Build and run agents you can see, understand and trust.
Python agentchatbotlarge-language-modelsllmllm-agentmcpmulti-agentmulti-modalreact-agentrealtime-agentvoice-agentvoice-assistant
import { ApiError, client, getBaseUrl, getUserId } from './client';
import type { UploadProgress } from './knowledgeBase';
import type {
AddFromLibraryResponse,
DirectoryListing,
WorkspaceStatus,
MCPClient,
MCPClientStatus,
Skill,
} from './types';
export interface UploadOptions {
/** Fired with byte-level progress while the body is streamed. */
onProgress?: (progress: UploadProgress) => void;
/** Aborts the upload; rejects with a `DOMException` named `AbortError`. */
signal?: AbortSignal;
}
/**
* XHR-based folder upload — `fetch` surfaces no byte-level send
* progress, so anything driving a progress bar has to use XHR.
*
* The manifest rides alongside the parts because the server builds a
* tar as they arrive, and a tar header needs each member's size before
* its bytes — which a multipart part does not declare.
*/
function uploadSkillXhr(
agentId: string,
sessionId: string,
files: File[],
options: UploadOptions = {},
): Promise<void> {
const { onProgress, signal } = options;
const formData = new FormData();
formData.append(
'manifest',
JSON.stringify({
entries: files.map((file) => ({
path: file.webkitRelativePath || file.name,
size: file.size,
})),
}),
);
for (const file of files) formData.append('files', file);
return new Promise((resolve, reject) => {
if (signal?.aborted) {
reject(new DOMException('Aborted', 'AbortError'));
return;
}
const xhr = new XMLHttpRequest();
const url = new URL('/workspace/skill/upload', getBaseUrl());
url.searchParams.set('agent_id', agentId);
url.searchParams.set('session_id', sessionId);
xhr.open('POST', url.toString(), true);
xhr.setRequestHeader('X-User-ID', getUserId());
const onAbort = () => xhr.abort();
signal?.addEventListener('abort', onAbort, { once: true });
const cleanup = () => signal?.removeEventListener('abort', onAbort);
if (xhr.upload && onProgress) {
xhr.upload.onprogress = (e) => {
onProgress({
loaded: e.loaded,
total: e.lengthComputable ? e.total : 0,
});
};
}
xhr.onload = () => {
cleanup();
if (xhr.status >= 200 && xhr.status < 300) {
resolve();
return;
}
let detail = xhr.responseText || xhr.statusText;
try {
const json = JSON.parse(xhr.responseText) as { detail?: unknown };
if (typeof json.detail === 'string') detail = json.detail;
else if (json.detail !== undefined) detail = JSON.stringify(json.detail);
} catch {
// keep raw text
}
reject(new ApiError(xhr.status, detail));
};
xhr.onerror = () => {
cleanup();
reject(new ApiError(0, 'Network error'));
};
xhr.onabort = () => {
cleanup();
reject(new DOMException('Aborted', 'AbortError'));
};
xhr.send(formData);
});
}
export const workspaceApi = {
/**
* List one directory level inside the session's workspace.
*
* `path` may be absolute or relative to the workspace root; empty
* lists the root itself. Not confined to the root — for a sandboxed
* backend the reachable filesystem is the sandbox, and for a local
* one the caller is already trusted with the host. The response
* echoes the resolved absolute path.
*/
directories: (agentId: string, sessionId: string, path = '') =>
client.get<DirectoryListing>('/workspace/directories', {
agent_id: agentId,
session_id: sessionId,
path,
}),
/**
* Where the session is pointed, plus the git state of that place.
*
* `silent` because this is fetched on the UI's own schedule: git
* being unavailable is an ordinary answer, not something to raise a
* toast over.
*/
status: (agentId: string, sessionId: string) =>
client.get<WorkspaceStatus>(
'/workspace/status',
{ agent_id: agentId, session_id: sessionId },
{ silent: true },
),
mcp: {
list: (agentId: string, sessionId: string) =>
client.get<MCPClientStatus[]>('/workspace/mcp', {
agent_id: agentId,
session_id: sessionId,
}),
add: (agentId: string, sessionId: string, mcp: MCPClient) =>
client.post<void>('/workspace/mcp', mcp, { agent_id: agentId, session_id: sessionId }),
/**
* Puts MCPs the user has already installed into this workspace.
* Ids, not configs — the rendered config never leaves the server,
* so the client has no way to reconstruct one.
*/
addFromLibrary: (agentId: string, sessionId: string, mcpIds: string[]) =>
client.post<AddFromLibraryResponse>(
'/workspace/mcp/from-library',
{ mcp_ids: mcpIds },
{ agent_id: agentId, session_id: sessionId },
),
remove: (mcpName: string, agentId: string, sessionId: string) =>
client.delete(`/workspace/mcp/${mcpName}`, {
agent_id: agentId,
session_id: sessionId,
}),
},
skill: {
list: (agentId: string, sessionId: string) =>
client.get<Skill[]>('/workspace/skill', { agent_id: agentId, session_id: sessionId }),
/** Uploads a picked folder as a skill, reporting send progress. */
upload: (agentId: string, sessionId: string, files: File[], options: UploadOptions = {}) =>
uploadSkillXhr(agentId, sessionId, files, options),
/** Installs skills the user already has, by library record id. */
addFromLibrary: (agentId: string, sessionId: string, skillIds: string[]) =>
client.post<AddFromLibraryResponse>(
'/workspace/skill/from-library',
{ skill_ids: skillIds },
{ agent_id: agentId, session_id: sessionId },
),
remove: (skillName: string, agentId: string, sessionId: string) =>
client.delete(`/workspace/skill/${skillName}`, {
agent_id: agentId,
session_id: sessionId,
}),
},
};