TabbyML, a self-hosted AI coding assistant designed as an open-source alternative to GitHub Copilot, is currently grappling with significant model integration challenges. The project, which operates without external databases or cloud services, has seen a notable increase in user-reported issues related to model loading and server stability.
The recent surge in GitHub issues highlights critical concerns around model integration and deployment environments. Users have reported problems with specific models like Codestral-22B and StarCoder-1B, which are not functioning as expected or causing server hangs (#2897). Additionally, there is a strong demand for enhanced documentation and features such as customizable context lengths and improved error handling.
Meng Zhang (wsxiaoys):
Aliang Liang (liangfung):
Zhiming Ma (icycodes):
Xiang Wallace (xxs-wallace):
Autofix CI Bot:
The TabbyML project is actively evolving but must address these critical integration and documentation issues to enhance user experience and maintain momentum.
| Developer | Avatar | Branches | PRs | Commits | Files | Changes |
|---|---|---|---|---|---|---|
| Meng Zhang |  |
18 | 79/67/6 | 120 | 531 | 15116 |
| aliang |  |
3 | 27/26/0 | 32 | 41 | 6986 |
| Zhiming Ma |  |
3 | 24/25/0 | 29 | 111 | 6206 |
| xxs-wallace |  |
4 | 9/6/0 | 14 | 42 | 2299 |
| Wang Zixiao |  |
2 | 1/2/0 | 4 | 16 | 673 |
| autofix-ci[bot] |  |
9 | 0/0/0 | 13 | 23 | 406 |
| Wei Zhang |  |
1 | 4/3/0 | 3 | 8 | 235 |
| Jackson Chen |  |
1 | 6/3/1 | 3 | 8 | 125 |
| Vladislav Bykov |  |
1 | 1/1/0 | 1 | 5 | 52 |
| Sherlock Xu |  |
1 | 1/1/0 | 1 | 3 | 7 |
| MeIchthys |  |
1 | 2/2/0 | 2 | 2 | 6 |
| Christian Clauss |  |
1 | 3/1/1 | 1 | 1 | 4 |
| richard-jfc |  |
1 | 1/1/0 | 1 | 1 | 2 |
| Richard Ginsberg |  |
1 | 1/1/0 | 1 | 1 | 2 |
| Vali Malinoiu (0x4139) | 0 | 1/0/0 | 0 | 0 | 0 | |
| ingend88 (jinal88) | 0 | 2/0/2 | 0 | 0 | 0 | |
| Ikko Eltociear Ashimine (eltociear) | 0 | 0/0/1 | 0 | 0 | 0 | |
| cinchen (eryue0220) | 0 | 1/0/0 | 0 | 0 | 0 | |
| Joshua Herrera (sirebellum) | 0 | 0/0/1 | 0 | 0 | 0 | |
| None (moqimoqidea) | 0 | 1/0/1 | 0 | 0 | 0 |
PRs: created by that dev and opened/merged/closed-unmerged during the period
| Timespan | Opened | Closed | Comments | Labeled | Milestones |
|---|---|---|---|---|---|
| 7 Days | 12 | 6 | 28 | 2 | 1 |
| 30 Days | 54 | 29 | 129 | 6 | 1 |
| 90 Days | 145 | 95 | 462 | 14 | 1 |
| 1 Year | 333 | 200 | 1039 | 48 | 3 |
| All Time | 641 | 503 | - | - | - |
Like all software activity quantification, these numbers are imperfect but sometimes useful. Comments, Labels, and Milestones refer to those issues opened in the timespan in question.
The TabbyML project has seen a surge in GitHub issue activity, with 138 open issues currently. Notably, several recent issues highlight critical bugs and enhancements, including problems with model loading, server crashes, and requests for new features like support for additional models and improved logging. A recurring theme is the integration of various models and the challenges users face when deploying in different environments, particularly regarding GPU utilization and configuration.
Several issues indicate that users are struggling with specific models returning unexpected responses or failing to load entirely. For instance, issues related to the Codestral model not functioning as expected and the StarCoder-1B model causing server hangs are significant concerns. Additionally, there are multiple requests for better documentation and features, such as customizable context lengths and improved error handling.
Issue #2923: Tabby Edit should be able to use tab or other key to select and edit prompt
Issue #2908: Stuck in the 'Starting' process
Issue #2902: Code generation frequently
Issue #2897: chat_template Not Functioning in Tabby Server v14+ Versions
Issue #2894: Integration with Notion Database
Issue #2908
Issue #2902
Issue #2897
Issue #2894
Issue #2877: [VSCode] Enhance Answer Engine to Auto-Add References from Open File
The data reveals several themes among the issues:
These themes suggest that while TabbyML is rapidly evolving, there are significant areas for improvement in both functionality and user guidance.
This analysis provides a snapshot of current issues affecting the TabbyML project, highlighting user concerns and areas where enhancements could significantly improve the overall experience.
The dataset contains a comprehensive list of 50 open pull requests (PRs) for the TabbyML/tabby repository, along with a history of 2,164 closed PRs. The current open PRs focus on various features, bug fixes, and enhancements, particularly in the areas of user interface improvements, backend functionality, and integration with development environments.
PR #2928: feat(webserver): add support for deleting expired ephemeral threads
Created by Meng Zhang, this PR introduces functionality to delete expired ephemeral threads. It is marked as a cherry-pick and has a size increase of approximately 67 KB.
PR #2927: ci: reuse llama.cpp distributions for windows cuda packaging
This draft PR aims to optimize the CI process by reusing existing distributions for Windows CUDA packaging.
PR #2926: feat(vscode): add an edit button in inline edit history
This PR adds an edit button to the inline edit history in VSCode, addressing issue #2923.
PR #2924: chore(auth): allow delete_thread_message_pair auth with auth_token
This PR enhances authentication capabilities by allowing deletion of thread message pairs using an auth token.
PR #2921: feat(vscode): add proxy support for vscode
A draft PR that implements proxy support in VSCode, addressing issue #2893.
PR #2919: refactor(crawler): add timeout for crawler job
This PR introduces a timeout feature to prevent jobs from blocking the background loop indefinitely.
PR #2917: feat(vscode): use CommandPalette to display tabby menu
This PR integrates the Tabby menu into the VSCode Command Palette for improved accessibility.
PR #2910: feat(ui): integrate thread api in Chat side panel
A draft PR aimed at integrating the thread API into the chat side panel UI.
PR #2903: blog: add introduce tabby edit
This PR adds a blog post introducing the Tabby editing feature.
PR #2898: feat(document): add web document
Introduces a new web document feature, which includes unique constraints on fields and preset documents for crawling.
PR #2888: feat(crawler): add preset web crawler urls
Adds preset URLs for users to crawl, enhancing the crawler's functionality.
PR #2878: feat(vscode): adding auto-generate button in source control
This PR improves user experience in VSCode by adding an auto-generate button for commit messages.
PR #2852: refactor(webserver): remove deprecated /v1beta/answer route
A cleanup effort that removes outdated API routes from the web server.
PR #2839: Fix typos discovered by codespell
A minor update that fixes various typos across the codebase.
PR #2312: chore: adjust answer response interface
This long-standing open PR is significantly behind main and requires attention to bring it up-to-date with ongoing changes.
16-50. Additional older PRs range from feature additions (like support for various programming languages) to minor fixes and documentation updates, indicating ongoing development and refinement of the Tabby project.
The current landscape of open pull requests within the TabbyML/tabby repository reveals several key themes and areas of focus:
A significant portion of the open pull requests is dedicated to enhancing user experience across various platforms, particularly within VSCode and other IDEs. For instance, PRs like #2926 (adding an edit button) and #2919 (adding proxy support) indicate a strong emphasis on improving usability and functionality within these environments. Additionally, features such as thread integration (#2910) and enhanced command palette support (#2917) reflect a commitment to making Tabby more accessible and efficient for developers.
Several pull requests focus on backend functionality, such as managing ephemeral threads (#2928), refining authentication processes (#2924), and enhancing crawler capabilities (#2898). These improvements suggest an ongoing effort to optimize performance and reliability within the system's core functionalities, which are critical for maintaining user trust and satisfaction.
The presence of refactoring efforts (e.g., PRs #2919 and #2852) indicates a proactive approach to code quality and maintainability. By removing deprecated routes and improving existing code structures, the team is ensuring that the codebase remains clean and manageable as new features are added.
The active participation in discussions around pull requests demonstrates a collaborative development environment where contributors are encouraged to provide feedback and suggestions. The presence of comments from multiple team members on various PRs indicates a healthy review process that fosters knowledge sharing and collective problem-solving.
Some older pull requests, such as #2312, have not seen recent activity or merging attempts despite being open for an extended period (over 80 days). This could indicate potential issues with alignment on implementation details or priorities within the team. Addressing these older requests should be prioritized to avoid stagnation in development progress.
While there is a steady influx of new pull requests, there appears to be a lack of recent merge activity on some fronts, particularly concerning older contributions that may require additional review or updates before they can be integrated into the main branch. This could lead to contributor frustration if not addressed promptly.
Overall, the current state of open pull requests in the TabbyML/tabby repository reflects a vibrant development environment focused on enhancing user experience through feature-rich updates while maintaining code quality through regular refactoring efforts. However, attention should be given to older pull requests that may need resolution or closure to keep the project moving forward effectively.
Meng Zhang (wsxiaoys)
Aliang Liang (liangfung)
Zhiming Ma (icycodes)
Xiang Wallace (xxs-wallace)
Autofix CI Bot (autofix-ci[bot])
Richard Ginsberg (richginsberg)
Christian Clauss (cclauss)
Wei Zhang (zwpaper)
Jackson Chen (Sma1lboy)
Sherlock Xu (Sherlock113)
Vladislav Bykov (VladislavNekto)
Wang Zixiao (wwayne)
The development team is highly active, with a strong focus on feature enhancements, bug fixes, and improving user experience across different platforms. Collaboration among team members is evident, contributing to a dynamic development environment that supports rapid iteration and continuous improvement of the Tabby project.
Summary
Quantified Reports
Detailed Reports