version to reuse the extracted
helpers.
Content uploads reuse s3Service.putObject; mime type is derived from the
filename extension with a text/plain fallback, and the name is sanitized
for the S3 key.
Two bugs found while testing skill loading as a non-owner:
1. The injected bash tool was never actually usable: AgentHarness.setTools
only updates the tool registry unless activeToolNames is passed, so bash
was missing from the active tool set sent to the model on the next turn.
Pass the full post-update tool list as activeToolNames.
2. The active_tools_change session entry (written by setTools) has no chat
representation, so mapEntryToMessage fell back to fabricating an empty
user message that appeared as a stray empty bubble in the chat stream.
Render active_tools_change as a typed custom message and skip entry
types that have no chat representation instead of emitting empty text.
A userId that is not a team member previously resolved to isOwner=true
(role === undefined || role === 'owner'), bypassing bash restrictions.
Treat only flows without a user context (autofill etc.) and explicit
owner memberships as unrestricted; unknown users are now restricted.
Editor/reviewer users can no longer ask the agent to run arbitrary bash
commands. Bash execution for them is gated behind approved team skills:
- Add required source param (enum user/skill) to the bash tool; commands
with source=user are blocked when the user is not a team owner.
- Append a Restricted User Context instruction to the system prompt for
non-owner users: never run user-requested bash, only skill-required
commands with source=skill.
- Delay bash tool injection for editor/reviewer: it is only added after
a skill is successfully loaded via read_skill (or restored when a
skill was loaded earlier in the session). Team owners are unaffected.