openai/codex · warning · io::Error

generated image destination already exists

Error message

generated image destination already exists

What it means

Immediately before writing, save_image_generation_result stats the destination (derived from session_id and call_id): any successful stat returns AlreadyExists. Only NotFound proceeds — an existing destination (e.g. a hardlink) could otherwise overwrite files outside the workspace, so overwriting is never attempted.

Source

Thrown at codex-rs/ext/image-generation/src/tool.rs:361

                        .get_metadata(&parent_uri, Default::default(), sandbox)
                        .await?;
                    if metadata.is_symlink || !metadata.is_directory {
                        return Err(io::Error::new(
                            io::ErrorKind::PermissionDenied,
                            "generated image directory is not a real directory",
                        ));
                    }
                }

                // Existing destination hardlinks could otherwise overwrite files outside the workspace.
                let path_uri = PathUri::from_abs_path(&path);
                match environment
                    .file_system
                    .get_metadata(&path_uri, Default::default(), sandbox)
                    .await
                {
                    Ok(_) => {
                        return Err(io::Error::new(
                            io::ErrorKind::AlreadyExists,
                            "generated image destination already exists",
                        ));
                    }
                    Err(error) if error.kind() == io::ErrorKind::NotFound => {}
                    Err(error) => return Err(error),
                }

                environment
                    .file_system
                    .write_file(&path_uri, bytes, Default::default(), sandbox)
                    .await?;
                Ok(path)
            }
            .await;
            (output_dir, save_result)
        }
    };

View on GitHub (pinned to 339751715c)

Solutions

1. Generate a fresh call_id when retrying the image tool call.
2. Delete the stale artifact named in the warning if the run is genuinely new.
3. Treat as informational: the model still receives the generated image; only persistence is skipped.

// Before replaying a tool call, ensure the artifact name is free
if artifact_path.exists() {
    std::fs::remove_file(&artifact_path)?; // or mint a fresh call_id
}

When it happens

Trigger: The same image tool call is retried or replayed with the same session_id and call_id, or a file already exists at generated_images/<artifact-name>.

Common situations: Client retry of a timed-out tool call reusing the call id; two executions sharing a session id; leftover artifacts from a previous run.

Related errors

• generated image exceeds the executor file size limit
• generated image directory is not a real directory
• pid-managed app-server shutdown is unsupported on this platf
• pid-managed updater shutdown is unsupported on this platform
• failed to read start time for pid-managed app server {pid}

AI-assisted analysis of openai/codex@339751715c (2026-08-25). Data as JSON: /api/errors/37b9505fe15d3b13. Report an issue: GitHub.