Feat/cloud turn - #282
Merged
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
anton.cloud_turn is the entrypoint that runs a single anton turn headlessly inside an isolated cloud sandbox (one turn per request). This is how anton runs in the multi-tenant cloud product, as opposed to the local desktop CLI.
What this PR does: it makes that entrypoint safe to run in a shared sandbox where the agent can execute arbitrary code. Concretely, when it builds the agent session, it leaves out the desktop-only behavior that would leak data across tenants — it doesn't load the workspace's .env into the process, doesn't touch personal memory on disk, and doesn't inject datasource credentials into the environment; conversation history comes from the request instead of local disk. It also strips secrets (like the model API key) out of the environment the agent's generated code runs in, isolates the output stream at the OS level so stray prints or library output can't corrupt the JSON protocol, returns short scrubbed error messages instead of raw tracebacks, and restricts the pod to a small reviewed set of tools.
Not included yet: the model API key still has to be provided to the pod via environment variables, and model selection is passed through as-is; both are tightened in a later Gateway change(model calls go through a gateway with a short-lived, run-scoped token; )