Hi Anna team,
I would like to request a configurable timeout option for anna.tools.invoke in Anna App frontends.
## Context
I am building an Anna App where the frontend calls an Executa tool with:
await anna.tools.invoke({
tool_id,
method,
args,
});
Inside the backend tool, the tool may call Anna LLM sampling or another LLM-backed workflow. That backend-side LLM call can have its own timeout or retry policy, but the frontend-to-tool invocation currently appears to have a fixed host/runtime timeout. When the backend work takes longer than that, the `tools.invoke` chain fails even if the backend operation is still valid and intentionally long-running.
The workaround is to redesign the tool as an asynchronous job API: start a job, return immediately, then have the frontend poll `get_status` / `get_result`. That pattern works, but it adds a fair amount of frontend and backend boilerplate for cases where the developer simply wants one longer-running tool call.
## Request
Could anna.tools.invoke accept an optional timeout parameter, for example:
await anna.tools.invoke({
tool_id,
method,
args,
timeoutMs: 180_000,
});
or an equivalent option shape:
await anna.tools.invoke(
{ tool_id, method, args },
{ timeoutMs: 180_000 }
);
The important part is that the app frontend can request a longer timeout for known long-running tool calls, while the platform can still enforce a maximum allowed value.
## What I observed locally
In my local environment:
- @anna-ai/cli: 0.1.14
- pinned local runtime: anna-app-runtime-local@0.2.0a2
- current docs/examples show tools.invoke args as { tool_id, method?, args }, with no timeout option
From the local runtime source I found that the local Executa JSON-RPC wait uses a fixed default:
async def _rpc(…, timeout: float = 65.0):
return await asyncio.wait_for(fut, timeout=timeout)
So the effective behavior seems to be a fixed ~65 second limit in local development. I am not sure whether production uses exactly the same value, but the public App UI Host API currently does not appear to expose any frontend-controlled timeout for `tools.invoke`.
## Why this would help
A configurable timeout would make it much easier to build Anna Apps that delegate substantial work to an Executa tool, especially workflows involving LLM planning, summarization, retrieval, report generation, or multi-step processing. Developers could still choose the async job/polling model when it is the better UX, but they would not be forced into that architecture solely because the frontend invoke timeout is fixed.
Thanks for considering this.