The bot's id never reached its prompt: `composeBotPrompt` only took its
instructions, the appendix, its memory and its workspace, and
`grabBotPrompt` never passed `bot.id` or `bot.name` on. A bot could only
learn its id by accident, from a tool description that named it, so
"which bot am I" was unanswerable in most contexts.
- Open the prompt with who the bot is (`composeIdentity`): its name,
whether it's the coordinator or a worker, its role, and its id, said
to be how tools and refs point back at it. `grabBotPrompt` now passes
the bot, which covers chat replies, tasks, routines and delegated
tasks, since both runners build their prompt there.
- Turn the notes' example refs into functions bound to the running bot:
`chatReplyNote`, `botFilesNote` and `coordinatorReplyNote` take its id,
so the folder-ref example no longer points at a hardcoded `owner_id: 5`
(another bot's folder) nor `[bot:3]`. ComputerFilesNote keeps the
computer it's given.
- Stop `create_task` from hiding the worker's own id in its own chat.
Tests: four new `composeBotPrompt` cases pin the identity block (worker,
coordinator, unnamed, no bot at all), and a new `harness-prompts.test.ts`
checks each note's own-folder and self refs point at exactly the running
bot, for three ids.
The two limits were compiled in: a task run stopped at
AppData.HarnessMaxToolSteps (200) and a chat reply at
AppData.BotReplyMaxToolSteps (8), so an install that needed a longer task
had to rebuild.
Both are now `app_settings` rows (`harness_max_tool_steps` and
`bot_reply_max_tool_steps`), read at the two loop call sites through
grabHarnessMaxToolSteps and grabReplyMaxToolSteps, and edited by the super
admin on Settings > General beside the prompt appendix. Zero clears a
limit back to the code default, like the bot file quota.
readStepLimit keeps the saved number only when it is a whole number of at
least 1 and no more than AppData.BotStepLimitMax (1000), so a blank, a
zero or a typo falls back instead of stopping every run before its first
turn, and a huge number still ends in the wrap-up report.
runTextLoop keeps AppData.HarnessMaxToolSteps as its default, so it stays
free of database reads and its tests run as they did.
Nothing anywhere recorded what a model call cost: the harness used a
provider's usage numbers only to decide whether a turn had finished, then
dropped them. So a reply could not say what it took to make, one model
could not be compared with another on price, and there was no way to ask
what a day, a week or a month of the crew's work came to.
- `token_usages` records one row per call: user, bot, workspace, session,
job and message, the harness, the provider and model that actually
answered, the prompt and completion counts, and the cost when the
provider reports one.
- `harness-usage-scope.ts` carries those ids in an AsyncLocalStorage
scope that `run-bot-job.ts` enters once, so a run's turns, its session
summary, its transcriptions, images and embeddings are all attributed
without threading ids through every harness function. Nested runs merge
rather than replace, and a null genuinely clears a field (a bot's own
chat has no workspace).
- Work with no run behind it is charged to its owner instead:
`run-with-fallback-scope.ts` reaches the workspace's creator else the
bot's, and deliberately does not override a real run, so an ingest a
bot asked for is not misattributed to the bot's owner.
- All six harnesses record through `withCandidates`'s `on_success`.
Recording never throws: a model that answered must not lose its answer
to a failed insert.
- A job's calls land before its message exists, so `complete-bot-job.ts`
back-fills `message_id` once the message is inserted.
- The message's actions row shows what it cost, beside the timestamp and
the play button, with the breakdown in its title.
- `/admin/usage` totals the install's spending for 7, 30 or 90 days: the
total against the previous period, a day-by-day chart, a split per kind
of call, and a table per model with the price per thousand tokens
worked out from cost and tokens, which is what makes two models
comparable. Crew managers only, since it is the whole install's work
rather than one user's conversation. It sits with Routines and Activity
in the user popover rather than under Settings, because it is a record
of what happened, not something to configure.
Days are grouped in local time (`date(created_at / 1000, 'unixepoch',
'localtime')`), matching the local midnights the window builds and the
`YYYY-MM-DD` buckets it looks up. Grouping by the bare UTC day put a late
evening's calls under a name no bucket carried, so they vanished and the
chart's total silently disagreed with the totals card; a test pins it.
`canSeeRailItem` now decides who sees a popover or rail entry. The two
each filtered with their own inline rule and only one understood
`can_view`, so a gated page could be offered in the popover to everyone.
The thinking row and the header menu both carry Stop while a bot has work
open for the user anywhere: every pending or processing job of that bot in
their conversations is failed as stopped, its task closed, a Stopped step
recorded and one closing message posted per session, so the thread stops
waiting and the bot goes idle. The handler running the job is aborted from
this process through a registry, so a job stuck waiting on a model stops at
once instead of on its next heartbeat, and a late face change after the
abort is dropped.
The release stage is now said in the app too: Settings > General reports
"Cobble v0.1.5 (alpha)" next to development, and the README and RELEASING.md
say what alpha means for installs.
A coordinator asked to set an uploaded logo as a workspace picture could
not: the picture tools take a media id, and nothing ever gave the bot one.
Attachments reached the model as "[Attached image \"logo.png\"]", every
file tool answers with a [file:ID] ref, and an SVG is deliberately stored
as a document rather than an image.
- copyMediaAsPicture copies an approved image into a new picture row
(bot-image/workspace-image). The source keeps its row and bytes, so a
message that attached it still loads and changing a picture can never
take a file away from a conversation.
- grabPictureSource resolves a picture by [file:ID] (the same scope gate
every file tool uses) or by the user's own upload; parseFileRef reads
the [file:12] form.
- The new set_picture tool ships on every run: a bot sets its own face,
the coordinator can also dress a worker. update_workspace gained
picture_file_id, and setBotPicture/setProfileImage share the copy and
the delete.
- removeReplacedPicture deletes a replaced picture only when it really is
one, never a chat upload.
- uploadAndRecordMedia writes images in their own format: PNG stays PNG
with its transparency, WebP stays WebP, GIF becomes PNG, everything
else JPEG, and the row records the real mime type (it used to say jpeg
for every image).
- Pictures change live: BotUpdatedData carries image_media_id and a new
server:workspace-updated pushes workspace pictures, relayed between
processes.
- Attachments are labelled with their media id, and BotFilesNote tells a
bot that a [file:ID] can become a picture.
- grabRunScopes is now the single source of a run's reachable folders,
shared by the file tools, imports and the picture tools.
`text_content` is an extracted-text cache, but nothing stopped it from
holding a dump of bytes:
- `fileToText` ended its switch with `default: return buffer.toString("utf-8")`
and defaulted a missing extension to "txt", so any binary was decoded.
- Both extraction callers gated on `media_type == "document"`, and
`attachmentFileKind` classes everything non-image/audio/video as a
document: an SVG, a zip, a spreadsheet, a .bin.
- Nothing capped the length, so a 20 MB attachment could leave 20 MB of
text in the database.
- `storeChatAttachments` copied the text onto a second `media` row, and
`SELECT m.*` pulled the column into page props and API responses.
On a real install an SVG held 6852 characters of markup and a
`.designcraft` file held 142 KB of raw bytes.
Changes:
- `file-to-text.ts` lists the text extensions explicitly and returns
undefined for anything else; no extension is no text.
- `can-extract-text.ts` is the one allowlist (reusing
`BotTextFileExtensions`), used by both upload routes and by
`uploadAndRecordMedia`; `extension-of.ts` is shared with the storage
catalog instead of duplicated.
- `BotAttachmentTextStoreMaxChars` (200k) caps the column via
`cap-text-content.ts`.
- `store-chat-attachments.ts` no longer copies the text; ingestion
re-reads the bytes once.
- `media-columns.ts` selects every media column but `text_content`, which
now has to be asked for (`grab-messages-media`'s reply path and
ingestion do); the media CRUD routes sanitize rows they write.
- `ingest-bot-file.ts` decides on size alone, since a cached slice made
an oversized file look already read.
- `bun run media:prune-text` cleans rows that were already stored
(rerunnable, with a VACUUM hint).
Verified on the dev database: both bad rows cleared (146 KB), rerun
reports nothing to do.
A skill is a runbook: a name, a unique slug, a one-line description of when
it applies, and the markdown body. A run only ever sees the slug and
description (one load_skill tool carries them), and reads the whole body
only when it calls the tool, so a long runbook costs nothing until needed.
Skills are the third extension kind beside MCP servers and custom tools, on
the same botExtensionsSql path, so per-bot, per-workspace and all-bots
grants work the same way, and attachDefaultExtensions hands an every-bot
skill to new bots. load_skill is a reserved tool name, so a custom tool
cannot shadow it, and a built-in wins a clash.
Managed on Settings > Skills; ExtensionRow links each kind to its own page
and a skill never reports a failure, since nothing connects or runs.
A bot's identity was one of seven hues, and the hue picked both its colour
and its stone shape. Now its colour is free and its stone is its own:
- `bots.color` holds any `#rrggbb`, `bots.stone` the flat-cut shape it's
drawn on. `bots.hue` stays as the palette column they replaced, read only
as a fallback; `src/scripts/move-bot-colour-to-owners.ts` backfills an
install from it (safe to rerun).
- `src/utils/bot-color.ts` is the only place a colour is validated or
resolved (`resolveBotColor`/`resolveBotStone` carry the old hue forward,
`botColorNeedsPaperFace` flips the eyes to paper on a very dark body), and
`src/utils/bot-stone-props.ts` gives every row its props, so a row saved
before this still renders as it did.
- `BotStone` takes the colour as the body's SVG `fill`, the one documented
exception to the token rule: an arbitrary hex has no token to point at.
`UnreadDot` follows it.
- The picker keeps the palette tiles (still shape and colour together) and
adds a native swatch and a hex field, so a custom colour keeps its stone.
Same place as before: the header avatar's modal and Settings > Bots, for
crew managers. The coordinator stays hue-less.
- A recolour reaches other open tabs through `pushBotUpdated`.
There is no public media any more. `/media/public/<id>` served a media
row with no session at all, so anyone with the id could read a file.
The public/private split is gone: the path is `/media/<id>` (and
`/media/<id>/thumbnail`), every request is signed in first, and
`grabUserMedia` decides whether the user may see it: their own upload,
a bot's or workspace's picture or file they reach, or a file on a task
they reach. Anything else is a 404 before a path is built.
The `media.is_private` column and the two per-user directories went
with it: uploads live in `media/<user_id>/` and there is nothing to
pick between. The files browser's media volume is now the viewer's
own folder whatever their role.
Bots also say which of the two things they mean. A `[file:ID]` ref is a
saved file and its chip opens the bytes; a folder is a new
`[folder:volume:owner:path]` ref (`src/utils/folder-ref.ts`) whose chip
opens the files browser there and downloads nothing. The old `[path:...]`
ref and the extension sniffing that guessed file from folder are gone,
so `import_file` names the folder holding a file it copied rather than
the file itself.
The prompts say what the Assistant got wrong: list before importing or
saving, one copy per artifact, never split a file into parts unless
asked, never claim the store is empty without a fresh list, and never
invent a size cap. A re-import now says the path already existed
instead of silently overwriting and reporting a fresh copy.
Envs had three scopes, general, workspace and computer. A bot had no
scope of its own, so the only way to give one worker a key was to set it
on a computer it happened to share.
The envs table gets a bot_id, and resolution now layers bot over
computer over workspace over general, so a bot's own key wins wherever it
runs. Every computer call site passes the run's bot_id, and grabRunEnvs
does the same for the {{env.KEY}} a custom tool or an MCP server's headers
read.
A bot's and a workspace's settings page can now edit their own envs, in
place, through the same save-env and delete-env routes: sending `scoped`
makes them answer with just that scope instead of the whole install's
overview. Super admins only, like the Environment page.
Both runners now pass the files a run wrote back into the reply, and a
computer file ref renders as a chip like a bot file ref does, so a
reply can point at what it produced instead of naming a path.
import_file sent a computer's bytes back with base64 over /v1/shell/exec,
whose stdout is capped at 30,047 characters. Every import was therefore
cut at 22,528 bytes and saved anyway, with the tool reporting success: a
1.12 MiB archive became a 22 KB file that would not open, and a crawler's
crawl.json lost half its pages.
The bytes now come from the sandbox's streaming GET /v1/file/download.
The size is read first so nothing over BotFileImportMaxBytes is
transferred, the download asks for change_policy=abort so a file that
changes mid-send is refused rather than half-saved, and the received
length is checked against the size read, so a cut-short transfer fails
loudly. Folders go the same way, as a gzipped tar unpacked into one row
per file.
Bots also name a path as [path:volume:owner:path] now, so what a bot
leaves on a computer is a link into the files browser instead of text.
Bots and workspaces each keep a list of work tracked until it's done, and
the bots move it along themselves as they work.
- tasks (one collection per bot or per workspace), task_notes (the
progress log) and task_media (a deliverable or a reference file, its
bytes a private media row served at /media/private/<id>)
- statuses todo/doing/blocked/done/cancelled, with blocked_reason while
only the user can unblock it
- visibility as SQL in src/utils/task-access: a user's own bots and
workspaces, a run's own list plus its workspaces, everything for the
coordinator; every tool and endpoint resolves through it
- list_tasks, create_task and update_task on every run, so a chat reply,
a task and a routine all read the list, log progress and file what they
produced
- TasksNote on every run and TaskReminderNote in chat, so a greeting or a
proposed next step brings pending work up in one measured line
- tasks group on the workspace dashboard and a tasks modal in the
conversation header, sharing one form, row and detail panel
- a bot's or workspace's own chat uploads are read before the delete so
none are stranded; bot files keep their folder
Also corrects the README's RAM guidance and table alignment.
Scheduled runs are pre-approved: they carry out their brief instead of
drafting for confirmation. A run that can't do its brief calls
report_run_failed and is retried 5 and 10 minutes later; only then does
the user get one "failed after N tries" message, and the routine keeps
why in last_error.
Every reply and task now gets ActionCheckNote: look for an existing
record before creating one, and verify an action took effect before
saying it's done. Memory and instructions saves are read back, and
create_bot, create_schedule, start_task and delegate_task refuse
duplicates.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
The composer now writes markdown with bold, italic, strike, code,
links, quotes and lists, from shortcuts, markdown typing or a bubble
over the selection. @ mentions a bot and # a workspace, shown as chips
and saved as [bot:ID] / [workspace:ID]. Text selected in a message can
be quoted into the composer. User messages render as markdown, and
replies tell the bot who and what the user mentioned.
Adds an ESM stand-in for use-sync-external-store, since bunext keeps
React external in page bundles and the CommonJS original requires it.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Every message has a row under it to reply or react. Replying quotes the
message above the composer, and the sent message stores
reply_to_message_id (a plain id, not a foreign key, so deleting the
quoted message only drops the quote). send-message accepts a quote from
any session of the same conversation, so a rollover or a thread that
moved on still sends; the conversation drops a picked quote when its
session changes. The bot reads the quote as "[Replying to ...]" before
the user's turn, file-only replies included, and relayed messages carry
it too.
Reactions are a fixed set (BotMessageReactions) in
bot_message_reactions, one per message, emoji and user or bot. Users
toggle their own through api/admin/bots/react-message, and every bot
gets react_to_message for the session it answers (the user's latest
message by default). Both push the fresh list as server:bot-reaction,
with mine worked out for the session's user.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
- Resolve the crew and workspaces fresh inside each tool run, so a bot or
workspace created earlier in the same reply is usable: create_bot,
assign_computer, add_bot_to_workspace, delegate_task and
set_bot_workspace_instructions.
- Add coordinator-only update_bot, find_bot, find_workspace, delete_bot,
delete_workspace and remove_bot_from_workspace tools.
- Push server:workspace-removed when a workspace is deleted, and
workspace_ids on server:bot-updated so a workspace's crew drops a bot
that was taken out of it.
Sessions no longer take their title from the first message: once the first exchange is sent, the text model writes the title and summary, which the rail and sessions list show. The summary records how many messages it covers so a stale one is rewritten before it is carried into the next session.
The sessions hook now reloads on live message and session events, so a session started by the first message, a rollover or another tab appears without a reload. New sessions are pushed as server:bot-session, both from the process that changed them and through the live relay.
Add a generated manifest, icon set, offline page and service worker that never caches authenticated HTML, plus install and update UI. Turn the rail into a drawer below md with container-query internals, dvh and safe-area handling, and reconnect the socket after visibility/network changes. Add Web Push subscriptions, presence-aware delivery, notification settings and tests.
A reply that works through tools now shows its latest tool step under
"is thinking…" (not its closing note, which is the answer). While a reply
is due, the open conversation checks the session's latest messages every
few seconds (api/admin/bots/session-messages) and replays any it missed
as server:bot-message, so the thread, rail and unread all take it and the
thinking row can't stay up until a reload. MCP and custom tool steps read
in words ("Using Exa search (web search exa)").
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Replies were locked per bot, so a long reply in one session held up every
other session's reply with nothing shown. The lock is now per session. A
chat shows the thinking row whenever the user has the last word, rather
than from the bot's face (one per user, shared by every session), so it
shows even while the bot works elsewhere. A task waiting for its busy
computer says so, and a job finishing while another still runs puts back
the running one's face.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Two new settings pages, MCPs and Tools. Super admins define them; super
admins and admins attach them to bots, to workspaces (every bot there),
or to every bot. Bots get them in chat replies and tasks.
- Custom tools: a JSON-schema of arguments plus an HTTP request, a script
on the bot's computer, or a prompt, templated with {{args.x}} and
{{env.KEY}}, secrets redacted from results.
- MCP servers: remote (URL and headers) or stdio commands run on the host
with a setup command and an isolated env, in a per-process pool. Tools
are listed into a cache and connect on first call.
- Add, edit, access and try open in modals.
- Markdown, JSON and command fields use twui's AceEditor (CodeEditor),
its scripts loaded on demand.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
create_workspace gives the coordinator a way to start a workspace with its shared context, and the tool list is shared, so a new id can be used by create_bot or add_bot_to_workspace in the same reply.
A coordinator's tasks now get create_workspace and update_workspace too: the research task that had the material could only report it, never write it into the workspace, so the coordinator promised an update it had no turn to make.
A workspace created from chat is pushed as server:workspace-created, from the tool and from the live relay, so the rail shows it without a reload.
A bot's face and status line live in bot_user_faces, one row per bot and
user, set by jobs for the user whose conversation they're in. Needs-you
and working only show to that user; every bot read overlays the viewer's
face (botFaceSql) and pushes go to the user a face change is for.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Deleting a bot only changed the page after a reload. The delete route now
reads who could see the bot before removing it, since its assignments and
workspace places go with it, then pushes server:bot-removed
(BotRemovedData) to them. useLiveBots drops the row, so every tab and
every workspace the bot was in updates together.
The rail was built from page props and only knew bots it already had, so a
bot the coordinator created did not appear until a reload.
create_bot and add_bot_to_workspace now push server:bot-created
(BotCreatedData) to everyone who sees the bot: its rail row, the
workspaces it is in, and whether it belongs in that user's own chat list.
useLiveBots inserts it, after the pinned coordinator outside a workspace
and only when the open workspace is one it joined. liveRelay pushes created
bots too, so a bot added by another process still shows.
Bots, workspaces and computers now belong to the install. Users reach
workers only through bot_users (assigned by a super admin or admin) and
workspace_users; the coordinator and main computer are one per install,
and only the super admin sees the coordinator. Conversations are per
user (bot_sessions.user_id), jobs run for the session's user, and live
updates go only to who should see them.
A bot's core memory is now general knowledge shared by every user it
works for; what it learns about a user lives in bot_user_memories.
One-off scripts move existing installs over: move-bot-memory-to-owners
(run first), then share-crew.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
- Rail: Add icon with stacked form modals (bot, workspace, computer),
coordinator, recent bots with "See all bots", Workspaces section, and
a workspace mode listing only its crew
- Workspaces: page for context, bots (per-workspace instructions and
memory), computers (default, link, unlink), models and delete;
conversations scoped per bot per workspace
- Sessions: conversations split into searchable, reopenable sessions
with carry-over summaries; sessions and new-session header icons
- Computers: created and provisioned over the docker socket, home dir
kept at DATA_DIR/computers/<id> after delete; super admins only
- Models: workspace settings win over a bot's own, then the defaults
- Pictures for bots and workspaces; bots editable from the header
- Coordinator tools: update_workspace, set_bot_workspace_instructions
Co-Authored-By: Claude Opus 5.5 <[email protected]>
- Serve each computer to its owner under /computers/<id>/ (HTTP and
websockets proxied to the AIO sandbox) and show it in a shared
ComputerView with browser, terminal and code modes, opened from the
computers modal or a take-over button in the thread.
- ask_user_to_take_over lets a bot hand a computer to the user for a
login or captcha; the message gets a take-over button and needs-you,
and the viewer hands it back with a chat message.
- While a task runs the thread shows "<bot> is working..." with the
task's latest step; steps read as plain words, not tool names.
- A coordinator with no workers no longer claims to delegate, and bots
describe started tasks in the first person.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
- Bots have a current computer (bots.computer_id), switched from the chat
header beside the model button or by clicking a computer the bot names
([computer:ID]); tasks run on it by default
- search_computers runs one read-only command on all of a bot's computers;
chat replies know their computers (compose-computers-note)
- The coordinator can use every computer and grants or revokes crew access
with assign_computer
- Delegated jobs record requested_by_bot_id; the worker's result or failure
is reported back in the coordinator's conversation
- bots_queue_updates records each model turn of a job, pushed live and shown
in the working-on panel
- Live relay: every process pushes new messages, job steps and bot changes
from the database to its own sockets, so work done elsewhere shows live
- Bot messages render as markdown (react-markdown + remark-gfm, no raw HTML)
with twui components; rail previews are plain text
- Fix header button overflow in the model and computer selects
- Bump @moduletrace/twui to 0.1.2
Co-Authored-By: Claude Opus 5.5 <[email protected]>
- Harness: providers, model catalog synced hourly, automatic free picks,
cooldowns, tool loop with context trimming, MCP computer tools
- Bots queue runs replies and tasks through the harness; jobs wait
(attempt refunded) while every model is resting
- Envs: Settings > Environment, scoped general/workspace/computer,
written to computers and redacted from tool output
- Models: Settings > Models, per-harness provider select defaulting to
OpenRouter, and a models modal on each bot's chat (the coordinator's
edits the defaults)
- Chat: thinking row while a reply runs, file attachments (upload,
paste, drop) that replies read as images, audio and document text
- Fix media uploads failing after a user's first file, and private
files being served from the public media path
Co-Authored-By: Claude Opus 5.5 <[email protected]>
- Schema: workspaces (+ workspace_bots, workspace_computers), containers,
computers, bot_computers, bots.memory, workspaces.system_prompt, and
bots_queue as a bun-sqlite isQueue table
- Main computer (and its container record) ensured with the coordinator on
signup, seed and /admin/bots; the coordinator's access to it is locked
- Effective computer access: direct grants plus workspace computers
- Full bot prompt: core instructions, workspace prompts, then memory;
bots can rewrite their memory and system prompt
- Queue helpers on the native BunSQLite.queue API, plus a handler that
heartbeats, serializes per computer, recovers after restarts, reports
lost jobs, supports cancelling and pushes outcomes over the WebSocket
- Enable SQLite foreign keys at startup so schema cascades fire
- Prune finished jobs daily
Co-Authored-By: Claude Opus 5.5 <[email protected]>
tben
