feat: Improve Gemini error handling and add status indicators #24
Conversation
- Handle Gemini quota errors more gracefully, logging details at DEBUG. - Remove redundant "(Model stopped with no text)" message. - Add rich status spinner for thinking and tool execution phases. - Update tests to reflect logging changes and mock console status. - Restore missing add_to_history/clear_history methods. - Fix resulting test failures.
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Hello @ipv1337, 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!
Summary of Changes
This pull request focuses on improving the error handling and user experience of the Gemini agent. It introduces a rich status spinner to indicate thinking and tool execution phases, handles Gemini quota errors more gracefully by logging details and switching to a fallback model, and removes a redundant message. Additionally, it restores missing add_to_history and clear_history methods, and updates tests to reflect the changes.
Highlights
- Error Handling: Improved Gemini quota error handling with detailed logging and fallback mechanism.
- User Experience: Added a rich status spinner for thinking and tool execution phases to provide better feedback to the user.
- History Management: Restored missing
add_to_historyandclear_historymethods for managing conversation history. - Model Updates: Updates the fallback model to
gemini-2.0-flashand increases theOLLAMA_MAX_CONTEXT_TOKENSto 80000.
Changelog
Click here to see the changelog
- TODO_gemini_loop.md
- Removed outdated TODO items related to sequential tool calls, indicating progress on that front.
- src/cli_code/models/gemini.py
- Updated the fallback model to
gemini-2.0-flash. - Implemented a status spinner using
richto indicate agent thinking and tool execution. - Improved error handling for Gemini quota errors, including switching to a fallback model and logging details.
- Removed the redundant '(Model stopped with no text)' message.
- Refactored the agent loop to prioritize STOP reason checking and handle cases where the model stops without providing text content.
- Restored the
add_to_historyandclear_historymethods. - Enhanced tool execution confirmation by providing more detailed error messages and ensuring proper history management.
- Added detailed logging of quota errors at the debug level.
- Updated the fallback model to
- src/cli_code/models/ollama.py
- Increased
OLLAMA_MAX_CONTEXT_TOKENSfrom 8000 to 80000.
- Increased
- tests/models/test_gemini.py
- Updated the
FALLBACK_MODEL_NAME_FROM_CODEto match the new fallback model name insrc/cli_code/models/gemini.py. - Modified the
mock_consolefixture to provide a functional status context manager with anupdatemethod. - Refactored mocking strategy to use
monkeypatchinstead ofmockerfor better control and clarity. - Adjusted assertions in
test_generate_simple_tool_callto reflect the updated agent loop logic and history management. - Improved assertions in
test_generate_user_rejects_editto verify the presence of the rejection message in the history. - Fixed issues with mocking and assertion in multiple tests to ensure correct behavior and prevent false positives.
- Updated the
- tests/models/test_model_error_handling_additional.py
- Updated the
mock_count_tokensreturn value to 90000 to be higher than the updatedOLLAMA_MAX_CONTEXT_TOKENSvalue.
- Updated the
Using Gemini Code Assist
The 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 in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.
| Feature | Command | Description |
|---|---|---|
| Code Review | /gemini review |
Performs a code review for the current pull request in its current state. |
| Pull Request Summary | /gemini summary |
Provides a summary of the current pull request in its current state. |
| Comment | @gemini-code-assist | Responds in comments when explicitly tagged, both in issue comments and review comments. |
| Help | /gemini help |
Displays a list of available commands. |
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 .gemini/ folder in the base of the repository. Detailed instructions can be found here.
Limitations & Feedback
Gemini Code Assist is currently in preview and 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 to provide feedback.
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.
The agent thinks, a spinner bright,
Tools are called, in digital night.
Errors loom, but code stands tall,
A fallback waits, to heed the call.
Footnotes
-
Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩
There was a problem hiding this comment.
Code Review
The pull request introduces several improvements to the Gemini model integration, including enhanced error handling, status indicators, and test updates. The changes appear well-structured and address the issues outlined in the pull request description. However, some modifications require careful review to ensure correctness and maintainability.
Summary of Findings
- Model Name Update: The fallback model name has been updated from
gemini-1.5-flash-latesttogemini-2.0-flash. Ensure this change is intentional and aligns with the desired model behavior and capabilities. - STOP Reason Handling: The logic for handling the
STOPfinish reason has been modified to check for final text content before settingtask_completed. Verify that this change correctly handles cases where the model stops without providing text. - Tool Execution Status Update: The status spinner is now updated before and after tool execution. Confirm that these updates provide useful feedback to the user and do not introduce any performance issues.
- Quota Error Logging: Detailed quota error information is now logged at the DEBUG level. This is beneficial for debugging but ensure that sensitive information is not inadvertently logged.
- Help Text Generator: The
_get_help_textmethod has been replaced with apassstatement. This removes the detailed help text. Determine if this is intentional and if an alternative help mechanism is in place.
Merge Readiness
The pull request is not quite ready for merging. The changes related to the STOP reason handling and the removal of the help text generator should be carefully reviewed and verified. Additionally, the potential logging of sensitive information in quota errors should be addressed. Once these issues are resolved, the pull request can be considered for merging. I am unable to approve this pull request, and users should have others review and approve this code before merging.
| def _get_help_text(self) -> str: | ||
| """Generates comprehensive help text for the command line interface.""" | ||
| # Get tool descriptions for the help text | ||
| tool_descriptions = [] | ||
| for tool_name, tool_instance in AVAILABLE_TOOLS.items(): | ||
| desc = getattr(tool_instance, "description", "No description") | ||
| # Keep only the first line or a short summary | ||
| if "\n" in desc: | ||
| desc = desc.split("\n")[0].strip() | ||
| # Format as bullet point with tool name and description | ||
| tool_descriptions.append(f" • {tool_name}: {desc}") | ||
|
|
||
| # Sort the tools alphabetically | ||
| tool_descriptions.sort() | ||
|
|
||
| # Format the help text to be comprehensive but without Rich markup | ||
| help_text = f""" | ||
| Help | ||
| Interactive Commands: | ||
| /exit - Exit the CLI tool | ||
| /help - Display this help message | ||
| CLI Commands: | ||
| gemini setup KEY - Configure the Gemini API key | ||
| gemini list-models - List available Gemini models | ||
| gemini set-default-model NAME - Set the default Gemini model | ||
| gemini --model NAME - Use a specific Gemini model | ||
| Workflow: Analyze → Plan → Execute → Verify → Summarize | ||
| Available Tools: | ||
| {chr(10).join(tool_descriptions)} | ||
| Tips: | ||
| • You can use Ctrl+C to cancel any operation | ||
| • Tools like 'edit' and 'create_file' will request confirmation before modifying files | ||
| • Use 'view' to examine file contents before modifying them | ||
| • Use 'task_complete' to signal completion of a multi-step operation | ||
| For more information, visit: https://github.com/BlueCentre/cli-code | ||
| """ | ||
|
|
||
| return help_text | ||
| # Implementation of _get_help_text method | ||
| pass |
There was a problem hiding this comment.
The implementation of _get_help_text has been replaced with a pass statement. This removes the detailed help text. Is this intentional? If so, ensure that an alternative help mechanism is in place, or consider restoring the help text generation logic.
# Implementation of _get_help_text method
pass
src/cli_code/models/gemini.py
Outdated
| if final_text.strip(): # If there WAS text content with the STOP | ||
| log.info("Model stopped with final text content.") | ||
| final_summary = final_text.strip() | ||
| task_completed = True | ||
| break # Exit loop immediately on STOP with text | ||
| else: | ||
| # log.warning("Model stopped (finish_reason=STOP) but provided no text content. Letting loop continue or finish naturally.") # Removed warning | ||
| # Do NOT set final_summary here | ||
| # Do NOT set task_completed = True here | ||
| # Do NOT break here - let the loop potentially timeout or handle unexpected exit later | ||
| pass # Continue processing other parts or finish loop iteration |
There was a problem hiding this comment.
Consider adding a log message at the warning level to indicate that the model stopped without providing text content. This could help in debugging and understanding model behavior in such cases.
else:
# log.warning("Model stopped (finish_reason=STOP) but provided no text content. Letting loop continue or finish naturally.") # Removed warning
log.warning("Model stopped (finish_reason=STOP) but provided no text content.")
# Do NOT set final_summary here
# Do NOT set task_completed = True here
# Do NOT break here - let the loop potentially timeout or handle unexpected exit later| MAX_OLLAMA_ITERATIONS = 5 # Limit tool call loops for Ollama initially | ||
| SENSITIVE_TOOLS = ["edit", "create_file"] # Define sensitive tools requiring confirmation | ||
| OLLAMA_MAX_CONTEXT_TOKENS = 8000 # Example token limit for Ollama models, adjust as needed | ||
| OLLAMA_MAX_CONTEXT_TOKENS = 80000 # Example token limit for Ollama models, adjust as needed |
There was a problem hiding this comment.
The OLLAMA_MAX_CONTEXT_TOKENS has been increased from 8000 to 80000. Please confirm that this value is appropriate for the Ollama models being used and that it does not lead to excessive memory consumption or performance issues.
| MAX_OLLAMA_ITERATIONS = 5 # Limit tool call loops for Ollama initially | |
| SENSITIVE_TOOLS = ["edit", "create_file"] # Define sensitive tools requiring confirmation | |
| OLLAMA_MAX_CONTEXT_TOKENS = 8000 # Example token limit for Ollama models, adjust as needed | |
| OLLAMA_MAX_CONTEXT_TOKENS = 80000 # Example token limit for Ollama models, adjust as needed | |
| OLLAMA_MAX_CONTEXT_TOKENS = 80000 # Example token limit for Ollama models, adjust as needed |
Code Coverage Report📊 Current Coverage: 90.10% Detailed coverage analysis is available in SonarCloud Coverage Change DetailsThis shows code coverage for changes in this PR. To improve coverage, consider adding tests for new or modified code. |
Code Coverage Report📊 Current Coverage: 90.10% Detailed coverage analysis is available in SonarCloud Coverage Change DetailsThis shows code coverage for changes in this PR. To improve coverage, consider adding tests for new or modified code. |
|
Handle Gemini quota errors more gracefully, logging details at DEBUG.
Remove redundant "(Model stopped with no text)" message.
Add rich status spinner for thinking and tool execution phases.
Update tests to reflect logging changes and mock console status.
Restore missing add_to_history/clear_history methods.
Fix resulting test failures.