Problem
AgentV creates or reuses eval workspaces during runs, but the workspace location is only surfaced through verbose setup logs today.
Examples in the current code path:
- [setup] shared workspace created at:
- [setup] materialised static workspace at:
- [setup] reusing existing static workspace:
Those messages are emitted from the orchestrator only when --verbose is enabled. In normal agentv eval output, users may not know where the workspace lives, even though they often want to inspect generated files, hook side effects, or repo state after a run.
Why this matters
- Users cannot easily inspect artifacts created inside the eval workspace unless they already know AgentV's internal workspace layout.
- This is especially confusing when
workspace.path comes from env interpolation and may be blank, causing AgentV to fall back to a managed workspace.
keepWorkspaces / retention options are much more useful if the CLI tells users where the retained workspace is.
Proposal
Add a standard non-verbose log line that reports the workspace location whenever AgentV creates, materializes, or reuses a workspace for an eval run.
Suggested behavior:
- For shared/temp/static workspaces, print a concise line once per run such as:
Workspace: C:\Users\<user>\.agentv\workspaces\<run-id>\shared
Workspace: D:\path\to\static\workspace (reused)
- For pooled mode, print the acquired slot path or paths in a concise way.
- If no workspace is used, print nothing.
- Keep the detailed
[setup] ... messages behind --verbose.
Acceptance signals
- Running
agentv eval ... without --verbose shows the effective workspace path when a workspace exists.
- The message is concise and appears early enough for users to inspect files during or after the run.
- Verbose mode still includes the existing detailed setup messages.
Notes
This is a UX improvement rather than a behavioral change. The core ask is to make the effective workspace path discoverable in normal CLI output.
Problem
AgentV creates or reuses eval workspaces during runs, but the workspace location is only surfaced through verbose setup logs today.
Examples in the current code path:
Those messages are emitted from the orchestrator only when
--verboseis enabled. In normalagentv evaloutput, users may not know where the workspace lives, even though they often want to inspect generated files, hook side effects, or repo state after a run.Why this matters
workspace.pathcomes from env interpolation and may be blank, causing AgentV to fall back to a managed workspace.keepWorkspaces/ retention options are much more useful if the CLI tells users where the retained workspace is.Proposal
Add a standard non-verbose log line that reports the workspace location whenever AgentV creates, materializes, or reuses a workspace for an eval run.
Suggested behavior:
Workspace: C:\Users\<user>\.agentv\workspaces\<run-id>\sharedWorkspace: D:\path\to\static\workspace (reused)[setup] ...messages behind--verbose.Acceptance signals
agentv eval ...without--verboseshows the effective workspace path when a workspace exists.Notes
This is a UX improvement rather than a behavioral change. The core ask is to make the effective workspace path discoverable in normal CLI output.