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.
146 lines
4.7 KiB
TypeScript
146 lines
4.7 KiB
TypeScript
type WorkspacePrompt = {
|
|
name?: string | null;
|
|
description?: string | null;
|
|
system_prompt?: string | null;
|
|
};
|
|
|
|
type WorkspaceBotPrompt = {
|
|
system_prompt?: string | null;
|
|
memory?: string | null;
|
|
};
|
|
|
|
type BotPrompt = {
|
|
/**
|
|
* The bot's id: how it names itself, and how the user, the crew tools
|
|
* and every ref in a message point at it.
|
|
*/
|
|
id?: number | string | null;
|
|
name?: string | null;
|
|
kind?: string | null;
|
|
/**
|
|
* The one-line role it does that job as, when it has one.
|
|
*/
|
|
role?: string | null;
|
|
};
|
|
|
|
type Params = {
|
|
/**
|
|
* The bot itself. Its id is always in the prompt: it is how the bot
|
|
* names itself and how everything else (crew tools, refs, tasks,
|
|
* files) points back at it, in chat replies, tasks and routines alike.
|
|
*/
|
|
bot?: BotPrompt | null;
|
|
system_prompt?: string | null;
|
|
/**
|
|
* The install-wide text every bot gets, right after its own core
|
|
* instructions (Settings > General).
|
|
*/
|
|
appendix?: string | null;
|
|
memory?: string | null;
|
|
/**
|
|
* What the bot remembers about the user it's working with.
|
|
*/
|
|
user_memory?: string | null;
|
|
/**
|
|
* The workspace the bot is working in, if any.
|
|
*/
|
|
workspace?: WorkspacePrompt | null;
|
|
/**
|
|
* The bot's own instructions and memory for that workspace.
|
|
*/
|
|
workspace_bot?: WorkspaceBotPrompt | null;
|
|
};
|
|
|
|
/**
|
|
* The full prompt a bot runs with (markdown): who it is (its name and id,
|
|
* so it always knows which bot it is), its own core instructions, then the
|
|
* install-wide appendix, then, in a workspace, the workspace's
|
|
* context and the bot's instructions there, then its core memory, its
|
|
* memory about the user and its memory in the workspace (said to be the
|
|
* bot's own notes, not the workspace's context). Empty parts are left out;
|
|
* the workspace heading always shows, so the bot knows where it is.
|
|
*/
|
|
export default function composeBotPrompt({
|
|
bot,
|
|
system_prompt,
|
|
appendix,
|
|
memory,
|
|
user_memory,
|
|
workspace,
|
|
workspace_bot,
|
|
}: Params) {
|
|
const parts: string[] = [];
|
|
const name = workspace?.name?.trim() || "Untitled";
|
|
const identity = composeIdentity({ bot });
|
|
|
|
if (identity) {
|
|
parts.push(identity);
|
|
}
|
|
|
|
if (system_prompt?.trim()) {
|
|
parts.push(system_prompt.trim());
|
|
}
|
|
|
|
if (appendix?.trim()) {
|
|
parts.push(`# Install-wide instructions\n\n${appendix.trim()}`);
|
|
}
|
|
|
|
if (workspace) {
|
|
const context = [workspace.description?.trim(), workspace.system_prompt?.trim()].filter(
|
|
Boolean,
|
|
) as string[];
|
|
|
|
parts.push(
|
|
[
|
|
`# Workspace: ${name}`,
|
|
`You're working in the workspace "${name}". Follow its context and your instructions for it below, over your general habits where they differ.`,
|
|
context.length ? "The workspace's context, which the user sets:" : "",
|
|
...context,
|
|
]
|
|
.filter(Boolean)
|
|
.join("\n\n"),
|
|
);
|
|
|
|
if (workspace_bot?.system_prompt?.trim()) {
|
|
parts.push(`# Your instructions in ${name}\n\n${workspace_bot.system_prompt.trim()}`);
|
|
}
|
|
}
|
|
|
|
if (memory?.trim()) {
|
|
parts.push(`# Memory\n\n${memory.trim()}`);
|
|
}
|
|
|
|
if (user_memory?.trim()) {
|
|
parts.push(`# Memory about the user\n\n${user_memory.trim()}`);
|
|
}
|
|
|
|
if (workspace && workspace_bot?.memory?.trim()) {
|
|
parts.push(
|
|
`# Memory in ${name}\n\nYour own notes for this workspace: what you have learned here. They are not the workspace's description or context, which the user sets above; never call them the workspace's record. The user reads them in the workspace's settings, under this bot's row.\n\n${workspace_bot.memory.trim()}`,
|
|
);
|
|
}
|
|
|
|
return parts.join("\n\n");
|
|
}
|
|
|
|
/**
|
|
* Who the bot is, said once at the top of its prompt: its name, kind and
|
|
* id, and that the id is how it is named. Empty when the caller didn't say
|
|
* which bot it is.
|
|
*/
|
|
function composeIdentity({ bot }: { bot?: BotPrompt | null }) {
|
|
const id = bot?.id === null || bot?.id === undefined ? "" : String(bot.id).trim();
|
|
|
|
if (!id) return "";
|
|
|
|
const name = bot?.name?.trim() || "A bot";
|
|
const kind = bot?.kind == "coordinator" ? "the coordinator" : "a worker bot";
|
|
const role = bot?.role?.trim();
|
|
|
|
return [
|
|
`# You are ${name}`,
|
|
`You are ${kind} on this crew.${role ? ` Your role: ${role}.` : ""}`,
|
|
`Your bot id is ${id}. That id is you: when a tool takes a bot id for the bot it is about (your own tasks, your own picture), this is it, and when you name yourself in a message write the ref \`[bot:${id}]\`, never another bot's.`,
|
|
].join("\n\n");
|
|
}
|