Enhanced ChatGPT Clone: Features Agents, MCP, Skills, DeepSeek, Anthropic, AWS, OpenAI, Responses API, Azure, Groq, o1, GPT-5, Mistral, OpenRouter, Vertex AI, Gemini, Artifacts, AI model switching, message search, Code Interpreter, langchain, DALL-E-3, OpenAPI Actions, Functions, Secure Multi-User Auth, Presets, open-source for self-hosting. Active
TypeScript aianthropicartifactsawsazurechatgptchatgpt-cloneclaudeclonedeepseekgeminigooglegpt-5librechatmcpo1openairesponses-apivisionwebui
import { z } from 'zod';
import * as agentsSdk from '@librechat/agents';
import { tool } from '@librechat/agents/langchain/tools';
import type { GenericTool, LCTool, SubagentExecutionContext } from '@librechat/agents';
import type { RunFileSession } from './session';
export const LIST_RUN_FILES_TOOL = 'list_run_files';
export const PUBLISH_ARTIFACT_TOOL = 'publish_artifact';
/** Older SDKs cannot distinguish simultaneous copies of the same agent. */
export function isRunFileSharingSupported(): boolean {
return (agentsSdk as { SUBAGENT_CONTEXT_VERSION?: number }).SUBAGENT_CONTEXT_VERSION === 1;
}
/** Every configured tool must execute through host authorization and artifact capture. */
export function eventOnlyRunFileTools(
tools: GenericTool[] | undefined,
definitions: readonly LCTool[],
): GenericTool[] | undefined {
if (tools == null || tools.length === 0) return tools;
const eventTools = new Set(definitions.map((definition) => definition.name));
for (const candidate of tools) {
if (
candidate == null ||
typeof candidate.name !== 'string' ||
typeof candidate.invoke !== 'function' ||
!eventTools.has(candidate.name)
) {
throw new Error('Run file sharing requires host-dispatched tools with explicit definitions.');
}
}
// The definitions bind the model; removing direct instances sends execution
// through ON_TOOL_EXECUTE, including private workspace setup and capture.
return [];
}
/** In-graph tools receive execution metadata stamped by the SDK's owning ToolNode. */
export function createRunFileTools(
session: RunFileSession,
agentId: string,
runSignal: AbortSignal,
): GenericTool[] {
const list = tool(
async (_input, config) => {
const context = config.metadata?.executionContext as SubagentExecutionContext | undefined;
return JSON.stringify(await session.list(agentId, context, config.signal ?? runSignal));
},
{
name: LIST_RUN_FILES_TOOL,
description:
'List the files this execution may read and its unpublished output artifact IDs. ' +
'Use this to discover files published by other authorized agents during this run. ' +
'Private output versions remain available across sandbox calls during this run. ' +
'The same filename may have multiple artifact IDs, listed in creation order; publish the version you want to return when ready. ' +
"When delegating a task that returns files, ask the producing subagent to publish its outputs before returning. A parent cannot publish a child's private artifact IDs.",
schema: z.object({}).strict(),
},
);
const publish = tool(
async ({ artifact_id, recipient_agent_ids }, config) => {
const context = config.metadata?.executionContext as SubagentExecutionContext | undefined;
const file = await session.publish(
agentId,
context,
artifact_id,
recipient_agent_ids,
config.signal ?? runSignal,
);
return JSON.stringify({ file_id: file.file_id, filename: file.filename });
},
{
name: PUBLISH_ARTIFACT_TOOL,
description:
"Publish one of this execution's own output artifacts from list_run_files as a durable downloadable file for " +
'the parent and user when ready. Each artifact ID selects an immutable output version, even after the working file changes. ' +
'Outputs stay private until this succeeds. Optional sibling agent ' +
'recipients must be authorized by the run sharing policy.',
schema: z
.object({
artifact_id: z.string().min(1).describe('The artifact_id returned by list_run_files.'),
recipient_agent_ids: z.array(z.string().min(1)).optional(),
})
.strict(),
},
);
return [list, publish];
}