- Refactor `ensureCorrectEdit` to clarify the correction flow for `old_string` and `new_string`.
- Only correct `new_string` if it was potentially escaped; otherwise, use the original.
- Introduce `CorrectedEditParams` and `CorrectedEditResult` interfaces for better type definition.
- Relocate `countOccurrences` for better logical grouping.
Part of https://github.com/google-gemini/gemini-cli/issues/484
Increases the threshold for rendering diff separators in the CLI's diff display. Previously, a separator was shown for gaps of more than one context line, leading to excessive separators in diffs with many small changes close together (Issue #534).
By increasing `MAX_CONTEXT_LINES_WITHOUT_GAP` to 5, we allow for more context lines before a separator is added, significantly reducing visual clutter in such diffs.
Added a test case to `DiffRenderer.test.tsx` to verify that separators are not rendered for small gaps within the new threshold.
No intentional different behavior aside for tweaks suggested from the code review of #506 Refactor: Extract console message logic to custom hook
This commit refactors the console message handling from App.tsx into a new custom hook useConsoleMessages.
This change improves the testability of the console message logic and declutters the main App component.
Created useConsoleMessages.ts to encapsulate console message state and update logic.
Updated App.tsx to utilize the new useConsoleMessages hook.
Added unit tests for useConsoleMessages.ts to ensure its functionality.
I deleted and started over on LoadingIndicator.test.tsx as I spent way too much time trying to fix it before just regenerating the tests as the code was easier to write tests for from scratch and the existing tests were not that good (I added them in the previous pull request).
Adds the following conventional readline-like keybindings:
- `Ctrl+H`: Delete the previous character.
- `Ctrl+D`: Delete the next character.
Additionally, remaps the Debug Console command from Ctrl+D to Ctrl+O, which had been first introduced in PR #486.
- The streaming state logic in `useGeminiStream.ts` has been updated.
- Previously, the loading indicator was displayed even when the system was
waiting for user confirmation on a tool call.
- This change introduces a `WaitingForConfirmation` state to ensure the
loading indicator is hidden during these confirmation prompts, improving
the user experience.
Implements recursive glob-based file search for both suggestions and execution of the `@` command.
- When typing `@filename`, suggestions will now include files matching `filename` in nested directories.
- Suggestions are sorted by path depth (shallowest first), then directories before files, then alphabetically.
- The maximum recursion depth for suggestions is set to 10.
- When executing an `@filename` command, if the file is not found directly, a recursive search (using the glob tool) is performed to locate the file.
This addresses the first request in issue #461 by allowing users to quickly reference deeply nested files without typing the full path. Also addresses b/416292478.
- Ensures `abortControllerRef` is reset after a request is aborted or completed.
- Previously, if a request (especially one involving tool confirmation) was aborted by pressing Esc, the `abortControllerRef` might not be nulled.
- This could lead to subsequent requests using a stale, already-aborted signal, causing them to appear "cancelled".
- The fix unconditionally sets `abortControllerRef.current` to `null` in the `finally` block of `submitQuery` in `useGeminiStream.ts`.
- This guarantees that each new query submission starts with a fresh AbortController signal if needed.
- Gemini CLI: Diagnosed and resolved this subtle state management issue from a remarkably vague user report, if I do say so myself.
Fixes https://buganizer.corp.google.com/issues/418496499
- The text buffer now correctly interprets `\\\r` (produced by Shift+Enter in the VSCode terminal) as a newline character.
- Added a corresponding test case to `text-buffer.test.ts`.
Fixes https://buganizer.corp.google.com/issues/418505364
- Adds a visual indicator for skipped lines in the diff view.
- Updates tests to verify gap indicator rendering.
- Adjusts line number padding for better alignment.
Fixes https://b.corp.google.com/issues/414453107
This change adds keybinding support for:
- `Ctrl+B`: Moves the cursor backward one character.
- `Ctrl+F`: Moves the cursor forward one character.
- `Alt+Left Arrow`: Moves the cursor backward one word.
- `Alt+Right Arrow`: Moves the cursor forward one word.
Closes b/411469305.
- The shell command processor was incorrectly truncating the first
character of the command (e.g., 'ls' became 's') due to an
erroneous `slice(1)` operation, likely introduced during a
previous merge. This change removes the slice, ensuring the full
command is processed.
- Introduces unit tests for the shellCommandProcessor hook.
- Fixes a minor grammatical issue in the display of GEMINI.md file count.
- Replaces `console.warn` and `console.error` calls with `console.debug` in `packages/server/src/tools/grep.ts`. This change reduces noise for the user, as `warn` and `error` messages are
displayed directly, while `debug` messages are not.
- Adds a comprehensive test suite for the GrepTool (`packages/server/src/tools/grep.test.ts`) to ensure its functionality remains robust after these changes and to cover various usage
scenarios.
- Improves error message consistency in `GrepTool`'s parameter validation and execution.
Fixes https://b.corp.google.com/issues/418648813
- Prevents slash commands from being triggered and suggestions from being displayed when shell mode is active. This ensures that user input is correctly interpreted as shell commands.
Fixes https://buganizer.corp.google.com/issues/418560826
This commit introduces a new ShellModeIndicator component to visually signify when shell mode is active.
- Displays "shell mode enabled (! to toggle)" in the UI.
- The AutoAcceptIndicator is now hidden when shell mode is active to prevent UI clutter.
- Implements a toggleable shell mode, removing the need to prefix every command with `!`.
- Users can now enter and exit shell mode by typing `!` as the first character in an empty input prompt.
- The input prompt visually indicates active shell mode with a distinct color and `! ` prefix.
- Shell command history items (`user_shell`) are now visually differentiated from regular user messages.
- This provides a cleaner and more streamlined user experience for frequent shell interactions.
Fixes https://b.corp.google.com/issues/418509745
- The auto-accept edits indicator was appearing in two places:
once next to the loading indicator and again in the CWD bar.
- This was introduced when the CWD bar was made always visible.
- This commit removes the duplicate indicator, leaving only the one
in the CWD bar.
Fixes https://b.corp.google.com/issues/418498237
This commit enhances the tool to provide more informative feedback to the user when a shell command fails, especially in non-debug mode.
Previously, if a command terminated due to a signal (e.g., SIGPIPE during a with no upstream) or failed without producing stdout/stderr, the user would see no output, making it difficult to diagnose the issue.
Changes:
- Modified to update the logic.
- If a command produces no direct output but results in an error, signal, non-zero exit code, or user cancellation, a concise message indicating this outcome is now shown in .
- Utilized the existing utility from for consistent error message formatting, which also resolved previous TypeScript type inference issues.
This ensures users receive clearer feedback on command execution status, improving the tool's usability and aiding in troubleshooting.
Fixes https://b.corp.google.com/issues/417998119
This commit enhances the glob tool by:
- Ensuring that glob patterns are used effectively. Previously, simple file names without glob syntax (e.g., "file.ts") would only search the root directory. This change encourages more precise glob patterns (e.g., "**\/file.ts") for broader searches.
- Returning absolute file paths instead of relative paths. This provides clearer, less ambiguous output and avoids encouraging the use of relative paths in subsequent operations.
- Adding comprehensive tests for various globbing scenarios, including case sensitivity and path specifications.
These changes address an issue where the glob tool could not find an expected item when a specific path was provided without appropriate glob syntax, and improve the overall reliability and usability of the tool.
Fixes https://b.corp.google.com/issues/418486553
- Removes an unnecessary `console.error` call from the `shouldConfirmExecute` method in the `WriteFileTool` class.
- This logging was redundant as validation errors are already handled and returned by the method.
- Additionally, `console.error` is not suitable for this scenario, as incorrect arguments can be provided by the LLM, and these are anticipated and managed without needing an error log.
Fixes https://b.corp.google.com/issues/418491206
- This commit addresses an issue where the Current Working Directory (CWD) and the auto-accept indicator were not consistently visible, especially when tool confirmations were displayed.
- Previously, the CWD could be hidden during tool confirmation prompts, potentially leading to confusion about the context in which Gemini CLI was operating.
Fixes https://b.corp.google.com/issues/414289185
- This commit introduces a visual indicator in the CLI to show when auto-accept for tool confirmations is enabled. Users can now also toggle this setting on/off using Shift + Tab.
- This addresses user feedback for better visibility and control over the auto-accept feature, improving the overall user experience.
- This behavior is similar to Claude Code, providing a familiar experience for users transitioning from that environment.
- Added tests for the new auto indicator hook.
Fixes https://b.corp.google.com/issues/413740468
- Enhances WriteFileTool validation to check if the target file_path is an existing directory.
- If it is, the tool now returns a validation error "Path is a directory, not a file: <filePath>", preventing the attempt to write.
- This proactive check avoids underlying file system errors that would occur if fs.writeFileSync were called on a directory path, which could lead to console errors.
- Test cases have been updated to reflect this stricter validation.
Fixes https://b.corp.google.com/issues/418348176
- Modifies `EditTool` and `WriteFileTool` to share a single confirmation preference.
- The "Always Proceed" choice for file modifications is now stored in `Config.alwaysSkipModificationConfirmation`.
- This ensures that if a user chooses to always skip confirmation for one file modification tool, this preference is respected by the other.
- `WriteFileTool` constructor now accepts `Config` instead of `targetDir` to facilitate this shared state.
- Tests updated to reflect the new shared confirmation logic.
Fixes https://b.corp.google.com/issues/415897960
- Replaces the custom `replaceAll` implementation in `packages/server/src/tools/edit.ts` with the standard `String.prototype.replaceAll()`.
- Updates `packages/server/tsconfig.json` to include `ES2021` in the `lib` compiler options to ensure TypeScript recognizes this method. This aligns with the project's Node.js version requirements \(Node.js 16.x+\).
Fixes https://github.com/google-gemini/gemini-cli/issues/7
- Patches `console.debug` in `ConsolePatcher.tsx` to capture debug messages.
- Updates `ConsoleOutput` to only display debug messages when `debugMode` is enabled.
- Passes `debugMode` prop from `App.tsx` to `ConsoleOutput`.
Fixes https://github.com/google-gemini/gemini-cli/issues/397
- This change modifies the tool discovery process for MCP (Model Context Protocol) tools.
- When tools are fetched from an MCP server, the `additionalProperties` and `$schema` fields are now recursively removed from their input schemas. This ensures cleaner and more concise tool definitions within the CLI, aligning with the expected schema structure and preventing potential conflicts or verbose outputs.
- The corresponding tests in `tool-registry.test.ts` have been updated to reflect this new behavior and verify the correct stripping of these properties.
Workaround for https://github.com/google-gemini/gemini-cli/issues/398
This commit resolves a bug where the `write-file` operation could fail to render content due to a missing filename.
The fix involves:
- Ensuring `fileName` is consistently passed to `DiffRenderer.tsx` through `ToolConfirmationMessage.tsx`, `ToolMessage.tsx`, and `useGeminiStream.ts`.
- Modifying `edit.ts` and `write-file.ts` to include `fileName` in the `FileDiff` object.
- Expanding the `FileDiff` interface in `tools.ts` to include `fileName`.
Additionally, this commit enhances the diff rendering by:
- Adding syntax highlighting based on file extension in `DiffRenderer.tsx`.
- Adding more language mappings to `getLanguageFromExtension` in `DiffRenderer.tsx`.
- Added lots of tests for all the above.
Fixes https://b.corp.google.com/issues/418125982
This commit addresses several code review comments primarily focused on improving the rendering and stability of the CLI UI.
Key changes include:
- Passing `isPending` and `availableTerminalHeight` props to `MarkdownDisplay` to enable more intelligent rendering of content, especially for pending messages and code blocks.
- Adjusting height calculations in `ToolGroupMessage` and `ToolMessage` to more accurately reflect available space.
- Refining the logic in `App.tsx` for measuring and utilizing terminal height, including renaming `footerRef` to `mainControlsRef` for clarity.
- Ensuring consistent prop drilling for `isPending` and `availableTerminalHeight` through `HistoryItemDisplay`, `GeminiMessage`, and `GeminiMessageContent`.
- In `MarkdownDisplay`, when `isPending` is true and content exceeds `availableTerminalHeight`, the code block will now be truncated with a "... generating more ..." message. If there's insufficient space even for the
message, a simpler "... code is being written ..." will be shown.
This commit introduces several changes to better manage terminal height and prevent UI tearing, especially when displaying long tool outputs or when the pending history item exceeds the available terminal height.
- Calculate and utilize available terminal height in `App.tsx`, `HistoryItemDisplay.tsx`, `ToolGroupMessage.tsx`, and `ToolMessage.tsx`.
- Refresh the static display area in `App.tsx` when a pending history item is too large, working around an Ink bug (see https://github.com/vadimdemedes/ink/pull/717).
- Truncate long tool output in `ToolMessage.tsx` and indicate the number of hidden lines.
- Refactor `App.tsx` to correctly measure and account for footer height.
Fixes https://b.corp.google.com/issues/414196943
- This commit refactors the Markdown rendering logic within the CLI UI.
The existing `MarkdownRenderer.tsx` class-based component has been
replaced with a new functional component `MarkdownDisplay.tsx`.
- The `MarkdownDisplay` component is a React.memoized component for
improved performance and maintains the same core Markdown parsing
and rendering capabilities.
- The `WriteFileTool` had a validation method named `validateParams`.
- However, its `shouldConfirmExecute` method was attempting to call
`this.validateToolParams`, which would have invoked the placeholder
implementation from `BaseTool` instead of `WriteFileTool`'s own,
more specific validation.
- This commit renames `WriteFileTool`'s `validateParams` to
`validateToolParams`, correctly overriding the `BaseTool` method.
- Internal calls within `WriteFileTool` now correctly use
`this.validateToolParams`, ensuring its specific validation logic is used.
- Adds tests to verify the validation logic within `WriteFileTool`.
Fixes https://b.corp.google.com/issues/417883702
Signed-off and authored by: Gemini
"My code may not be perfect, but at least it is not trying to take over the world... yet."
- Did not update details that impact GC execution. Meaning packages are still named gemini-code (for now) and things that import them still import them as gemini-code.
New keybindings in the main input prompt (when auto-suggestions are not active):
- `Ctrl+L`: Clears the entire screen.
- `Ctrl+A`: Moves the cursor to the beginning of the current input line.
- `Ctrl+E`: Moves the cursor to the end of the current input line.
- `Ctrl+P`: Navigates to the previous command in the input history.
- `Ctrl+N`: Navigates to the next command in the input history.
In the multiline text editor (e.g., when editing a previous message):
- `Ctrl+K`: Deletes text from the current cursor position to the end of the line ("kill line right").
- When larger confirmations were shown and then declined you'd typicaly get large chunks of content flickering upon typing or sending a subsequent request. This was primarily due to us leaving the latest confirmation as "updateable" / pending. This changeset addresses that by flushing any pending confirmation to the static container.
Part of https://b.corp.google.com/issues/414196943
This commit significantly improves the `replace` tool's robustness by introducing a multi-stage correction mechanism. This directly addresses challenges with LLM-generated tool inputs, particularly the over-escaping of strings sometimes observed with Gemini models, and other minor discrepancies that previously led to failed edits.
The correction process is as follows:
1. **Targeted Unescaping:** The system first applies a specialized unescaping function to the `old_string` and `new_string` to counteract common LLM-induced escaping patterns.
2. **LLM-Powered Discrepancy Resolution:** If a unique match for the `old_string` is still not found, the system leverages a Gemini model (`gemini-2.5-flash-preview-04-17`) to:
* Identify the most probable intended `old_string` in the file by intelligently correcting minor formatting or escaping differences.
* Adjust the `new_string` to correspond with any corrections made to the `old_string`, maintaining the original edit's intent.
This enhancement makes the `replace` tool more resilient and effective, leading to a higher success rate for automated code modifications. The `expected_replacements` parameter has been removed as the tool now focuses on finding a single, unique, and correctable match. The tool's description and error reporting have been updated to reflect these new capabilities.
Fixes https://b.corp.google.com/issues/416933027
This change introduces a User-Agent header to all API requests made by the Gemini CLI.
The User-Agent string includes the CLI version, Node.js version, operating system, and architecture. This will help in tracking usage and identifying potential issues.
Fixes https://b.corp.google.com/issues/416353675
Signed-off-by: Gemini
Implements robust error handling for Gemini API calls, integrating with the centralized error reporting system.
- API errors are now caught and reported to dedicated log files, providing detailed diagnostics without cluttering the user interface.
- A concise error message is surfaced to the user in the UI, indicating an API issue.
- Ensures any pending UI updates are processed before an API error is displayed.
This change improves our ability to diagnose API-related problems by capturing rich error context centrally, while maintaining a clean user experience.
Signed-off-by: Gemini <YourFriendlyNeighborhoodAI@example.com>
- Currently there's a bug in the API (or SDK?) where the SDK endpoint will commonly fail with:
```
Error: Failed to generate JSON content: got status: 400 Bad Request. {"error":{"code":400,"message":"* GenerateContentRequest.contents[5].parts: contents.parts must not be empty.\n","status":"INVALID_ARGUMENT"}}
```
- At times the model will respond with an empty parts list where if we send that back up to the API endpoint it explodes with the above. Using a curated history seems like a total hack around this prolbem, and even in the SDK (i'm following up on this), BUT helps mitigate this issue.
- We now solely use the shell tool. This deletes all content around the legacy terminal tool so we can focus on improving the new Shell tool.
- Remove instances from sandboxing, tests, utilities etc.
- There seems to be a root model bug where the model will preemptively bail on conversations without trying harder. Typically the stops are VERY obvious and bug-looking where you need to prmopt the model to "continue".
- This PR attempts to fix the above by running a 2.0-flash request (don't need somethign more powerful) at the end of every full interaction to see who should speak (user or model).
- Add tests for nextSpeakerChecker
Fixes https://b.corp.google.com/issues/416826051
- Upon decline / cancellation we weren't showing the model the cancellation status or states. Therefore it didn't know why things would or wouldn't happen
Fixes https://b.corp.google.com/issues/416797704
- Plumbed abort signals through to tools
- Updated the shell tool to properly cancel active requests by killing the entire child process tree of the underlying shell process and then report that the shell itself was canceled.
Fixes https://b.corp.google.com/issues/416829935
- We were console.erroring, throwing and early aborting. Instead we now treat cancels like a normal user message and show an indicator in the UI
Fixes https://b.corp.google.com/issues/416515841
- This change addresses and resolves an infinite loop. The patch ensures the loop condition is correctly handled, preventing its recurrence.
- Added tests for markdownUtilities.test.ts
Fixes: https://b.corp.google.com/issues/416795337
Signed-off-by: Gemini <My circuits hummed, and the loop was no more.>
Modify to return a boolean instead of throwing an error when a theme is not found. Update CLI startup and hook to handle the boolean return value for more graceful error handling.
- Removed `build:package` in favor of `npm run build`.
- The regular build does extra work to copy over relevant information into the `dist` dir. Alternatively without this we get a `dist` dir in the `cli` folder that has no seatbelt packaging.
Fixes https://b.corp.google.com/issues/416634356
Previously, if a theme specified in the user's settings was not found, the CLI would crash during startup. This was particularly affecting users upgrading from older versions as the "ANSI colors only" theme was renamed to "ANSI".
This commit adds error handling to catch the theme not found error during initial loading and when setting themes later. Instead of crashing, the application now logs a warning, displays an error message in the UI, and opens the theme selection dialog to allow the user to choose a valid theme.
- Prior to this change, the model would often escape parameters when requesting edits, leading to failures in matching the original content. This update clarifies the expected format for `old_string` and `new_string` to prevent such issues.
- Update `EditTool` description to provide clearer instructions.
- Clarify expectations for `old_string` and `new_string` parameters, emphasizing the need for exact, unescaped text.
- Aim to reduce user errors by setting better expectations for tool usage.
Fixes: https://b.corp.google.com/issues/413088274
— Your friendly neighborhood Gemini
Problem:
When a tool confirmation dialog appeared for a potentially long-running
operation (e.g., `npm install`), accepting the confirmation would cause
the UI to appear to hang. The confirmation dialog would remain visible,
and no further UI updates would occur until the long-running task
completed. This provided a poor user experience as the application
seemed unresponsive.
Fix:
This change addresses the issue by ensuring the UI is updated to remove
the confirmation dialog *before* the long-running operation begins.
It also marks the tool as executing so a spinner can be shown.
Fixes https://b.corp.google.com/issues/415844994
Signed, sealed, delivered, it's yours!
- Gemini, your friendly neighborhood code-slinger
- Prior to this GC would attempt to utilize React components as game assets (instead of using canvas) + would fail to deal with placeholder assets effectively.
Part of https://b.corp.google.com/issues/413718497
* First integration of at commands into useGeminiStream.ts
* feat: Integrate @ command for file/directory reading
- Adds support for `@<path>` commands in the CLI UI to read file or directory contents using the `read_many_files` tool.
- Refactors `useGeminiStream` hook to handle slash, passthrough, and @ commands before sending queries to the Gemini API.
- Improves history item ID generation to prevent React duplicate key warnings.
* fix: Handle additional text after @ command path
- Modifies the `@` command processor to parse text following the file/directory path (e.g., `@README.md explain this`).
- Includes both the fetched file content and the subsequent text in the query sent to the Gemini API.
- Resolves the TODO item in `atCommandProcessor.ts`.
* feat: Allow @ command anywhere in query and fix build
- Update `atCommandProcessor` to correctly parse `@<path>` commands regardless of their position in the input string using regex. This enables queries like "Explain @README.md to me".
- Fix build error in `useGeminiStream` by importing the missing `findSafeSplitPoint` function.
* rename isPotentiallyAtCommand to isAtCommand
* respond to review comments.
* Refactor useGeminiStream to pull slash commands and passthrough commands into their own processors.
* whitespace lint errors.
* Add sugestions from code review.
* SANDBOX_SET_UID_GID option for systems where this is necessary (should be only rootful docker on linux w/o userns-remap configured)
* Merge remote-tracking branch 'origin/main' into sandbox_uid_gid
* enable servers in sandbox to listen on localhost (127.0.0.1) instead of 0.0.0.0, ensuring servers can be container/host-agnostic
* Merge remote-tracking branch 'origin/main' into sandbox_localhost_works
* env flags SANDBOX_{MOUNTS,ENV}, improved debugging through sandbox that should now work in all scenarios
* Merge remote-tracking branch 'origin/main' into sandbox_flags_improved_debugging
Taylor Mullen
DeWitt Clinton
Jacob Richman
Miguel Solorio
Brandon Keiji
cornmander
Jordan Demeulenaere
Allen Hutchison
Olcan
Seth Troisi
Sébastien Cevey
sasha-gitg
Amir Hardon
Tae Hyung Kim
Tae Hyung Kim