Moduletrace/cobble
2
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases 9 Wiki Activity

Add workspaces, conversation sessions, app-managed computers and pictures

Browse Source 
- 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]>
This commit is contained in:
tben
2026-10-19 08:39:33 +01:00
co-authored by Claude Opus 5.5
parent 56daf0c462
commit 0ce61cdb49
223 changed files with 8166 additions and 585 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/bots-queue/compose-computers-note.ts
+17 -6
View File
@@ -10,6 +10,11 @@ type Params = {
* The coordinator can use every computer and give the crew access.
*/
is_coordinator?: boolean;
/**
* In a workspace the computers are the workspace's, and its default
* is where tasks run.
*/
in_workspace?: boolean;
};
/**
@@ -17,9 +22,11 @@ type Params = {
* one it's on, and how it reaches them (markdown). Without it a bot asked
* "can you see the computer?" says it can't.
*/
export default function composeComputersNote({ computers, is_coordinator }: Params) {
export default function composeComputersNote({ computers, is_coordinator, in_workspace }: Params) {
if (!computers.length) {
return `# Your computers\n\nYou have no computer, so you can't run commands, open files or browse. If the user asks for that, say so.`;
return in_workspace
? `# Your computers\n\nThis workspace has no computer, so you can't run commands, open files or browse here. If the user asks for that, say so: they can add a computer to the workspace and set it as its default.`
: `# Your computers\n\nYou have no computer, so you can't run commands, open files or browse. If the user asks for that, say so.`;
}
const listed = computers
@@ -38,10 +45,14 @@ export default function composeComputersNote({ computers, is_coordinator }: Para
${listed}
Each is a Linux sandbox of your own (a shell, files, code runners and a browser), set up for the crew. It isn't the user's desktop: you can't see their screen or their own files unless they're put on it. The user picks which one you're on; your tasks run there.${
is_coordinator
? ` As the coordinator you can use every computer the user has, and you decide which ones each crew member can use with assign_computer.`
: ""
Each is a Linux sandbox of your own (a shell, files, code runners and a browser), set up for the crew. It isn't the user's desktop: you can't see their screen or their own files unless they're put on it. ${
in_workspace
? `These are this workspace's computers. Its tasks run on the default one (the first, "you're on it"); you can switch the default with set_workspace_computer when the work needs another.`
: `The user picks which one you're on; your tasks run there.${
is_coordinator
? ` As the coordinator you can use every computer the user has, and you decide which ones each crew member can use with assign_computer.`
: ""
}`
}
- To find which computer holds something (a file, a running app, a logged-in session), use search_computers: it runs one read-only command on all of them at once.
src/bots-queue/grab-session-carry-over.ts
+71
View File
@@ -0,0 +1,71 @@
import BunSQLite from "@moduletrace/bun-sqlite";
import type { BUN_SQLITE_COBBLE_BOT_SESSIONS } from "@/db/types/db";
import { AppData } from "@/src/data/app-data";
import { PreviousSessionNote, SessionSummaryPrompt } from "@/src/data/harness-prompts";
import grabBotMessages from "@/src/functions/backend/db/bots/grab-bot-messages";
import generateText from "@/src/harness/text/generate-text";
type Params = {
session_id: number | string;
bot_id: number | string;
workspace_id?: number | string | null;
signal?: AbortSignal;
};
/**
* The previous session's summary for a session's prompt (markdown), or ""
* when it doesn't carry on from one. The summary is written the first
* time it's needed and kept on that session.
*/
export default async function grabSessionCarryOver({
session_id,
bot_id,
workspace_id,
signal,
}: Params) {
const previous = (
await BunSQLite.sql<BUN_SQLITE_COBBLE_BOT_SESSIONS>({
sql: `SELECT p.* FROM bot_sessions s JOIN bot_sessions p ON p.id = s.previous_session_id
WHERE s.id = ? AND p.bot_id = s.bot_id LIMIT 1`,
values: [session_id],
})
).payload?.[0];
if (!previous?.id) return "";
if (previous.summary?.trim()) {
return `${PreviousSessionNote}\n\n${previous.summary.trim()}`;
}
const messages = await grabBotMessages({
session_id: previous.id,
limit: AppData.BotSessionSummaryMessages,
});
if (!messages.length) return "";
const transcript = messages
.map((message) => `${message.author == "user" ? "User" : "You"}: ${message.text || "(files only)"}`)
.join("\n\n");
const result = await generateText({
messages: [
{ role: "system", content: SessionSummaryPrompt },
{ role: "user", content: transcript },
],
bot_id,
workspace_id,
signal,
});
const summary = (result.text || "").trim().slice(0, AppData.BotSessionSummaryMaxChars);
if (!summary) return "";
await BunSQLite.sql({
sql: `UPDATE bot_sessions SET summary = ?, updated_at = ? WHERE id = ?`,
values: [summary, Date.now(), previous.id],
});
return `${PreviousSessionNote}\n\n${summary}`;
}
src/bots-queue/process-bot-job.ts
+5 -1
View File
@@ -84,7 +84,11 @@ export default async function processBotJob({
bot_id,
data: {
event: "server:bot-job-update",
data: { ...row, task_id: job.task_id || null } satisfies BotJobUpdateData,
data: {
...row,
task_id: job.task_id || null,
session_id: job.session_id || null,
} satisfies BotJobUpdateData,
},
});
} catch (error) {
src/bots-queue/run-bot-reply.ts
+43 -10
View File
@@ -1,6 +1,11 @@
import type { BUN_SQLITE_COBBLE_BOTS } from "@/db/types/db";
import { AppData } from "@/src/data/app-data";
import { ChatReplyNote, CoordinatorReplyNote, CoordinatorSoloNote } from "@/src/data/harness-prompts";
import {
ChatReplyNote,
CoordinatorReplyNote,
CoordinatorSoloNote,
WorkspaceLearningNote,
} from "@/src/data/harness-prompts";
import grabBotMessages from "@/src/functions/backend/db/bots/grab-bot-messages";
import grabBotPrompt from "@/src/functions/backend/db/bots/grab-bot-prompt";
import grabBotComputers from "@/src/functions/backend/db/computers/grab-bot-computers";
@@ -8,6 +13,8 @@ import runTextLoop from "@/src/harness/text/run-text-loop";
import grabBotChatTools from "@/src/harness/tools/bot/grab-bot-chat-tools";
import botMessagesToHarness from "./bot-messages-to-harness";
import composeComputersNote from "./compose-computers-note";
import grabSessionCarryOver from "./grab-session-carry-over";
import jobSessionId from "@/src/functions/backend/db/bots-queue/job-session-id";
import loadAttachmentParts from "./load-attachment-parts";
import withTakeOverRefs from "./with-take-over-refs";
import type { BotJobContext, BotJobOutcome } from "./types";
@@ -17,11 +24,14 @@ type Params = BotJobContext & {
};
/**
* Answers the user in chat: the bot reads the recent conversation (and
* the files of the latest messages, `loadAttachmentParts`) and
* replies, knowing its computers (`composeComputersNote`), with its chat tools (memory, instructions, start_task, and for
* the coordinator delegate_task). Finishes silently when the last message
* is already the bot's, i.e. an earlier reply answered everything.
* Answers the user in chat: the bot reads the session's recent messages
* (and the files of the latest ones, `loadAttachmentParts`) and replies,
* knowing where it is (its workspace, if the session is in one), what
* happened in the session before (`grabSessionCarryOver`) and its
* computers (`composeComputersNote`), with its chat tools (memory,
* instructions, start_task, and for the coordinator delegate_task).
* Finishes silently when the last message is already the bot's, i.e. an
* earlier reply answered everything.
*/
export default async function runBotReply({
bot,
@@ -29,8 +39,11 @@ export default async function runBotReply({
signal,
report,
}: Params): Promise<BotJobOutcome> {
const session_id = await jobSessionId({ job });
const workspace_id = job.workspace_id || null;
const history = await grabBotMessages({
bot_id: bot.id,
session_id,
limit: AppData.BotReplyHistoryMessages,
});
@@ -38,8 +51,19 @@ export default async function runBotReply({
return { result: "", silent: true };
}
const prompt = await grabBotPrompt({ bot_id: bot.id, user_id: bot.user_id });
const computers = await grabBotComputers({ bot_id: bot.id, user_id: bot.user_id });
const prompt = await grabBotPrompt({ bot_id: bot.id, user_id: bot.user_id, workspace_id });
const computers = await grabBotComputers({ bot_id: bot.id, user_id: bot.user_id, workspace_id });
/**
* A summary that can't be written now (no model free) is left out
* rather than holding up the reply.
*/
const carry_over = await grabSessionCarryOver({
session_id,
bot_id: bot.id,
workspace_id,
signal,
}).catch(() => "");
const take_overs: string[] = [];
@@ -47,6 +71,8 @@ export default async function runBotReply({
bot,
user_id: bot.user_id,
computers,
session_id,
workspace_id,
message_id: job.message_id || undefined,
onTakeOver: ({ ref }) => take_overs.push(ref),
});
@@ -64,8 +90,14 @@ export default async function runBotReply({
const result = await runTextLoop({
system: [
prompt,
carry_over,
ChatReplyNote,
composeComputersNote({ computers, is_coordinator: bot.kind == "coordinator" }),
workspace_id ? WorkspaceLearningNote : "",
composeComputersNote({
computers,
is_coordinator: bot.kind == "coordinator",
in_workspace: Boolean(workspace_id),
}),
crew_note,
`Now: ${new Date().toUTCString()}.`,
]
@@ -78,6 +110,7 @@ export default async function runBotReply({
tools,
max_steps: AppData.BotReplyMaxToolSteps,
bot_id: bot.id,
workspace_id,
signal,
onStep: report,
});
src/bots-queue/run-bot-task.ts
+30 -8
View File
@@ -1,4 +1,4 @@
import { TaskRunNote } from "@/src/data/harness-prompts";
import { TaskRunNote, WorkspaceLearningNote } from "@/src/data/harness-prompts";
import grabBotComputers from "@/src/functions/backend/db/computers/grab-bot-computers";
import grabBotPrompt from "@/src/functions/backend/db/bots/grab-bot-prompt";
import runTextLoop from "@/src/harness/text/run-text-loop";
@@ -8,6 +8,8 @@ import askToTakeOverTool from "@/src/harness/tools/computer/ask-to-take-over-too
import searchComputersTool from "@/src/harness/tools/computer/search-computers-tool";
import computerRef from "@/src/utils/computer-ref";
import composeJobBrief from "./compose-job-brief";
import grabSessionCarryOver from "./grab-session-carry-over";
import jobSessionId from "@/src/functions/backend/db/bots-queue/job-session-id";
import withTakeOverRefs from "./with-take-over-refs";
import type { BotJobContext, BotJobOutcome } from "./types";
@@ -16,7 +18,8 @@ type Params = BotJobContext & {
};
/**
* Works through a task: the bot follows its brief with its full prompt,
* Works through a task in its session's workspace (if any): the bot
* follows its brief with its full prompt,
* its own tools, search_computers across all its computers and, when the
* job has a computer, that computer's tools (over MCP), until it answers. That answer is the report for the user.
* When only the user can go on (a login, a captcha), the bot asks them to
@@ -30,13 +33,27 @@ export default async function runBotTask({
signal,
report,
}: Params): Promise<BotJobOutcome> {
const workspace_id = job.workspace_id || null;
const computer = job.computer_id
? await openComputerTools({ computer_id: job.computer_id, bot_id: bot.id, signal })
? await openComputerTools({
computer_id: job.computer_id,
bot_id: bot.id,
workspace_id,
signal,
})
: null;
try {
const prompt = await grabBotPrompt({ bot_id: bot.id, user_id: bot.user_id });
const computers = await grabBotComputers({ bot_id: bot.id, user_id: bot.user_id });
const session_id = await jobSessionId({ job });
const prompt = await grabBotPrompt({ bot_id: bot.id, user_id: bot.user_id, workspace_id });
const computers = await grabBotComputers({ bot_id: bot.id, user_id: bot.user_id, workspace_id });
const carry_over = await grabSessionCarryOver({
session_id,
bot_id: bot.id,
workspace_id,
signal,
}).catch(() => "");
const take_overs: string[] = [];
/**
@@ -50,10 +67,14 @@ export default async function runBotTask({
const result = await runTextLoop({
system: [
prompt,
carry_over,
TaskRunNote,
workspace_id ? WorkspaceLearningNote : "",
computer
? `You're working on the computer "${computer.name}" (a Linux sandbox with a shell, files, code runners and a browser). Use its tools; don't ask the user to run anything. When only the user can go on (a login, a captcha, a code sent to them), ask them to take over with ask_user_to_take_over.`
: `You have no computer for this task.`,
: workspace_id
? `You have no computer for this task: this workspace has none. If the task needs one, stop and say so in the report; the user can add a computer to the workspace.`
: `You have no computer for this task.`,
computers.length > 1
? `search_computers runs a read-only command on every computer you have. When you name one in the report, write its ref, e.g. ${computerRef({ id: computers[0]!.id || 0 })}: the user can click it to switch you to that computer.`
: "",
@@ -76,10 +97,10 @@ export default async function runBotTask({
},
],
tools: [
...grabBotTools({ user_id: bot.user_id, bot_id: bot.id }),
...grabBotTools({ user_id: bot.user_id, bot_id: bot.id, workspace_id, computers }),
...(computers.length
? [
searchComputersTool({ bot_id: bot.id, computers }),
searchComputersTool({ computers, workspace_id }),
askToTakeOverTool({
computers: take_over_computers,
in_task: true,
@@ -90,6 +111,7 @@ export default async function runBotTask({
...(computer?.tools || []),
],
bot_id: bot.id,
workspace_id,
signal,
onStep: (step) =>
report({ ...step, text: step.text && computer ? computer.redact(step.text) : step.text }),