oracle

To bundle repository files into an AI prompt, you attach selected files, directories, and globs to form a single contextual payload. This allows the AI model to consider code and documentation together in one pass for code context or task-specific guidance.

Yes, you can save and resume long-running AI prompts using session-aware runs. This session management feature lets you reattach to existing contextual requests, enabling iterative task execution without losing your previous repository file selections.

Yes, engine selection supports both browser and API paths for generating code context. You can choose your execution path and preview the bundled payload with dry-runs before execution to ensure reproducible and safe AI runs.

The best way to select specific code files is by using file attachment controls to attach files, directories, and globs. Choosing a tight file set ensures the AI receives only the necessary repository content for generating accurate explanations.

You preview a prompt payload before sending it to an AI model by using the dry-runs feature. Dry-runs allow you to check the bundled files and prompts to ensure safe executions and reproducible results.

You should avoid bundling entire directories when the repository content is too large, as it may exceed the AI model's context limits. Instead, select a tight file set using specific file attachments and globs to ensure safe executions.