Find meaningful ways to contribute

This issue is a roadmap task focused on researching and documenting vLLM's KV connector. The goal is to understand its internal workings, including its KV cache memory layout and how it differs from external KV caches, and to write comprehensive documentation for it.

This GitHub issue serves as a tracking list for Homebrew formulae that have available updates but cannot be updated due to build or test failures. It categorizes these issues and highlights where community assistance is needed to resolve them. The issue also includes links to pull requests that show progress in updating these formulae.

A test in the `PerformanceCounterCategory_GetCounters` test suite is failing on an ARM64 Windows build. The error indicates that a required performance counter category, 'PerformanceCounterCategory_GetCounters_Category', does not exist, suggesting a potential race condition or setup issue during test execution.

This feature request aims to improve clarity by replacing the generic `standalone` configuration option with a new, more descriptive option. This new option will allow users to explicitly specify the desired VISA backend, making the configuration more understandable and flexible.

This issue requests the addition of a new contributor, @Swar-42, to the project's contributor list. The task involves modifying the `en.yaml` file to include the contributor's GitHub username and image ID, following specific instructions for extracting the image ID from their GitHub profile.

This issue proposes the creation of a validating webhook for the Burrito project. The webhook will be responsible for validating resources before they are applied, specifically focusing on the `remediationStrategy` enum and complex interactions between various options.

The desktop UI is inefficiently re-rendering on every streaming update due to how session state is managed and subscribed to. Specifically, the entire session object is recreated on changes, and many components subscribe to this broad object or the entire store, leading to unnecessary re-renders of the sidebar, message list, and composer. The issue proposes batching thinking deltas, moving timer logic to a dedicated component, and optimizing sidebar subscriptions to improve performance.

This issue requests the implementation of a new Git provider for Bitbucket. It specifically mentions building upon the new structural changes introduced in a recent pull request, indicating a need to integrate with an existing framework.

This issue proposes adding a new configuration setting to the Burrito tool. This setting would prevent Burrito from executing 'apply' operations if the 'plan' output indicates that any resources are scheduled for destruction. The goal is to enhance safety and facilitate adoption on critical Terraform code by preventing accidental resource deletion.

This issue proposes optimizing the desktop application's startup time by implementing lazy loading for several heavy JavaScript libraries and UI components. Currently, large libraries like Mermaid and KaTeX, along with the terminal, diff viewer, and Doctor panel, are statically imported, contributing to a significant startup JavaScript bundle size. By using dynamic `import()` only when these features are actually needed, the application's launch time and initial session load can be improved.

This issue proposes enhancing the Review Gate section in the settings modal to expose and manage reviewer fan-out and reviewer agents. Currently, selecting multi-model review agents is a guessing game. The proposed changes will provide a clear checkbox list of configured agents with their engine/model, helper text for guidance, and warnings for potential configuration issues.

This issue proposes enhancing the relay-side audit logging by introducing a leveled stream instead of a single-shape log. It aims to capture more granular information about relay activities, including lifecycle events, executed commands, and the reasoning behind them, with robust redaction for sensitive data. The changes will be reflected in both the `bin/contributor-relay.js` script and the `hivectl hives audit` command.

This issue proposes two enhancements for the Auto Optimize tool in Avogadro. Firstly, it should mirror the Manipulate tool's behavior by using the Alt key to increase the step size when moving selected atoms. Secondly, the tool should only signal a molecule change when atoms are actually moved, rather than for every key press, including ignored ones.

The SECURITY.md file instructs users to report vulnerabilities privately but fails to provide any contact information or channels for doing so. This makes it impossible for users to follow the stated policy, likely leading to public disclosure of security issues. The proposed solution is to update the SECURITY.md file with a valid contact method, such as GitHub's Private Vulnerability Reporting feature and an email address, similar to another repository within the same organization.

This issue is a beginner-friendly task for Hacktoberfest, asking contributors to add a new Japanese "false friend" pair to a JSON file. No coding is required, and the contribution can be made directly from the browser in under a minute.

This issue proposes adding a new configuration option `blocked-realm-prefixes` to `registries.conf` for ORAS. This option will allow users to specify a list of realm prefixes that should be blocked for specific registries, supporting both CIDR notation and named sets like 'default' and 'private'. The implementation involves modifying configuration structures, validation logic, and documentation.

The `responseOverride` feature in Gateway API is failing to redirect requests when the origin scheme (HTTPS) does not match the scheme of the `redirect.hostname` (HTTP). This prevents internal redirects to HTTP-only hostnames from HTTPS origins.

A bug in `BoundedRangeFunction` allows for integer overflow when calculating range sizes near `Long.MAX_VALUE` or `Long.MIN_VALUE`. This bypasses memory limits, enabling Pebble to allocate an excessive number of elements and crash the JVM with an OutOfMemoryError.

On Windows, the `repowise uninstall` command incorrectly displays configured database URLs with a backslash instead of a forward slash. This is due to the database locator being stored as a `Path` object, which on Windows converts forward slashes to backslashes. The fix involves ensuring the database locator is consistently treated as a string across all platforms.

This issue points out a significant omission in the MDN documentation for the Popover API: the lack of information regarding accessibility semantics and screen reader exposure. The reporter expects details on how `popovertarget` and `popover` attributes are announced, and whether traditional ARIA attributes like `aria-controls`, `aria-haspopup`, or `aria-role` are still necessary when using the Popover API.

This issue proposes adding ambient detection for llama.cpp's `llama-server` to mirror the existing Ollama detection. The goal is to automatically identify a locally running `llama-server` by probing `localhost:8080` with a short-timeout TCP connection, making it a readily available provider within the Omnigent setup.

The user wants to dynamically adjust the number of output images generated per scene based on the number of frames within that scene. Currently, the `save-image -n` parameter applies a uniform number of images to all scenes, leading to inefficient image generation for scenes with varying frame counts. The proposed solution is to allow a parameter that dictates image output relative to scene length, such as one image per a specified number of frames.

When a GitHub Pull Request (PR) is re-bucketed via a webhook, it temporarily loses its closing references. This means that for a short period, until the next evaluation cycle, the PR will appear as if it has no linked issue in the activity monitor. This issue affects both draft/exempt PRs and pre-existing actionable PRs, and the fix aims to bring them to parity.

This issue aims to enable the `--lego` command to correctly read the method contract of Go interfaces. Currently, Go interface methods are captured but suppressed due to language allow-listing. The fix involves updating `src/serialize.h` to include Go in the method contract check and adding new tests to verify the functionality.

This issue details recommendations for improving JAFF based on an end-to-end user workflow experiment. The experiment involved importing reaction networks from GPUAstroChem, converting them, checking model properties, generating code, and integrating with CPU components. Several issues were identified across these steps, leading to suggestions for enhancements.

This issue proposes to enhance the tags pass in the language architecture cleanup by enabling the evaluation of string predicates like #eq?, #any-of?, and #not-any-of?. This change aims to reduce the need for C++ code to manage keyword lists for languages with keyword-headed calls, allowing these definitions to be moved into their respective SCM files. The goal is to achieve byte-identical output for all existing languages and to add new test fixtures to verify the predicate functionality.

This issue proposes adding a new environment variable, `COLLECTOR_RUNNING_TASKS_HEALTHY_THRESHOLD`, to define a limit for the number of concurrently running tasks. If this limit is exceeded, the collector will be reported as unhealthy. The default threshold is set to 120, with a rationale provided based on task creation and timeout intervals.

Currently, when a build-from-source component is created via the Portal, the initial build does not trigger automatically. Users are required to manually initiate the first build, as the existing auto-build setting only applies to commit-triggered builds. The suggested improvement is to automatically trigger the first build upon component creation.

The user is experiencing an issue where pressing Shift+Enter while typing in all caps in Chinese input mode results in an unintended conversion to full Pinyin. They are requesting the ability to disable this Shift+Enter shortcut, as they believe most experienced double Pinyin users do not require this auto-conversion feature.