* perf(api): filter peer-run activity in the transcript page query
The transcript loaded a raw page, dropped hidden peer-run rows in memory and
rescanned when nothing visible was left, so a long delegated stretch cost two
queries per hidden page. The page query now joins runs and excludes hidden
peer activity in Postgres, so every page is filled with visible rows. The
kept block kinds live in one list shared by paging, the in-memory filter and
live event forwarding.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YTJRkJrXAS9e3aEFcFYTy8
* test: port the main-conversation journey fix from #126
Journey 5c fails on main since every space got its own main conversation:
the lookup picked the Lab conversation instead of the home one. Same change
as #126 (including its chart-links warm-up), so this PR can go green; it
no-ops once #126 lands.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01YTJRkJrXAS9e3aEFcFYTy8
---------
Co-authored-by: Claude <noreply@anthropic.com>
While onboarding wrote the main conversation, its sidebar row, project rows and Heute showed the
server preview: nothing for a pending card (a card has no text of its own, so the row fell back
to "Ready") and the English fallback of the closing note, even for a German user.
Groups now carry `previewBlock` next to `preview`: the latest message when it is an onboarding
card or note, validated with the same block schemas the messages use (now named exports). The
web words it with the helpers that render the card and the note, so no copy is duplicated: the
question while a card waits, the answer once given, and the closing note in the user's language
and clock. A dismissed card keeps the plain preview, as the conversation no longer shows it.
The field is left out for every other message, and mobile keeps using the plain `preview`.
The chosen-answer lookup moves from the card into onboarding-copy so the card and the preview
share it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
The coach check-ins are main-bot routines scheduled in the profile time zone, but a later
zone change (About you, onboarding, the main bot's update_user_profile tool, or forgetting
the whole profile) left them in the old zone until the rhythm was saved again, so they fired
at the wrong local time.
syncCoachTimeZone, next to the rhythm code, moves the main bot's coach routines that are not
in the current profile zone into it, recomputes the next run of armed ones and re-arms their
wakeup. Paused check-ins move but stay paused, routines the user made are never touched, users
without check-ins get none, and a save that leaves the zone as it is changes nothing. The
profile update and forget RPCs and the profile tool call it whenever they write the zone.
Quiet hours, the daily nudge cap and coach dates are evaluated in the profile zone when they
are used, so they need no migration.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
Main now gives every project its own main agent. Coaching stays with the
user's personal agent, the main bot of the home space: only it gets the
goal, habit, check-in, brief and nudge tools, the coach summary and the
open-onboarding note, and the tools refuse a project's main bot too. The
onboarding questions start with the personal agent only. A project the
scan creates gets its main agent like any other, and the inbox lists asks
from another space's main conversation.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
* chore: remove legacy orchestrator, trading and one-off script leftovers
The repository still carried the pre-evOo prototype: a Python orchestrator,
the old HTML "evoo-platform" cockpit (whose test ran in the root unit
suite), an apex-trading package, the hot-reload deploy.sh for it, a
permissions manifest, an architecture note with local file links, agent
handoff notes, a Steve Jobs persona template and about ninety one-off
scripts that probe a developer database with hard-coded local URLs. None
of it is used by the product, CI or the real deploy in ops/evo-deploy.
Delete those files, drop the evoo-platform test glob from the root vitest
config, the Cockpit section from .env.example and the dead biome and
gitignore entries, and update the docs that listed the removed tools.
scripts/start-tmux.sh no longer starts the removed sentinel and resolves
the checkout from its own location instead of a machine-specific path;
scripts/seed-ollama.ts requires DATABASE_URL instead of falling back to a
local URL with a username.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
* fix(web): drop trading and persona prompts from the conversation shell
The web shell still showed leftovers from the trading prototype: composer
chips with German trading and "Steve Jobs" prompts, trading suggestion
chips under the empty conversation line, a German trading hero for the
empty transcript, and a German "data protection" status panel in the
sidebar. They suggested actions the product does not have and added
persistent chrome.
Remove the composer chips and their helper, the empty-state chips and the
hero (the empty conversation keeps its existing one-line start hint), the
sidebar trading indicator and the data trust panel. The new-bot E2E now
asserts the plain empty state; the sidebar spec no longer expects the
trading status, and the placeholder Steve Jobs spec with no assertions is
gone.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
Co-authored-by: cynarAI <cynarAI@users.noreply.github.com>
* refactor: remove the trading bot status integration
With the sidebar trading indicator gone, the only remaining consumer of
the trading bot status was a banner in General settings that reused the
same component. evOo is a personal assistant; a status proxy and poller
for one external trading service, configured through service-specific
TRADING_BOT_URL and VITE_TRADING_BOT_URL variables, does not belong in
the core product.
Remove the settings banner and its prop plumbing, the web poller, the
indicator component, the /api/trading-bot/status proxy, the shared core
contract, the env loading and both variables from .env.example, together
with their tests.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
* chore: remove one-off database scripts and a stale review note
packages/db carried four hand-run scripts outside the package build that
query or seed a developer database with a hard-coded connection string,
one of them for a specific account. docs/VOICE-FIRST-REVIEW.md was a
one-time verification note for a workflow that no longer exists and
named private machines and local paths. Nothing references them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
* chore: replace machine-specific paths and addresses with placeholders
ORIGIN.md pinned the intended checkout to an absolute path under a
personal home directory, docs/LOCAL-SETUP.md listed folders on one
developer's machine, and two tests used what looks like a real tailnet
address. Use a home-relative path, drop the machine-specific folder list,
and use 100.100.100.100 in the tests; it is still inside 100.64.0.0/10,
so both tests check the same thing as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
* test(web): drop the trading status from the shell contrast test
main's contrast test now waits for the sidebar trading status before
measuring; this branch removes that status.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
---------
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: cynarAI <cynarAI@users.noreply.github.com>
Each project (space) now has a main agent and a main conversation of its
own. The user talks to it, and it delegates to the project's specialists.
Creating a project sets both up and opens the conversation, and deleting
an otherwise empty project removes them with it.
Claude-Session: https://claude.ai/code/session_01R8hSsKTB3W3e61wD6kvzur
Co-authored-by: Claude <noreply@anthropic.com>
The onboarding scan keeps the consent-first rewrite. Since main's folder
check moved out of wireProjectWorkspace, applying a scan now checks that
each chosen project is still a folder (not only inside the scan folder)
before anything is written; a discovery test covers a project replaced by
a file after the scan.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
spaces.create created the space first and only then checked the folder,
so a folder that was missing or not a folder returned 400 but left the
new space behind. The New project dialog shows that error, so fixing the
path and submitting again created a second project with the same name.
The folder is now resolved and checked right after the deployment-owner
check, before the space exists, and the space stores that resolved path.
wireProjectWorkspace takes the resolved folder instead of resolving it
again after the space was created. The onboarding scan, its other
caller, resolves and checks each repository folder the same way before
creating the project, and skips a folder that cannot be linked.
The authorization test has the deployment owner submit a missing folder
and a file (both 400) and then the real folder under the same name, and
checks that exactly one project with that name exists.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012HG4pFTUBDpuCS2Y44TV8x
The consent-first scan replaces the automatic project creation that main
fixed in #116; main's git checkout fix and sidebar classes are kept.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
spaces.create linked any workspacePath to the new project, and
computer.files and computer.readFile then listed and read files under
that folder for anyone in the space. On a deployment with signups, any
member could link / or the API's own checkout and read whatever the API
process can read, such as its .env file. Linking also ran git in the
chosen folder and stored its remote URL.
A linked folder lives on the host, and the host is the deployment
owner's machine, the rule the onboarding scan already follows:
- spaces.create refuses a folder from anyone but the deployment owner,
before the space is created or the folder is touched.
- Project files are listed and read only for the deployment owner, so a
space a member linked before this change stays closed too. Space
navigation no longer shows a linked folder's path and git remote to
other members.
- A linked folder is stored as its real path, and listing and reading
follow symlinks only while they stay inside it. Symlinks that leave
the folder or lead nowhere are left out of a listing instead of
failing it.
- The web New project dialog offers the folder field only to the
deployment owner.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EyQw3AEaHYXegvgvTPAsi1
The scan's project phase called prisma.space with fields the Space model
does not have (title, path). Both calls threw, a per-repository
try/catch swallowed the errors, and no project was ever created. With
the field names fixed, the space would still have had no organization
or membership, so nobody could have seen it.
The phase now goes through createSpaceForMember and
wireProjectWorkspace, the helpers behind spaces.create: each repository
becomes a project of the deployment owner in their current
organization, linked to its folder with its git remote, branch and
head. It still runs only in the owner's scan with ONBOARDING_SCAN_HOME
set. A repository whose real folder one of the owner's projects already
links is skipped, so scanning again creates nothing; the phase stops at
the per-member space limit; and a second start while a scan runs (a
StrictMode remount, a retry) joins it instead of linking everything
twice. Discovery no longer descends into dot folders, where shell and
editor plugin checkouts would otherwise become projects. The Prisma
client is typed, so a field the schema lacks is a compile error again
instead of a swallowed one.
readGitCheckout loaded child_process with require(), which does not
exist in the ES module runtime the API runs in, so linking any folder
with a .git directory threw "require is not defined", in the scan and
in the New project dialog alike. It now imports spawn statically; the
module is only reachable through its server-only subpath, so browser
bundles still never load it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ebcq9seT9eCoHtqxuWEUzg
AGENTS.md asks for top-level `import type` instead of inline `type`
specifiers; this applies it to the import statements the coach work added
or changed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
The onboarding cards, the closing line and the coach routine names were English server text
for every user. The server keeps the state and an English fallback; the web now words them
through Lingui, so there is one translation source.
- Choice cards: the web says the question, the option labels and the placeholder from the
step and option ids, and keeps the server text for anything it does not know (quiet-hour
ranges are already written in the user's locale). Another bot's focus card is recognized
by its option ids. Mobile and agent history keep the server text.
- The "Got it. … are a good place to start." line is gone: the app cards speak for
themselves, and with no cards nothing is said.
- The closing line becomes an intake_note block (note "done", nextCheckInAt, English text).
The web words it with the first check-in in the user's clock; mobile blockText and
blocksToAgentHistoryText read its text through their generic fallback.
- Routines carry coachKey. The routine list and editor show a check-in under its label in
the UI language until the user renames it; saving an untouched label keeps the stored
name. The English names move to contracts (COACH_ROUTINE_NAMES); prompts stay English.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
AGENTS.md asks for top-level `import type` instead of inline `type`
specifiers; this applies it to the import statements the media work added
or changed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
The main conversation lives in the home space, but the user can have another
space selected while it is on screen, and the card's actions sent that space:
media.jobs.start resolved the conversation in the selected space and failed,
and media.jobs.get/cancel only found jobs of the selected space.
The server now takes the space from the card itself. start finds the card's
message in one of the actor's own threads (the conversation named by botId or
groupId), requires the actor to still be a member of that thread's space, and
resolves the conversation and the source artifacts there. get and cancel find
the actor's job by id and require membership of the job's space.
media.capabilities takes an optional job id and answers for that job's space;
the card asks through its job and shares the answer per conversation. Every
other case is a 404, as before.
A Postgres test runs a card in the home-space main conversation while another
space is selected (start, get, capabilities, cancel) and checks that another
user and a user who left the card's space get 404 on every procedure.
This also completes the media settings section committed earlier as work in
progress (01eea83); its code was final, only that commit message said
otherwise.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
Node's default 5-minute requestTimeout cut off a 500 MiB video upload on
any uplink slower than about 14 Mbit/s, so the video cap was unusable over
ordinary connections.
- The API server's requestTimeout is 30 minutes, enough for the largest
upload at about 2.5 Mbit/s. headersTimeout keeps its 60-second default.
- Each streamed upload aborts when the client sends no bytes for 60
seconds: the route answers 408, the partial file is deleted and no row
is written, so a slow-drip client cannot hold a connection for the full
request timeout. Time spent writing to disk does not count.
- The store and the upload guard no longer wait for a body to confirm its
cancellation, which a stalled or hostile client may never do.
- A guard test checks that every file the executor attaches records the
thread's group, which the render_plot chart fallback once dropped.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
Dictation falls back to the owner's transcription server when the voice
provider cannot transcribe, so that server is a voice recipient too: it is
listed for voice use (as "Transcription (<origin>)") whenever it would take
the recording, and not when the voice provider transcribes itself.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
Brings in profile editing, the cross-space inbox and consent-first discovery. The wizard
keeps both changes: no preferences step, and the new scan step without the main bot lookup.
The harness lists both new onboarding Postgres suites.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
- onboarding-intake.spec walks focus, goal (after typing into the composer, which still
works), check-ins, quiet hours and the notifications card when the browser offers it,
with the onboarding-intake-goal, -rhythm and -done screenshots.
- completeOnboarding and the wizard specs drop the preferences step; the scan step is
optional (skipped when it asks, otherwise it advances by itself).
- The user-profile spec now checks the browser-filled profile and the focus fact; the
conversation spec checks the shorter app suggestion and the following goal question.
- The hover spec measures a reply that fills its row (a bot row's rail sits at the row's
end, so the old long onboarding line was what made it pass) and scopes Send to the
composer, since an open goal card has its own.
- Mobile shows a choice card as its question and answer instead of an empty bubble.
- The focus options drop their unused summary.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
Brings in ffmpeg (probing, extractAudio) and the video operations. The
transcribe registration now gets the ffmpeg audio tools, so video is
transcribed through extractAudio with the demuxer forced from the MIME type,
and long recordings are converted and uploaded in ranges.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
- The GPU guard waited for free VRAM at the operation's floor even on a card
whose total memory is below it (an 8 GB card and a 12 GB video floor), so
such jobs deferred until their budget ran out. The requirement is now capped
at 90% of the card's total; a card with less than half the floor fails at
once with "The GPU does not have enough memory for this job."
- Tests pin that media tools reach a run only through the space's
capabilities and within the bot's allowlist now that the image tool
definitions exist.
- Type-only imports are top-level `import type` in the media runtime, tools,
RPCs and tests (AGENTS.md).
- The ComfyUI settings test gives saveMediaSettings a store with nothing saved,
since the save reads the previous settings before validating.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
The card's follow-ups now reach the media job RPCs, and it only offers what
the space can run.
- media.capabilities answers which operations the actor's space can run
(booleans only). The card asks once per minute per space, and only when it
shows finished video or audio, so other cards cost nothing.
- Export and Transcribe call media.jobs.start with the card's message, so the
same bot posts the new card; Cancel calls media.jobs.cancel. A refused
start shows the server's message under the card.
- New files use top-level import type.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
The onboarding scan started on its own, discarded what it found, claimed
to read email and calendars, simulated progress for everyone who could
not scan, and its project creation always failed (wrong Space fields and
a space lookup by title across tenants).
Discovery now asks first and creates nothing without a choice:
- The scan step appears only for the deployment owner when
ONBOARDING_SCAN_HOME is set; everyone else skips it. The simulated
scan is gone.
- It shows the folder and asks. Only directory listings are read: git
projects up to four levels deep, their languages and a few tools
(Docker, GitHub Actions, Terraform). Symlinks are never followed and
each chosen folder is resolved again before use, so nothing outside
the folder is reached. Results stay with the account that scanned.
- Found projects are a checklist (none preselected); found languages and
tools are facts (preselected). "Add" creates one project space per
chosen project with createSpaceForMember and wireProjectWorkspace in
the user's own organization, reusing the user's own space already
linked to that folder, and saves the chosen facts as onboarding facts
under the profile lock and fact limit.
- When a connection becomes connected (direct connect, authorization
round trip, or catalog reconciliation), "Uses <App>" is remembered as
a tools fact from the connector. Revoking keeps it.
The E2E harness now points ONBOARDING_SCAN_HOME at a fixture folder and
starts without a deployment owner, so owner-only onboarding steps appear
only in specs that take the seat through /__e2e/deployment-owner.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
Uploads with a text/plain, text/csv or text/markdown body are CORS "simple"
requests that a browser sends cross-site without a preflight. The session
cookie's SameSite default already keeps it off such requests, but a
deployment that relaxes it would accept uploads from any page. The route now
applies the same origin rule as CORS before reading the session.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
Brings in the media job RPCs, generic media tool dispatch and the image
operations; transcribe joins the operation list and the tool catalog next to
them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
Agents only received the documents of a turn on their computer and could
attach workspace files up to 10 MiB, so they could neither cut a clip a user
sent nor post the video they rendered.
- Current-turn images, video and audio are copied into the bot's workspace
(Team Computers under bots/<botId>/) like documents. Remote computers move
files as base64 JSON, so anything above their 12 MiB transfer cap stays in
storage and the agent is told it by artifact id instead of failing the
turn; local computers copy up to the kind cap.
- attach_file accepts video and audio by extension and attaches up to the
kind cap when the store streams (still bounded by the transfer cap of a
remote computer), stored through the shared artifact helper with a real
SHA-256.
- The render_plot fallback that attaches a chart PNG now records the group,
so the chart opens in group conversations.
- Agent history names image artifacts by id, like video and audio, so a
later turn can refer to them.
- artifacts.get refuses files above 25 MiB: base64 reads hold the whole
file in memory, and larger media is served by the artifact route.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
Media tools and card actions send prompts, images, video and audio to the
deployment owner's ComfyUI and transcription servers. Like the embeddings
server behind built-in memory, those self-hosted endpoints are recipients in
the AI data-sharing disclosure: they appear with the model uses ("ComfyUI
(<origin>)", "Transcription (<origin>)", keyed by the full endpoint so a new
server asks again), and starting a card job asks for model consent on clients
that run the foreground check.
Also adds unit tests for the media settings store: URL normalization and
refusals, encryption, the key-free view, one decryption per saved version, key
retention per server and removing empty settings.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
With the media runtime in place, transcription becomes an operation like the
others: bots call the transcribe tool, the job runs in the background and
the card shows the subtitle files.
- createTranscriptionOperations registers transcribe: available when a
transcription server is configured, refuses artifacts the user cannot
read before the job exists, and writes <name>.srt and <name>.vtt next to
the source (a .txt when the server returns no timings). With ffmpeg the
audio is converted to 16 kHz mono WAV and long media is transcribed in
upload-sized ranges on one timeline; without it audio is sent as stored
and video is refused with a clear message. A server that is down or busy
(unreachable, 429, 502, 503) defers the job instead of failing it, since
nothing has been stored yet. The bot gets the transcript, cut at a word.
- The transcribe tool definition uses the shared media schema helper and
continue_after; dispatch stays generic.
- The main bot's run creates Studio once a media server is configured,
right before the specialist directory is built.
- Dictation: when the user's voice provider cannot transcribe (or there is
none), /api/voice/transcribe uses the self-hosted transcription server,
and the voice status reports server dictation so the web app uses it
instead of the browser's speech recognition. A voice provider that
transcribes keeps being used as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
Bots and users need to start and follow media jobs, and the deployment owner
needs to point the deployment at their GPU and transcription servers.
- Media tools are named like their operations. A run sees one only when the
space can run that operation (MediaRuntime.capabilities); the per-bot
allowlist still applies. applyTool hands every media tool to one
dispatcher (media-tools.ts) that splits off continue_after, starts the job
under the tool call's effect key (a replay returns the same job) and returns
at once; refusals (invalid input, unknown artifact, unavailable) go back to
the model as tool errors.
- media.jobs.start starts an export, transcription or image edit from a card:
the conversation is resolved like every thread procedure, the card's bot
(or the group's lead) posts the new card, and the runtime refuses source
artifacts outside the actor's space. media.jobs.get and media.jobs.cancel
only see the actor's own jobs.
- media.settings.get/update are deployment-owner only, never return keys and
answer invalid URLs with a 400 that says what is wrong.
- createApp accepts media operations as a test seam.
- Tests: a Postgres suite over the RPCs (owner-only settings, encryption,
key retention per server, URL policy, start/get/cancel isolation between
users, card jobs end to end) and unit tests for tool exposure and dispatch.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MRtGG9ARfqqhma3t85Ro2u
