feat(main): add non-interactive --ask-agent mode to the CLI#403
feat(main): add non-interactive --ask-agent mode to the CLI#403
Conversation
Summary of ChangesHello, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed! This pull request enhances the CLI by introducing a non-interactive mode, enabling users to execute single queries programmatically. This feature is particularly useful for scripting and automation, as it allows the agent to process a query and output the result to standard output without launching the full interactive REPL. The change includes a new CLI option, a dedicated function for single query execution, and refined logging to support this new mode. Highlights
Changelog
Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here. You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension. Footnotes
|
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request introduces a non-interactive --ask-agent mode to the CLI, a valuable addition for scripting and automation. However, the current implementation lacks critical security controls, allowing the RAG agent to execute sensitive tools (like file editing and shell commands) without user confirmation, making it highly vulnerable to prompt injection. It also exposes a potential arbitrary file read vulnerability through unvalidated image path handling. The implementation is otherwise logical and correctly separates logging from the main output, though there are minor style guide suggestions regarding docstrings which are disallowed by the project's general rules. These security issues should be addressed before merging.
| response = await rag_agent.run(question_with_context, message_history=[]) | ||
|
|
||
| # Output response to stdout | ||
| print(response.output) |
There was a problem hiding this comment.
The main_async_single_query function executes the RAG agent with user-supplied input but does not implement any confirmation logic for sensitive operations such as file modifications or shell command execution. Since the agent has access to powerful tools (shell_command_tool, file_writer_tool, etc.), an attacker can use prompt injection via the --ask-agent flag to perform unauthorized actions on the system. In non-interactive mode, the application should either block these operations by default or require an explicit override via the --no-confirm flag. Additionally, ensure that the shell_command tool is included in the EDIT_TOOLS list in config.py to be properly detected by is_edit_operation_response.
| response = await rag_agent.run(question_with_context, message_history=[]) | |
| # Output response to stdout | |
| print(response.output) | |
| response = await rag_agent.run(question_with_context, message_history=[]) | |
| # Check for edit operations in non-interactive mode | |
| if confirm_edits_globally and is_edit_operation_response(response.output): | |
| console.print("\n[bold red]Error: The agent attempted to perform file modifications or sensitive operations in non-interactive mode. Use --no-confirm to allow these changes.[/bold red]") | |
| return | |
| # Output response to stdout | |
| print(response.output) |
| rag_agent = _initialize_services_and_agent(repo_path, ingestor) | ||
|
|
||
| # Handle images in the question | ||
| question_with_context = _handle_chat_images(question, project_root) |
There was a problem hiding this comment.
The main_async_single_query function passes the user-supplied question to _handle_chat_images, which identifies absolute file paths ending in image extensions and copies them to a temporary directory without any path validation. This could allow an attacker to read arbitrary files on the system if they can be tricked into having an image extension or if the agent is manipulated to read them. It is recommended to validate that any identified file paths are within the allowed project directory in the _handle_chat_images function.
| """Common setup logic for both main and optimize functions. | ||
|
|
||
| Args: | ||
| repo_path: Path to the repository | ||
| question_mode: If True, suppress INFO/DEBUG/WARNING logs (only show errors and direct output) | ||
| """ |
There was a problem hiding this comment.
According to the project's general rules, docstrings are not allowed. Please replace this multi-line docstring with a regular comment to adhere to the project's coding standards.
# Common setup logic for interactive, optimization, and single-query modes.
# When question_mode is True, logs are suppressed and only errors are sent to stderr.References
- Docstrings are not allowed in this project, as enforced by a pre-commit hook.
| async def main_async_single_query( | ||
| repo_path: str, batch_size: int, question: str | ||
| ) -> None: | ||
| """Initializes services and runs a single query in non-interactive mode.""" |
There was a problem hiding this comment.
The project's general rules state that docstrings are not allowed. To maintain consistency with the project's standards, please replace this docstring with a regular comment.
# Initializes services and runs a single query in non-interactive mode.References
- Docstrings are not allowed in this project, as enforced by a pre-commit hook.
Greptile SummaryAdds a Key changes:
Minor style notes:
Confidence Score: 5/5
Important Files Changed
Last reviewed commit: a879523 |
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
| def _setup_common_initialization(repo_path: str, question_mode: bool = False) -> Path: | ||
| """Common setup logic for both main and optimize functions. | ||
|
|
||
| Args: | ||
| repo_path: Path to the repository | ||
| question_mode: If True, suppress INFO/DEBUG/WARNING logs (only show errors and direct output) | ||
| """ |
There was a problem hiding this comment.
docstring added to existing function
Custom instructions state "No Comments or Docstrings" - code should be self-documenting. While the codebase has existing docstrings, consider whether this addition is necessary or if the code is clear without it.
Prompt To Fix With AI
This is a comment left during a code review.
Path: codebase_rag/main.py
Line: 162-168
Comment:
docstring added to existing function
Custom instructions state "No Comments or Docstrings" - code should be self-documenting. While the codebase has existing docstrings, consider whether this addition is necessary or if the code is clear without it.
How can I resolve this? If you propose a fix, please make it concise.Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!
| async def main_async_single_query( | ||
| repo_path: str, batch_size: int, question: str | ||
| ) -> None: | ||
| """Initializes services and runs a single query in non-interactive mode.""" |
There was a problem hiding this comment.
docstring added to new function
Custom instructions state "No Comments or Docstrings". While the existing codebase has docstrings, consider if this one is needed or if the function name and signature are self-explanatory.
Prompt To Fix With AI
This is a comment left during a code review.
Path: codebase_rag/main.py
Line: 787-790
Comment:
docstring added to new function
Custom instructions state "No Comments or Docstrings". While the existing codebase has docstrings, consider if this one is needed or if the function name and signature are self-explanatory.
How can I resolve this? If you propose a fix, please make it concise.Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!
| response = await rag_agent.run(question_with_context, message_history=[]) | ||
|
|
||
| # Output response to stdout | ||
| print(response.output) |
There was a problem hiding this comment.
uses print() instead of logger
Custom instructions prefer loguru over print(). However, this is likely intentional for clean stdout output in non-interactive mode (logs go to stderr). This is a standard CLI pattern for scriptable tools.
Prompt To Fix With AI
This is a comment left during a code review.
Path: codebase_rag/main.py
Line: 807
Comment:
uses `print()` instead of logger
Custom instructions prefer `loguru` over `print()`. However, this is likely intentional for clean stdout output in non-interactive mode (logs go to stderr). This is a standard CLI pattern for scriptable tools.
How can I resolve this? If you propose a fix, please make it concise.Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!
2 participants
Add --ask-agent / -a option to the start command. When provided, runs a single query via main_async_single_query (which uses question_mode logging to stderr) and prints the result to stdout, then exits. Useful for scripting without launching the full interactive REPL.
Dependency graph
Merge in this order: