zhayujie/CowAgent
38.7
Weak · 14 November 2026
145k
lines of production code
Python
with TypeScript, JavaScript
8
measurements over time
What this system is
Features
Add Azure Cognitive Services voice integration
Users can now use Azure Cognitive Services for speech-to-text and text-to-speech capabilities. The new \voice/azure\ module supports automatic language detection for text-to-speech synthesis, allowing the system to select appropriate neural voices for Chinese, English, Japanese, Korean, German, French, and Spanish, while also providing configurable speech recognition settings.
voice/azure · high confidence
Add Baidu Qianfan chat and vision support
Users can now interact with Baidu's Qianfan platform (ERNIE models) via a new bot integration. This adds support for text-based chat using the ERNIE-5.1 model and multimodal vision capabilities for image analysis using models like ERNIE-4.5-Turbo-VL. The integration handles session management, token counting, and API configuration via standard settings.
models/qianfan · high confidence
Add Baidu Voice service with long-text synthesis and request timeouts
Users can now use Baidu for both speech recognition (ASR) and text-to-speech (TTS). The new implementation supports long-text synthesis via an asynchronous task API, allowing responses longer than the 1024-byte limit of the standard SDK. To improve reliability, all outbound HTTP requests are now bounded by a specific timeout, preventing stalled connections from hanging the conversation turn. Configuration for TTS parameters (speed, pitch, volume, voice type) is managed via a local \config.json\ file in the \voice/baidu\ directory.
voice/baidu · high confidence
Add Baidu and Youdao translation providers
Users can now select Baidu or Youdao as translation providers. The Baidu translator supports automatic language detection and includes a 10-second request timeout with clear reporting of HTTP failures and API errors after retries. The Youdao translator implements the v3 signature scheme, maps ISO 639-1 language codes to Youdao-specific codes (including support for Cantonese), and validates responses to ensure translations are returned.
translate · high confidence
Add Banwords plugin for filtering sensitive content
A new Banwords plugin has been added to the system, allowing users to filter sensitive words in both user messages and ChatGPT replies. The plugin supports two main actions: 'ignore' to silently drop messages containing banned terms, and 'replace' to mask sensitive words with asterisks and notify the user. It also includes a configurable 'reply\_filter' option to apply the same filtering logic to the AI's responses, with a default configuration provided in config.json.template and a sample word list in banwords.txt.template.
plugins/banwords · high confidence
Add Claude API integration with thinking model support
Introduces a new Claude API bot implementation that supports both legacy 'enabled' thinking modes and the newer 'adaptive' thinking controls for Claude 4.6+ models (such as claude-opus-5 and claude-sonnet-5). The integration handles streaming responses, surfaces real provider token usage, and includes specific logic to manage thinking budgets and max\_tokens defaults for different model generations.
models/claudeapi · high confidence
Add DashScope ASR and TTS support
Users can now use the DashScope API for speech recognition and text-to-speech via the new DashScopeVoice implementation, which supports the qwen3-asr-flash and qwen3-tts-flash models and handles audio format conversion for compatibility.
voice/dashscope · high confidence
Add Discord channel support
Users can now connect the bot to Discord via a new channel implementation. The feature supports direct messages and guild (group) chat, handling text, images, and file attachments. It uses the Discord Gateway for persistent connections, allowing the bot to operate behind NAT without requiring a public IP or callback URL. In guilds, messages are processed only when the bot is @mentioned or replied to, while direct messages are always processed. The implementation runs the Discord client in a dedicated thread to maintain compatibility with the existing synchronous architecture.
channel/discord · high confidence
Add Doubao (Volcengine Ark) model support
Users can now use the Doubao (Volcengine Ark) API as a chat and vision model provider. This change introduces the DoubaoBot implementation, which supports text-based conversations with session management, image analysis via the vision endpoint (with a 180-second timeout), and configuration via the \ark\_api\_key\ and \ark\_base\_url\ settings. The default model is set to \doubao-seed-2.1-pro\, and the implementation includes error handling for rate limits and authentication failures. Also in this area: Unified error-body parsing for DeepSeek, Doubao, and Moonshot bots; Fixes for model error handling, credential masking, and configuration fallbacks.
models/doubao · high confidence
Add Dungeon text-adventure plugin
A new 'Dungeon' plugin is introduced, enabling users to play interactive text-adventure games via chat. Users can start a session with \$开始冒险\ (optionally providing a custom background story) and continue the narrative by sending messages, which the AI incorporates into the story. The game can be ended at any time with \$停止冒险\. The plugin is disabled by default and requires a compatible chat bot backend (such as OpenAI or ChatGPT). Additionally, a 'Finish' plugin is added to handle unknown command prefixes by returning an error message.
plugins/dungeon · high confidence
Add Edge TTS voice synthesis support
Users can now use Microsoft Edge TTS for voice synthesis via the new EdgeVoice implementation in the voice/edge module. This adds a new text-to-speech provider that generates audio files using the edge-tts library, with error handling that returns a localized failure message instead of dropping the reply if synthesis fails.
voice/edge · high confidence
Add ElevenLabs text-to-speech support
Users can now generate voice responses using the ElevenLabs API. This change introduces a new voice provider that utilizes the \elevenlabs\ Python SDK (version 1.0.3) and the \eleven\_multilingual\_v2\ model. The implementation requires configuration of the \xi\_api\_key\ and \xi\_voice\_id\ in the application settings to enable text-to-voice conversion.
voice/elevent · high confidence
Add Google Voice support for speech recognition and text-to-speech
Users can now use Google services for voice interactions. The new GoogleVoice component enables speech-to-text conversion using the Google Speech Recognition API (defaulting to Chinese) and text-to-speech synthesis via gTTS. It also includes logic to generate unique filenames for audio outputs to prevent conflicts in multithreaded environments.
voice/google · high confidence
Add Hello plugin for customizable group interactions
Introduces the 'Hello' plugin, which allows users to customize automated messages for group events such as new member welcomes, departures, and 'pat-pat' interactions via configurable prompts. It supports setting fixed welcome messages for specific groups and integrates with LinkAI character descriptions if enabled. The plugin also handles basic text commands like 'Hello', 'Hi', and 'End' (which triggers an image creation).
plugins/hello · high confidence
Add LinkAI voice provider with multi-vendor TTS and configurable ASR
Users can now use the LinkAI platform for voice interactions, supporting text-to-speech via OpenAI, Doubao, and Baidu engines, and speech-to-text via Whisper, Doubao, or Baidu. The ASR engine is no longer hardcoded to Whisper-1; it respects the configured \voice\_to\_text\_model\ setting. Additionally, the TTS implementation now supports an optional \app\_code\ for workspace overrides and normalizes unsupported audio formats (like .silk) to .mp3 before sending to the API.
voice/linkai · high confidence
Add MiniMax TTS support
Introduces a new voice provider for MiniMax text-to-speech, implemented in \voice/minimax/minimax\_voice.py\. The implementation calls the MiniMax \/v1/t2a\_v2\ endpoint, handles SSE streaming of hex-encoded MP3 chunks, validates API error responses, enforces file size limits, and writes the resulting audio to a temporary directory.
voice/minimax · high confidence
Add MiniMax bot integration with M3 vision support
Users can now interact with the MiniMax language model family. The new \MinimaxBot\ implementation defaults to the MiniMax-M3 model and enables native multimodal vision capabilities for image analysis. A dedicated session manager handles conversation history and token counting for this provider.
models/minimax · high confidence
Add ModelScope bot integration with streaming and thinking model support
Users can now interact with ModelScope AI models (defaulting to Qwen/Qwen3.5-27B) via a new bot integration. This feature supports standard text replies, streaming responses, and 'thinking' models that expose internal reasoning content. It also includes image generation capabilities and session management with automatic token-based context truncation. Also in this area: Fix ModelScope stream error handling to properly signal failures.
models/modelscope · high confidence
Add OpenAI voice integration for speech-to-text and text-to-speech
Introduces the OpenAI voice backend (voice/openai/openai\_voice.py) to enable voice interactions using OpenAI's API. Users can now convert speech to text via the /audio/transcriptions endpoint (defaulting to gpt-4o-mini-transcribe) and convert text to speech via the /audio/speech endpoint (defaulting to the TTS-1 model). The implementation uses the native requests library instead of the OpenAI SDK, includes bounded request timeouts to prevent thread blocking, and ensures temporary audio files are properly managed.
voice/openai · high confidence
Add WeCom Bot channel with WebSocket and webhook support
Introduces a new WeCom (Enterprise WeChat) bot channel that supports both WebSocket long-connection and HTTP callback (webhook) modes. The implementation handles text, image, file, and mixed message types, including downloading and decrypting media payloads via AES-256-CBC. It includes built-in safeguards such as bounded media downloads, control character escaping for JSON parsing, and transient file management within the agent's temporary directory.
_channel/wecom\bot · high confidence
Add Xiaomi MiMo TTS provider support
Users can now use the Xiaomi MiMo service for text-to-speech synthesis. A new \MimoVoice\ implementation has been added to the \voice/mimo\ module, which calls the MiMo \/chat/completions\ API to generate audio. This feature requires configuring the \mimo\_api\_key\ and optionally \mimo\_api\_base\ and \tts\_voice\_id\ in the application config. Note that speech recognition (voice-to-text) is not supported by this provider.
voice/mimo · high confidence
Add Xunfei Voice integration for speech recognition and synthesis
Introduces a new voice provider backed by the Xunfei (iFlytek) cloud APIs, enabling both speech-to-text (ASR) and text-to-speech (TTS) capabilities. The change adds the core implementation files (\xunfei\_voice.py\, \xunfei\_asr.py\, \xunfei\_tts.py\) and a configuration template (\config.json.template\) requiring APPID, APIKey, and APISecret. The TTS stream includes a ping/pong mechanism to prevent hanging on silent providers, and the ASR module handles streaming audio frames for transcription.
voice/xunfei · high confidence
Add ZhipuAI voice integration for ASR and TTS
Users can now use ZhipuAI for voice recognition (ASR) and text-to-speech (TTS) by configuring the \zhipu\_ai\_api\_key\ and optionally \zhipu\_ai\_api\_base\. The new \ZhipuAIVoice\ class handles transcription via the \glm-asr-2512\ model and speech synthesis via \glm-tts\, including automatic audio format conversion to MP3 for ASR input and magic-byte sniffing for TTS output extensions.
voice/zhipuai · high confidence
Add custom OpenAI-compatible voice provider support
Users can now route Automatic Speech Recognition (ASR) and Text-to-Speech (TTS) through external OpenAI-compatible vendors. This new custom provider allows configuration of a specific API base and key, supports extra headers for authentication, and enforces strict timeouts to prevent stalled requests from hanging the conversation turn. It requires explicit model configuration for both voice-to-text and text-to-voice operations.
voice/custom · high confidence
Add offline text-to-speech support via Pytts
Introduces a new offline voice service using the Pytts3 library, allowing the system to convert text to audio files without external APIs. The implementation handles platform-specific initialization (Windows vs. Linux/macOS), sets default speech rate and volume, and ensures thread-safe file generation by using unique filenames based on timestamps and text hashes to prevent collisions during concurrent requests.
voice/pytts · high confidence
Add on-demand current time tool to preserve prefix cache
A new \TimeTool\ has been added to the agent's toolset, allowing the system to retrieve the current local date, time, weekday, and timezone on demand. This change moves time retrieval out of the static system prompt (which only contains today's date) to prevent cache invalidation on every turn, ensuring that time-sensitive queries are handled accurately without degrading conversation performance. The tool supports localization for weekdays in Chinese.
_agent/tools/current\time · high confidence
Add support for Alibaba Cloud (Aliyun) voice services
Users can now use Alibaba Cloud for both text-to-speech (TTS) and speech-to-text (STT) recognition. This change introduces the \AliVoice\ integration, which handles token generation, audio file management, and API communication with Aliyun's NLS gateway, allowing the system to process voice inputs and generate voice outputs using this provider.
voice/ali · high confidence
Added Tencent Cloud voice integration
Users can now use Tencent Cloud for voice services, including speech-to-text (ASR) and text-to-speech (TTS). This change introduces the \TencentVoice\ implementation and a configuration template (\config.json.template\) requiring Tencent Cloud credentials (SecretId, SecretKey) and voice type settings.
voice/tencent · high confidence
Added banwords search library
The banwords plugin now includes the WordsSearch module, which provides a trie-based algorithm for efficiently detecting and locating specific keywords within text. This addition enables the plugin to perform pattern matching against a set of defined ban words.
plugins/banwords/lib · high confidence
Centralized model routing, multi-provider support, and per-model reasoning controls
The models directory has been restructured to support multiple custom (OpenAI-compatible) providers via ID-based routing (bot\_type: "custom:\<id\>") while maintaining backward compatibility with the legacy single-provider mode. A new model catalog system allows per-provider model overrides and hidden presets, stored atomically to prevent data loss. Reasoning capabilities are now provider-specific, exposing native effort settings (e.g., low/high/max) for DeepSeek, Claude, Qwen, Kimi, and GLM models, with automatic handling of hybrid thinking models and Responses API routing for models like gpt-6-astra. The bot factory centralizes instance creation, and session management includes token-aware context trimming.
models · high confidence
CowAgent initial release with multi-agent, skills, and unified web console
The repository has been rebranded from chatgpt-on-wechat to CowAgent, introducing a comprehensive Agent Harness architecture. This release adds multi-agent team support with per-agent memory and workspaces, a Skills system for one-click tool installation, and a self-evolution subsystem for automatic memory distillation. A unified Web console (default channel) now serves as the central hub for chat, model configuration, and channel management. The platform supports a wide range of LLM providers (DeepSeek, Claude, GPT, Gemini, etc.) and integrates with multiple IM channels including WeChat, Feishu, DingTalk, Telegram, and Discord. Deployment is simplified via a one-line installer script (run.sh) and Docker support, with a new config template and extensive documentation.
(repo-wide) · high confidence
Initial renderer shell with theme-aware layout and asset support
The desktop renderer now includes its foundational HTML shell and TypeScript asset declarations. Users will see the application load with a dark theme by default, respecting system preferences if configured, and the interface is structured to support a three-column layout with platform-specific branding. The renderer is also equipped to handle static assets like images and fonts locally without external CDNs.
desktop/src/renderer · high confidence
Introduce 'ls' tool for listing directory contents
Adds a new 'ls' tool that allows users to list directory contents with support for relative and absolute paths, alphabetical sorting, and entry size limits. The tool includes security checks to prevent direct access to sensitive configuration directories (e.g., \~/.cow) and provides helpful error hints for path resolution issues.
agent/tools/ls · high confidence
Introduce AgentAvatar and multi-agent UI components for desktop
The desktop client now supports multi-agent conversations with a dedicated visual identity and management interface. A new AgentAvatar component renders agent faces using uploaded images, the brand logo for the default agent, or deterministic colored initials, with cache-busting on avatar updates. This is integrated into new selector components: AgentSelector allows switching the active agent or adding teammates to group chats directly from the composer; AgentScopeSelect scopes data views (like Knowledge and Memory) to specific agents; and ChannelTeamSelect binds agents to channel instances with an owner-first hierarchy. The chat experience is enhanced with a ChatTimeline message navigator for quick jumping between questions, a ContextUsagePopover with a donut chart for monitoring context limits, and a ComposerChip for consistent selector styling. These changes enable users to manage agent rosters, collaborate in teams, and monitor context usage within the desktop app.
desktop/src/renderer/src/components · high confidence
Introduce Baidu Wenxin and Baidu Unit bot integrations
Added new bot implementations for Baidu's AI services: \BaiduWenxinBot\ integrates with the Wenxin (Ernie) text-generation API, supporting session management, model configuration, and explicit error handling for unsupported message types like image creation; \BaiduUnitBot\ provides a simpler interface to the Baidu Unit conversational API. Both implementations include token acquisition logic and basic session handling.
models/baidu · high confidence
Introduce Bash tool with Windows support, background execution, and robust output handling
The new Bash tool allows executing shell commands with platform-aware behavior: on Windows it uses cmd.exe (with PowerShell fallback for specific cases) and handles GBK/CP936 encoding to prevent mojibake, while on Unix it uses bash. It supports running long-lived processes in the background via run\_in\_background, with a registry to track, read, and kill jobs. Output is safely truncated and decoded, with soft exit codes (e.g., grep no matches) handled to avoid false failure reports. Security is enforced by blocking direct access to \~/.cow/.env and stripping sensitive environment variables. The tool also adapts its description and safety prompts based on the working directory and desktop mode.
agent/tools/bash · high confidence
Introduce CowAgent CLI with version 2.1.9
The CowAgent CLI is now available as a standalone command-line tool (invoked via \cow\) for managing the agent instance. This release introduces commands to start, stop, restart, and check the status of the agent, as well as view logs. It includes a \cow update\ command that performs a one-click source update from GitHub (git pull, dependency install, and restart) for supported environments, and a \cow version\ command that reports the current version (2.1.9). Additional capabilities include managing skills and knowledge bases, performing portable data backups and restores, and installing browser tools (Playwright + Chromium). The CLI also supports localizing its output based on the user's configured language.
cli · high confidence
Introduce CowAgent Desktop client with platform-specific code signing and data-driven theming
This change introduces the CowAgent Desktop application, a cross-platform Electron client built with React and TypeScript. It features a data-driven theme system that allows for platform-aware styling and customizable user bubbles, alongside a three-column layout. The release includes robust platform-specific code signing: macOS builds automatically sign nested Python backend binaries to satisfy Apple notarization, while Windows builds integrate a custom signing hook for all executables, supporting dry-run validation. The desktop client communicates with a local Python backend via HTTP and Server-Sent Events.
desktop · high confidence
Introduce DashScope integration with multimodal and custom base URL support
Users can now connect to the DashScope API via a new bot implementation. This change adds support for Qwen3.5+ multimodal models (e.g., qwen3.8-max) by routing them to the MultiModalConversation API, while legacy models continue to use the standard Generation API. It also introduces a configurable base URL (dashscope\_api\_base) to support dedicated Bailian deployments and sets qwen3.7-plus as the default model.
models/dashscope · high confidence
Introduce DingTalk channel with Stream mode, file/image support, and AI card streaming
This change adds the DingTalk channel implementation (channel/dingtalk) to support the DingTalk Stream protocol. Users can now receive and process text, images, files, and rich-text messages from DingTalk, with inbound media safely downloaded and cached. Outbound replies are streamed into DingTalk AI markdown cards for a live-typing experience, with a fallback to webhook markdown when cards are disabled or streaming fails. The implementation includes robust error handling for non-zero API error codes, group-chat whitelist filtering, bounded media downloads, and credential masking in logs.
channel/dingtalk · high confidence
Introduce Godcmd plugin for user and admin command management
A new 'Godcmd' plugin is added to the system, providing a command-line interface for both regular users and administrators. Users can authenticate via the \\#auth\ command using a password configured in \config.json\, after which they gain access to administrative commands such as \\#stop\, \\#resume\, \\#reconf\, and plugin management tools like \\#installp\, \\#uninstallp\, and \\#reloadp\. The plugin also exposes general commands like \\#help\ and \\#reset\ for session management, with help text dynamically generated based on user role and channel type.
plugins/godcmd · high confidence
Introduce Google Gemini bot integration with multimodal support
Added a new Google Gemini bot implementation that enables text-based conversations and supports multimodal inputs via image markers (e.g., \[图片: path\]) embedded in user messages. The bot handles image processing by converting local files to base64 inline data for the Gemini API, manages session state, and includes logic to handle API errors and safety ratings. It defaults to the 'gemini-3.8-flash' model if no specific model is configured.
models/gemini · high confidence
Introduce LinkAI bot integration with improved API configuration handling
Added a new LinkAI bot implementation (models/linkai/link\_ai\_bot.py) that integrates the LinkAI agent platform using an OpenAI-compatible interface. The integration includes robust API base URL normalization to prevent duplicate path segments (e.g., /v1/v1) and converts cryptic 404 errors into actionable configuration hints for users. It supports text and image creation contexts, manages sessions, handles client-specific metadata (client\_id, sender info), and includes specific error handling for authentication failures and rate limits.
models/linkai · high confidence
Introduce LinkAI plugin for Midjourney, document summarization, and knowledge base management
Adds the \linkai\ plugin, enabling integration with the LinkAI platform for three main capabilities: Midjourney image generation (with commands like \$mj\, \$mju\, \$mjv\, \$mjr\ and configurable rate limits), document summarization and multi-turn chat for files (txt, pdf, docx, etc.), sharing links (WeChat articles), and images, and knowledge base management that allows mapping specific group chats to different LinkAI applications via the \$linkai app\ command. The plugin includes a configuration template (\config.json.template\) for enabling/disabling features, setting API keys, and managing permissions, along with administrative commands (\$linkai open\, \$linkai close\, \$linkai sum open\) to control functionality dynamically.
plugins/linkai · high confidence
Introduce MCP (Model Context Protocol) tool integration
Adds support for integrating external MCP servers as agent tools, enabling the agent to discover and execute tools provided by remote or local MCP services. The implementation includes a custom client supporting stdio, SSE, and Streamable HTTP transports, along with OAuth 2.1 authorization for remote servers. Configuration is managed via mcp.json with validation, atomic writes, and BOM-tolerant reading, while security hardening ensures that sensitive environment variables are not leaked to child processes.
agent/tools/mcp · high confidence
Introduce Moonshot (Kimi) bot integration with advanced reasoning and coding support
Added a new MoonshotBot implementation that supports Kimi-specific models including K2.x, K2.5, K2.6, K2.7-code, K3, and coding plans. The integration handles model-specific parameters such as stripping temperature/top\_p for K2.x/coding models, enabling deep reasoning via effort settings for K3/K2 models, and using a specific User-Agent for the Kimi Coding Plan. It also includes session management tailored for Moonshot's token limits and error handling for API failures.
models/moonshot · high confidence
Introduce QQ Bot channel with WebSocket support
Adds a new QQ Bot channel implementation that connects via WebSocket long-connection, supporting group chat (@bot), single chat (C2C), guild channels, and guild DMs. The channel handles text, image, file, and voice messages, including automatic transcription of voice messages via QQ's official ASR and immediate agent execution when transcripts are available. It includes robust connection management with heartbeat keep-alive, auto-reconnect with session resume, and exponential backoff for failed reconnections. Attachment downloads are bounded and filenames are safely normalized to prevent path traversal and excessive length.
channel/qq · high confidence
Introduce Read tool for file inspection with security and parsing improvements
The new Read tool allows agents to inspect file contents, supporting text, images, PDFs, and Office documents (Word, Excel, PowerPoint). It includes security hardening by blocking credential file access and bypasses via /proc/environ or symlinks, and improves usability with features like PDF page range selection, negative offset support for reading from the end of files, and preservation of table structures and formulas in Office files.
agent/tools/read · high confidence
Introduce Telegram channel support
Users can now connect the bot to Telegram via the Bot API. The new channel handles text, photos, voice, video, and documents in both private and group chats, with group triggering via @mention or reply. It exposes the same slash-command menu as the Web channel (help, status, context, tasks, skill, memory, knowledge, config, cancel, steer, logs, version) and supports optional HTTP/SOCKS5 proxies. Markdown output is converted to Telegram-compatible HTML, including code blocks, bold, italic, strikethrough, links, and tables.
channel/telegram · high confidence
Introduce Telegram channel support
Users can now connect the bot to Telegram via the Bot API. The new channel handles text, photos, voice, video, and documents in both private and group chats, with group triggering via @mention or reply. It exposes the same slash-command menu as the Web channel (help, status, context, tasks, skill, memory, knowledge, config, cancel, steer, logs, version) and supports optional HTTP/SOCKS5 proxies. Markdown output is converted to Telegram-compatible HTML, including code blocks, bold, italic, strikethrough, links, and tables.
channel/weixin · high confidence
Introduce WeChat Customer Service (wechat\_kf) channel
Adds a new \wechat\_kf\ channel that enables the agent to handle external WeChat user inquiries via the WeCom Customer Service API. This channel operates independently from the internal \wechatcom\ app, using separate configuration keys (\wechat\kf\\*\) and a dedicated callback port (default 9888). It supports inbound text, image, voice, and file messages (with automatic download and caching), and outbound replies including text (with UTF-8 splitting), images, voice (converted to AMR), and video. The implementation includes a persistent cursor store to prevent message duplication across restarts and uses a background thread pool to handle the WeCom 5-second callback timeout.
_channel/wechat\kf · high confidence
Introduce WeChat Official Account (WeChatMP) channel with voice, image, and encryption support
Users can now connect the application to WeChat Official Accounts (both Subscription and Service accounts) via a new \wechatmp\ channel. This addition enables two-way interaction beyond simple text, supporting voice input (using WeChat's speech-to-text) and voice/image replies. The implementation handles WeChat's strict passive-reply timing constraints by caching responses and merging text segments, while also supporting AES message encryption and local file:// image URLs. It includes a custom client that manages API rate limits by automatically clearing quotas when exceeded, and is compatible with Python 3.13 by replacing the deprecated \imghdr\ module with manual image signature detection.
channel/wechatmp · high confidence
Introduce WeCom App channel with robust token management and media handling
Adds the \wechatcom\_app\ channel, enabling users to connect the bot to a WeCom (WeChat Work) self-built application for receiving text, image, and voice messages. The implementation includes a custom client that proactively refreshes the \access\_token\ in a background thread to prevent expiration errors, splits long text replies into multiple messages, and handles voice conversion and splitting for audio files longer than 60 seconds. It also supports automatic image compression for large uploads and validates the corporate ID on callback endpoints to ensure security. Also in this area: WeChat Com: Fix for temporary file cleanup after long voice replies; Enforce media download size limits and improve error handling in WeChat channels.
channel/wechatcom · high confidence
Introduce common module with multi-agent identity, cloud hand-offs, and localization
The new common package establishes the runtime foundation for multi-agent coordination and cloud integration. It adds a process-wide channel manager registry to fix delivery bugs, a cloud client that enables agent hand-offs across different deployments via a LinkAI console, and a runtime identity system using ContextVars to track agent, user, and session context across threads. The update also introduces a global language resolution module supporting Simplified and Traditional Chinese, a token bucket for rate limiting, and SSL certificate fixes for PyInstaller builds.
common · high confidence
Introduce dedicated DeepSeek bot module with V4 model support
A new independent DeepSeek bot module has been added to handle DeepSeek API interactions, establishing \deepseek-flash\ (V4.1 Flash) as the default model. This module supports the full V4 series (\deepseek-v4-flash\, \deepseek-v4-pro\) and includes native multimodal capabilities for image understanding via \deepseek-flash\ and \deepseek-v4-flash-vision-exp\. It implements explicit thinking mode toggles and reasoning effort controls for V4 models, automatically stripping incompatible parameters like temperature during reasoning, and increases the vision request timeout to 180 seconds to prevent premature failures.
models/deepseek · high confidence
Introduce dedicated chat service with prompt optimization and session management
Adds a new agent/chat module that provides a ChatService for streaming agent responses via the CHAT protocol, a SessionService for managing conversation sessions (including AI-generated titles and context clearing), and a prompt optimization feature that rewrites user inputs into structured instructions using configurable rules from prompts.json. This enables richer, more reliable chat interactions with persistent session handling and improved input clarity.
agent/chat · high confidence
Introduce dedicated web console channel with modular backend and split frontend
The web console is now delivered as a standalone channel (\channel/web\) that can be enabled by setting \channel\_type\ to \web\ in \config.json\ and listening on port 9899. The backend is restructured into a clear URL table (\web\_channel.py\) that wires 77 routes to per-view API modules (e.g., \api/chat.py\, \api/sessions.py\, \api/scheduler.py\) and shared core logic (\core/channel.py\, \core/\_common.py\). The frontend is split from monolithic files into concern-based fragments: server-side assembled HTML templates under \templates/\ (including views for chat, agents, config, skills, memory, knowledge, channels, tasks, and logs) and classic, defer-loaded JavaScript modules under \static/js/\ (with strict load-order rules to avoid temporal dead zone issues). Assets are versioned by file modification time to prevent stale caches, and vendor libraries are vendored locally without external CDN dependencies.
channel/web · high confidence
Introduce desktop renderer with native UI, theming, and artifact management
The desktop application now includes a full React-based renderer that provides a native user interface. This introduces a data-driven theming system with semantic design tokens, supporting both light and dark modes as well as a wallpaper/glass aesthetic. The UI features a new Artifacts page for browsing, pinning, and previewing files generated by agents, alongside a configurable menu editor that allows users to customize the application's navigation. The renderer also handles first-run onboarding, manages authentication for web\_password-protected backends, and integrates with the main process for features like auto-updates, OS notifications, and system-level actions.
desktop/src/renderer/src · high confidence
Introduce desktop-specific React hooks for backend connectivity, platform detection, and notifications
Added five new hooks in the renderer to manage core desktop client behaviors: \useBackend\ handles backend health probing, port resolution, and error diagnosis to prevent false startup failures; \usePlatform\ detects the OS and applies CSS classes for platform-specific styling; \usePushPoll\ and \useSchedulerNotifyPoll\ implement distinct polling loops to receive and display OS notifications for general session updates and scheduled task completions respectively; and \useTheme\ manages a data-driven theme system with light/dark modes, wallpapers, and frosted-glass effects.
desktop/src/renderer/src/hooks · high confidence
Introduce env\_config tool for secure API key management
Added a new \env\_config\ tool that allows users to securely manage API keys and environment variables (such as OPENAI\_API\_KEY, CLAUDE\_API\_KEY, etc.) via set, get, list, and delete actions. The tool stores credentials in a user-specific \\~/.cow/.env\ file, ensuring security by restricting file permissions to owner-only (0o600) and writing updates atomically to prevent corruption. It also supports hot-reloading of environment variables and refreshing associated skills immediately upon configuration changes.
_agent/tools/env\config · high confidence
Introduce granular session permission modes and cross-process agent delegation
Users can now restrict an agent session to read-only or workspace-write modes, preventing unintended file modifications or shell commands outside the working directory, while full-access remains the default. Additionally, agents can now delegate tasks to teammates hosted in separate processes, enabling multi-agent teams where work is distributed across distinct runtime instances.
cowagent · high confidence
Introduce in-process sub-agent tool for parallel task delegation
Users can now delegate self-contained tasks to in-process sub-agents that run in parallel and report back their conclusions. This new \SubagentTool\ allows the main agent to spawn multiple sub-agents simultaneously, each working in its own context, with results formatted for both machine parsing and human readability. The tool handles parallel execution, timeout management, and error reporting, providing a way to offload independent work while maintaining visibility into each sub-agent's progress and outcome.
agent/tools/subagent · high confidence
Introduce knowledge base service with background reindexing
A new KnowledgeService is added to manage the knowledge base, providing a unified interface for listing, reading, and graphing knowledge files. It supports document creation, import, and deletion, while automatically maintaining the index.md file to keep it in sync with the directory tree. The service includes background reindexing capabilities to ensure search indexes are updated without blocking user interactions, and it handles edge cases such as symlink traversal, protected files, and path validation.
agent/knowledge · high confidence
Introduce local file send tool with credential and path safety guards
A new 'send' tool is added to the agent, allowing users to send local files (images, videos, audio, documents) or pass through remote URLs for inline rendering. The tool enforces strict security by refusing credential paths, directories, and special files, and automatically copies files from outside the workspace (e.g., /tmp) into the workspace so they remain accessible in the web console.
agent/tools/send · high confidence
Introduce multi-agent architecture with dedicated team and routing services
This change introduces the core infrastructure for supporting multiple agents within a single runtime, including data isolation and team-based interactions. It adds a new \team.json\ file to store agent profiles and channel bindings separately from the legacy \config.json\, automatically migrating existing single-agent setups. A new \AgentRouter\ handles deterministic message routing, binding channel instances to specific agents and resolving @mentions to the correct agent workspace. Additionally, an \AgentAdminService\ is added to manage agent profiles, enforce core file size limits, and ensure that deleting an agent also cleans up its associated conversations and files.
agent · high confidence
Introduce multi-agent delegation and inbound hand-off support
This change introduces the \agent.multiagent\ package, enabling agents to delegate tasks to teammates hosted in other processes and handle incoming hand-offs. It adds inbound request handling (\inbound.py\) that validates delegation policies, enforces allowlists, checks for cycles, and derives delegation depth from the trace chain. The implementation supports three modes: \delegate\ for sub-tasks, \speak\ for answering turns addressed to a local teammate, and \clear\ for clearing a teammate's context. It also ensures that permission modes are narrowed (not widened) across peer hand-offs and that the caller's permission mode is carried correctly.
agent/multiagent · high confidence
Introduce native desktop shell with auto-updates, themes, and system integration
The desktop application now features a complete native Electron shell that manages the backend lifecycle, system tray, and application menu. Users benefit from automatic background updates via electron-updater (with support for legacy Windows builds and China CDN mirrors), a data-driven theme system allowing bundled and user-provided themes, and native OS integrations including launch-at-login toggles, custom app icons, and system notifications. The shell also provides a generic HTTPS relay to bypass renderer CORS restrictions, robust backend health monitoring with specific error diagnostics, and persistent logging to a local run.log file for troubleshooting.
desktop/src/main · high confidence
Introduce new agent protocol module with streaming execution and robust error handling
A new \agent.protocol\ package has been added, establishing the core execution layer for the agent. This includes \AgentStreamExecutor\ for multi-turn tool-call streaming, \StepWriter\ for persisting run messages incrementally to prevent data loss on crashes, and \message\_utils\ for sanitizing and repairing broken tool-use/tool-result pairs in conversation history. The module also introduces \artifact.py\ to distinguish user-facing workspace files from internal agent bookkeeping, \cancel.py\ for thread-safe cancellation of in-flight runs, and \steer.py\ for active-run steering. Additionally, \agent.py\ now contains a version-gated model specification table (\\_MODEL\_SPECS\) that automatically resolves context windows and output caps for various LLM families (e.g., GPT-5+, DeepSeek V4+, Gemini, Claude, GLM, Qwen), ensuring correct token budgeting.
agent/protocol · high confidence
Introduce plugin architecture with runtime management
The application now supports a plugin system that allows users to extend functionality without modifying core code. Plugins can be installed, enabled, disabled, and reloaded at runtime via the Godcmd admin interface (using commands like \#installp, \#scanp, \#reloadp, and \#updatep). The system supports plugin prioritization to control execution order and provides a global configuration template (plugins/config.json.template) for managing settings. A curated list of available plugins, including tools for AI image generation (Midjourney, SD WebUI), news queries, and task management, is maintained in plugins/source.json.
plugins · high confidence
Introduce portable CLI commands for backup, restore, and skill management
This change adds a new \cli/commands\ package containing portable command-line tools for managing CowAgent data and extensions. The \backup\ module provides commands to create and restore archives of user data, including support for multi-agent workspaces and cross-filesystem operations. The \skill\ module introduces a comprehensive system for installing, updating, and managing skills from various sources such as GitHub, GitLab, SSH, and local paths, with added security hardening for archive extraction. Additionally, new commands are provided for managing the browser tool installation (\install\), viewing knowledge base statistics and trees (\knowledge\), and handling conversation context (\context\).
cli/commands · high confidence
Introduce precise file editing tool with atomic writes and credential protection
Added a new \edit\ tool that allows users to replace exact text within files or append to the end. The implementation ensures file integrity through atomic writes (using a sibling file to prevent data loss on failure) and protects sensitive data by denying edits to credential paths. It also handles cross-platform line endings correctly by detecting them from raw bytes and preserves original formatting during fuzzy matches.
agent/tools/edit · high confidence
Introduce project workspaces and per-session model/permission preferences
Users can now assign a specific working directory to a conversation, keeping file paths, previews, and the \@\ picker anchored to that project rather than the agent's global state root. The system also supports per-session overrides for the AI model, permission mode, and team member rosters, allowing users to pin settings before the first message is sent. These preferences and project bindings are persisted in JSON files under the shared root, with automatic cleanup of stale references when agents or sessions are deleted.
agent/workspace · high confidence
Introduce scheduler tool for managing timed tasks
Adds a new scheduler tool that allows agents to create, manage, and execute scheduled tasks, including static reminders, periodic tasks via cron expressions or fixed intervals, and dynamic tool calls. The tool supports listing, retrieving, enabling, disabling, and deleting tasks, with persistent storage in a JSON file. It includes a background service for task execution, a recipient store for managing trusted delivery targets across channels, and utilities for handling timezones and cron recurrence.
agent/tools/scheduler · high confidence
Introduce search\_files tool for workspace content and filename search
A new \search\_files\ tool is added to the agent's toolset, allowing users to search file contents or locate files by name. The tool implements a four-tier backend strategy (ripgrep, grep, PowerShell Select-String, and pure Python) to ensure fast and reliable searches across different platforms, while respecting \.gitignore\ rules and excluding common VCS/dependency directories.
_agent/tools/search\files · high confidence
Introduce self-evolution subsystem for autonomous agent learning
A new background subsystem automatically reviews idle conversations to autonomously learn durable patterns, create or patch skills, and update memory or knowledge files. The feature is enabled by default, triggering after 6 turns or 10 minutes of idle time (or when context pressure is high), and uses a restricted, isolated review agent to make conservative changes. It includes safeguards such as file backups for instant rollback, workspace-level concurrency locks, and protection of built-in skills, while logging all changes to a dedicated evolution timeline for user visibility.
agent/evolution · high confidence
Introduce slash-command and 'cow' prefix interception with fuzzy resolution and custom aliases
The new cow\_cli plugin intercepts chat messages to handle slash commands (e.g., /skill list) and 'cow' prefixed commands (e.g., cow skill list). It supports fuzzy command resolution to suggest corrections for typos, handles ambiguous inputs by prompting the user, and allows users to define custom command aliases via the config file. This enables a more robust and user-friendly command interface within the chat environment.
_plugins/cow\cli · high confidence
Introduce structured long-term memory and conversation persistence
The agent now includes a dedicated memory module that persists conversation history in a global SQLite database (index.db) scoped by agent, and manages long-term memory via a hybrid vector/keyword search backend. This adds a pluggable reranker for precision reordering, a Deep Dream distillation process that periodically summarizes daily records into a refined MEMORY.md and dream diary, and a MemoryService to list and read memory/dream/evolution files. Configuration is now per-workspace rather than per-process, and the system supports multi-vendor embedding fallback and async LLM context summary injection on trim.
agent/memory · high confidence
Introduce the Skills module for agent extensibility
The agent/skills directory now provides a complete framework for discovering, loading, and managing modular skills. This includes a loader that parses SKILL.md files with YAML frontmatter, a manager that handles skill configuration and selection, and a service layer that supports creating, installing, and deleting skills. The implementation includes robust security measures for handling untrusted input, such as validating archive contents to prevent path traversal and limiting file counts and sizes during extraction.
agent/skills · high confidence
Introduce tool plugin for external capabilities
The new tool plugin enables the chatbot to access external resources and perform actions beyond standard text generation. It supports a variety of tools including Python execution, web browsing (url-get, browser), terminal command execution, weather queries (meteo), and multiple search engines (Bing, Google, SearXNG, Wikipedia, Arxiv). Additional capabilities include news aggregation, image-to-text analysis (visual), text-to-speech, speech-to-text, and messaging via email, SMS, and WeChat. Users can configure specific tools and API keys via a config file, with a template provided for setup.
plugins/tool · high confidence
Introduce unified tool registry with new search, delegation, and evolution capabilities
The agent now uses a centralized \ToolManager\ and a structured \agent/tools\ package to manage its capabilities. This introduces several new tools: \SearchFiles\ for searching file contents and names with a multi-backend strategy (ripgrep, grep, PowerShell, Python), \AgentDelegateTool\ and \SubagentTool\ for cross-agent and in-process task delegation, and \EvolutionUndoTool\ to roll back self-evolution changes. The system also integrates MCP (Model Context Protocol) tools via \McpTool\ and \McpClientRegistry\, and loads optional tools like \WebSearch\, \WebFetch\, \Vision\, \BrowserTool\, \EnvConfig\, and \SchedulerTool\ only when their dependencies are available. Base tool infrastructure includes \BaseTool\ with support for progress reporting, cancellation, and parallel execution flags.
agent/tools · high confidence
Introduces configurable product extension points for custom builds
The desktop renderer now supports a pluggable extension system that allows custom builds to override default behaviors and UI elements without modifying core code. This is achieved via a \ProductExtension\ interface and a default empty implementation, enabling alternate builds to inject custom authentication gates, UI slots (such as nav rail branding, assistant avatars, and home/status logos), additional routes, onboarding controls, model selection pickers, navigation link visibility, and chat suggestions. This change provides the infrastructure for OEMs or specific product builds to tailor the application's appearance and functionality while keeping the core codebase unchanged.
desktop/src/renderer/src/product · high confidence
Introduces new Zustand-based state management for the desktop client
The desktop application now uses a dedicated set of Zustand stores (agent, artifact, chat, session, workspace, etc.) to manage UI and application state, replacing the previous approach. This change introduces multi-agent support by tracking agent rosters and session ownership, adds an artifact center for managing generated files, and implements features like chat history persistence, draft recovery, and task completion notifications.
desktop/src/renderer/src/store · high confidence
Introduction of per-session permission modes for agent tool access
The agent now supports configurable permission modes (read-only, workspace-write, and full-access) that bound what tools are allowed to do during a session. This change introduces the core permission module, exposing constants for the available modes, a decision type, and functions to check tool calls, describe modes, and manage the global or normalized mode state, allowing users to control the scope of agent actions.
agent/permission · high confidence
New Feishu channel with dual event modes and rich media support
This change introduces a new Feishu channel implementation that supports both webhook and WebSocket event reception modes, configurable via \feishu\_event\_mode\. It adds support for receiving and processing rich message types including images, files, audio (voice), and rich-text (post) messages, with automatic downloading to the workspace temporary directory. The channel features streaming progress cards that display agent reasoning, tool execution steps, and accumulated text responses in real-time, as well as scheduler action cards for managing scheduled tasks directly within Feishu. It includes a one-click QR-scan app creation flow for simplified setup, lazy loading of the \lark\_oapi\ SDK to reduce startup time, and on-demand SDK bundle provisioning for desktop builds. The implementation also handles message recall events, group chat mention gating, and secure credential management.
channel/feishu · high confidence
New Role plugin for AI persona switching
A new 'Role' plugin has been added, allowing users to switch the bot's behavior to specific personas (e.g., 'Writing Assistant', 'Catgirl', 'Buddha') via commands like \$role\ or \$角色\. The plugin includes a built-in library of roles defined in \roles.json\ and supports fuzzy matching for role names. Users can also extend the library by placing custom JSON files in a \roles/\ directory, where files with matching titles override built-in roles and new titles are appended. The plugin supports filtering roles by tags and works with a wide range of supported AI backends.
plugins/role · high confidence
New Windows installer and Linux service management scripts
The repository now includes a comprehensive PowerShell installer script (run.ps1) for Windows that handles one-line installation, service management (start/stop/restart), and configuration. This script features internationalization (i18n) for Chinese and English UI, fixes for Windows-specific issues like console QuickEdit hangs and UTF-8 encoding, and simplified model selection menus. Additionally, new shell scripts (start.sh, shutdown.sh, tout.sh) are provided for Linux users to easily start, stop, and monitor the application's background process and logs.
scripts · high confidence
New ZhipuAI integration with vision and image generation support
This change introduces a new ZhipuAI bot implementation (models/zhipuai) that adds support for ZhipuAI's GLM models, including multimodal vision capabilities and image generation. The integration allows users to configure a custom API base URL (e.g., for the international z.ai endpoint) via the \zhipu\_ai\_api\_base\ configuration. It features intelligent vision handling: if the main model is vision-capable (like glm-5.3-flash), it is used directly for image understanding; otherwise, it falls back to the dedicated glm-5v-turbo model. Additionally, the bot supports image creation via the CogView-3 model and includes session management with token-based message truncation to handle context limits.
models/zhipuai · high confidence
New browser tool for web navigation and interaction
Added a new browser tool that allows the agent to control a Chromium browser for web navigation, element interaction, and content extraction. The tool supports multiple launch modes, including using the user's system-installed Chrome or Edge via CDP (to preserve logins and avoid automation prompts) or using Playwright's bundled Chromium. It includes SSRF protection by blocking link-local and cloud-metadata addresses, persists login state across sessions via a dedicated user profile directory, and optimizes resource usage with memory caps and automatic shutdown on inactivity.
agent/tools/browser · high confidence
New desktop console pages for agents, knowledge, memory, channels, and artifacts
The desktop client now includes a full set of console pages mirroring the web interface, allowing users to manage their agent roster, browse and edit memory files, organize knowledge bases, configure communication channels, and view conversation artifacts. These pages introduce multi-agent scoping, tabbed editors for agent profiles and skills, live log streaming with download, and a dedicated artifacts viewer with search and pinning, all integrated into the desktop navigation.
desktop/src/renderer/src/pages · high confidence
New desktop navigation rail, session list, and window controls
The desktop app now features a redesigned layout with a new navigation rail (NavRail) that includes a brand logo on Windows/Linux, theme and language toggles, and an update status indicator. A new session list component (SessionList) organizes conversations by project or time, supports pinning, renaming, deleting, and reordering sessions, and persists collapsed project states. Custom window controls (WindowControls) are rendered for the frameless titlebar on Windows, providing minimize, maximize, and close buttons while remaining hidden on macOS.
desktop/src/renderer/src/layout · high confidence
New desktop workspace, menu, and notification utilities
This change introduces several new utility modules in the desktop renderer to support workspace file handling, customizable navigation menus, and task completion notifications. The \fileKind\ module defines file types, previewable/editable sets, and icon/color mappings, while also providing logic to route workspace-relative links to the preview panel and parse attachment markers from chat history. The \menu\ module manages the navigation rail, supporting both built-in and user-customized menu structures shared with the web console. The \taskNotify\ module handles OS-level notifications for completed or failed agent runs, respecting user settings for sound and visibility. Additionally, \handoff\ logic identifies agent delegation steps, \newChat\ centralizes session creation logic, and \permission\ defines access modes for workspace interactions.
desktop/src/renderer/src/lib · high confidence
New keyword-matching plugin with file and image support
A new 'Keyword' plugin has been added that matches user text input against a configurable list of keywords. When a match is found, the plugin replies with the configured content, supporting plain text, image URLs (including WebP), and video URLs. For downloadable files (PDF, Office docs, archives), it fetches the content to a temporary directory and sends the file to the user, with bounded timeouts and error reporting. Configuration is managed via a per-plugin config.json file.
plugins/keyword · high confidence
New memory search and retrieval tools for agents
The agent now includes dedicated tools for accessing long-term memory and knowledge. The \memory\_search\ tool allows agents to find relevant information using semantic and keyword queries, while the \memory\_get\ tool enables reading specific content from memory or knowledge files with line-range support. These tools handle cross-platform path resolution and validate file access to ensure agents only read files within their allowed workspace or shared knowledge roots.
agent/tools/memory · high confidence
New modal dialogs for web console configuration and management
The web console now includes a suite of new modal interfaces for managing skills, MCP tools, and workspace settings. Users can add skills from the Cow Skill Hub, GitHub, or ClawHub, or create them via a form with file attachments. MCP tools can be configured using a form or raw JSON, with support for testing connections and advanced settings like timeouts and prefixes. Additional dialogs allow users to customize the main menu, rename channel instances, manage knowledge documents, pick project folders, review execution run details, and edit scheduled tasks with cron or interval scheduling. Also in this area: 6 commits (1 fix) modifying channel.
channel/web/templates/modals · medium confidence · unverified
New shared utility library for agent tools
A new \agent/tools/utils\ package introduces shared infrastructure for file operations, security, and output handling. It adds credential-path guards (\credentials.py\) to prevent tools from bypassing the \env\config\ tool by reading secrets from \/proc/\/environ\ or symlinks to \\~/.cow/.env\. SSRF protection (\url\_safety.py\) is introduced as an opt-in feature (disabled by default) to block requests to private, loopback, or link-local addresses, including re-validation of redirect hops. File editing reliability is improved via \diff.py\ (unified fuzzy matching that preserves indentation and line endings) and \file\_state.py\ (conversation-scoped staleness warnings when files change on disk). Structured data writes are now validated for syntax (\syntax\_check.py\), and tool outputs are consistently truncated (\truncate.py\).
agent/tools/utils · high confidence
New skills management UI with MCP configuration and multi-source installation
The Skills page now features a comprehensive modal-based interface for managing capabilities. Users can install skills from multiple sources—including a product catalogue, GitHub, and external hubs—via a new SkillAddModal that supports previewing and uploading. A new SkillCreateForm allows users to define and create custom skills directly. Additionally, a new McpEditorModal enables users to view, edit, and test MCP server configurations through a form or raw JSON interface, with built-in validation and testing capabilities.
desktop/src/renderer/src/pages/skills · high confidence
New web console layout with sidebar, header, and session panel
The web console now uses a new layout structure consisting of a collapsible sidebar, a top header bar, and a session history panel. The sidebar provides navigation to Chat, Agents, Configuration, Skills, Memory, Knowledge, Channels, Tasks, and Logs, with collapsible groups that persist user preferences. The header includes controls for toggling the sidebar, session panel, artifacts view, workspace panel, and message navigator, along with language selection, theme toggle, and links to documentation, the website, and GitHub. The session panel allows users to view and switch between conversation history and start new chats. Also in this area: 6 commits (1 fix) modifying channel.
channel/web/templates/layout · medium confidence · unverified
New web console views for agents, artifacts, channels, and configuration
The web interface now includes dedicated views for managing agent teams (with creation, profile, skills, and core file editing), browsing conversation artifacts with filtering and preview, managing message channels, and configuring model providers and agent parameters (tokens, turns, thinking). These templates establish the structure for the new console capabilities.
channel/web/templates/views · high confidence
New workspace panel and in-place file editing capabilities
The web interface now includes a dedicated workspace panel that allows users to browse project directories, preview various file types (code, markdown, images, etc.), and edit text-based files directly in the browser. This change introduces new JavaScript modules (boot.js, doc-editor.js, workspace.js) to handle file management, inline editing with conflict detection, and session-scoped workspace access. Users can now view and modify memory files and skill definitions, and agent-generated file links will route to this new preview panel for immediate inspection and editing.
channel/web/static/js · high confidence
New write tool for creating and overwriting files
A new 'write' tool has been added to the agent's toolset, allowing it to create new files or completely overwrite existing ones. The tool automatically creates parent directories as needed and includes safety checks to prevent writing to credential files (such as \~/.cow/.env) and to reject content that resembles line-numbered output from the read tool. It also supports optional workspace confinement via configuration and integrates with the memory manager to mark the memory index as dirty when writing to memory-related files.
agent/tools/write · high confidence
Redesigned desktop settings with per-provider model catalogs and chat fallback chains
The desktop settings interface has been rebuilt into a tabbed layout (Basic and Models) featuring a new Models tab that organizes capabilities (chat, vision, image, etc.) into individual cards. Users can now configure per-provider model catalogs with an overlay editor that tracks changes against vendor presets, and set up an ordered fallback chain for chat to automatically switch models if the primary fails. The Basic settings tab now includes a launch-at-login toggle for macOS and Windows, live-editable context budget controls, and a global permission mode setting.
desktop/src/renderer/src/pages/settings · high confidence
Support for multiple concurrent channel instances
The system now allows running multiple instances of the same channel type (e.g., multiple Feishu or DingTalk bots) simultaneously, each bound to a specific AI agent. This is configured via the new \channel\_instances\ list in \team.json\, which supports per-instance credentials, agent routing, and team delegation. The \channel\_factory\ and \channel\_instances\ modules handle the resolution of these instances, while the base \Channel\ class now supports instance-specific identity and credential overrides. Additionally, Discord and Slack channels have been added to the supported list, and the \ChatChannel\ base class has been refactored to use instance-level session queues to prevent cross-channel message interference.
channel · high confidence
Terminal channel now renders agent streaming events with rich formatting
The terminal channel has been updated to support real-time rendering of agent stream events, mirroring the experience of the web UI. Users will now see styled output for reasoning steps, tool calls (including arguments and execution results), file attachments, and errors, with ANSI color codes applied when running in a TTY environment. This change introduces the \TerminalAgentRenderer\ class to handle event types like \reasoning\_update\, \tool\_execution\_start/end\, and \message\_update\, providing immediate visual feedback during agent operations directly in the terminal.
channel/terminal · high confidence
Unified embedding provider factory with multi-vendor support and robustness fixes
The embedding subsystem now uses a shared factory to consistently initialize the configured embedding provider across all components (agent initialization, knowledge sync, index rebuild), preventing silent degradation to keyword-only search. It supports multiple vendors (OpenAI, LinkAI, DashScope, Doubao, Zhipu, and custom endpoints) with automatic fallback to LinkAI if OpenAI fails in the default path, and handles blank text inputs by substituting a placeholder to avoid API errors. Additionally, it tracks the chunker version to suggest index rebuilds when stale, cleans up legacy state files, and ensures batch results are ordered correctly.
agent/memory/embedding · high confidence
Web chat interface modularized and enhanced with drag-and-drop, context usage, and voice support
The web chat interface has been refactored by splitting the monolithic console.js into focused, per-concern scripts (composer-input, context-usage, message-actions, new-chat, render, scheduler-notify, send, session-settings) to improve maintainability and performance. Users now benefit from a drag-and-drop file upload overlay in the chat view, a hover-activated context usage popover showing a donut chart of token consumption (system, tools, history, free) with actions to clear, compact, or adjust budget, and support for sending voice messages with ASR captions. The chat also features a slash command menu for quick access to skills, knowledge, and configuration, lazy-loading of folded tool-call step bodies to prevent UI freezing on long sessions, and cross-session notifications for scheduled tasks. Additionally, the new chat flow now supports multi-agent team chats, and message editing/deletion cascades properly to subsequent turns.
channel/web/static/js/chat · high confidence
Web search is now a built-in tool with multi-provider support
The web search capability is now a built-in tool that supports nine backends (Bocha, Qianfan, Zhipu, LinkAI, AnySearch, Serply, Tavily, SearXNG, and Keenable) with a unified response format. Users can configure specific providers via API keys or environment variables, or enable keyless modes for AnySearch and Keenable through explicit opt-in flags. The tool automatically selects the best available provider based on a configured priority order or allows users to fix a specific provider, ensuring robust fallback behavior if credentials are missing. Also in this area: 3 commits (1 fix) modifying agent; Browser tool stability, search reliability, and OAuth configuration fixes; Fix crash on blank or non-string web search config values; Bash tool reports written files; MCP client fixes orphaned processes and timeouts; Send tool blocks credential paths; WebSearch handles non-string config.
_agent/tools/web\search · medium confidence · unverified
Removals
Removal of Baidu Unit Bot factory
The bot factory implementation for the Baidu Unit service has been removed. The \bot\_factory.py\ file, which previously handled the creation of \BaiduUnitBot\ instances, is deleted, meaning the application no longer supports or instantiates the Baidu Unit bot channel.
bot · high confidence
Removal of legacy WeChat (itchat) channel implementation
The legacy WeChat channel implementation based on the itchat library has been removed from the codebase. This file previously handled message reception and sending for WeChat Personal accounts via QR code login; its removal indicates that this specific integration is no longer supported or has been replaced by other channel implementations.
channel/wechat · high confidence
Removed Baidu Unit bot integration
The Baidu Unit bot implementation has been removed from the codebase, eliminating the ability to interact with the Baidu Unit API for chat responses.
bot/baidu · high confidence
Architecture
Centralize web console API handlers into channel/web/api
The web console's endpoint logic has been reorganized into a dedicated \channel/web/api/\ directory, introducing new modules for agents, artifacts, authentication, channels, chat, configuration, files, knowledge, and logs. This refactoring consolidates the URL routing and handler definitions, moving them out of the previous monolithic structure to improve code organization and maintainability for the web interface.
channel/web/api · high confidence
Web console JavaScript refactored into modular core components
The monolithic console.js has been split into distinct, load-ordered modules (auth, confirm, i18n, markdown, menu, nav, notify, router) to improve maintainability and separation of concerns. This refactoring introduces a dedicated authentication gate that prevents background pollers from running before login, implements address-bar routing for views and tabs, and adds a custom confirm dialog system. It also includes a comprehensive internationalization table, a markdown renderer with CJK emphasis and mermaid support, and a customizable sidebar menu system.
channel/web/static/js/core · high confidence
Web console split into modular view scripts
The monolithic console.js has been refactored into a modular script tree under channel/web/static/js/views/, introducing dedicated modules for agents, artifacts, channels (including Feishu, WeCom, and WeChat), and configuration. This change improves maintainability and load performance by separating concerns, while preserving the existing global-scope execution model required by the application's classic script architecture.
channel/web/static/js/views · high confidence
Web console styles split into per-area modules
The monolithic console.css has been refactored into eight distinct, load-bearing stylesheets (agents, artifacts, base, chat, components, knowledge, markdown, menu, sessions, workspace) to improve maintainability and organization. This structural change ensures that styles for specific console areas—such as the agent roster, artifact gallery, chat composer, and sidebar—are now isolated in their own files while preserving the existing visual design and layout.
channel/web/static/css · high confidence
Behavioural changes
14 commits (7 fixes) modifying channel/slack
A change to existing behaviour in channel/slack — 14 commits (7 fixs), 3 files: “fix(slack): keep same-named inbound files separate”; “feat(channels): run multiple channel instances, each bound to its own agent”; “feat(channel): support slack bot”; and 11 more.
channel/slack · medium confidence · unverified
2 commits (0 fixes) modifying bot/\_\_pycache\_\_, channel/\_\_pycache\_\_
A change to existing behaviour in bot/\_\pycache\\, channel/\\pycache\\_ — 2 commits, 4 files.
bot/\\pycache\\, channel/\\pycache\\, desktop/resources · medium confidence · unverified
2 commits (1 fix) modifying skills/skill-creator/scripts
A change to existing behaviour in skills/skill-creator/scripts — 2 commits (1 fix), 3 files: “fix(skill-creator): read and write skill files as UTF-8 so init\_skill.py and quick\_validate.py work on Windows”; “feat: support skills creator and gemini models”.
skills/skill-creator/scripts · medium confidence · unverified
3 commits (0 fixes) modifying channel/web/templates/guest
A change to existing behaviour in channel/web/templates/guest — 3 commits, 1 file: “feat: polish guest embed launcher and default to light theme”; “feat: add guest page language and loading states”; “feat: serve guest pages at /share/\<token\> with an embeddable widget”.
channel/web/templates/guest · medium confidence · unverified
3 commits (1 fix) modifying agent/prompt
A change to existing behaviour in agent/prompt — 3 commits (1 fix), 2 files: “feat: add guest share core for public Agent pages”; “fix: forbid guest agent from disclosing host secrets and infrastructure”; “feat: support skills creator and gemini models”.
agent/prompt · medium confidence · unverified
4 commits (0 fixes) modifying channel/web/static/guest
A change to existing behaviour in channel/web/static/guest — 4 commits, 4 files: “feat: polish guest embed launcher and default to light theme”; “feat: add data-open embed option and link the live demo”; “feat: add guest page language and loading states”; and 1 more.
channel/web/static/guest · medium confidence · unverified
5 commits (0 fixes) modifying agent/guest
A change to existing behaviour in agent/guest — 5 commits, 7 files: “feat: answer guest pages over the control connection”; “feat: report today's guest questions and tokens in the get action”; “refactor: drop the sign-out-visitors action”; and 2 more.
agent/guest · medium confidence · unverified
5 commits (0 fixes) modifying desktop/src/renderer/src/pages/settings/shares
A change to existing behaviour in desktop/src/renderer/src/pages/settings/shares — 5 commits, 4 files: “refactor: drop the sign-out-visitors action”; “feat(desktop): integrations tab with guest pages and listen address”; “feat: modify home links by language”; and 2 more.
desktop/src/renderer/src/pages/settings/shares · medium confidence · unverified
8 commits (4 fixes) modifying models/mimo
A change to existing behaviour in models/mimo — 8 commits (4 fixs), 3 files: “fix(minimax): close streaming responses on every exit”; “fix(vision): bump vision timeout from 60s to 180s to avoid premature failures”; “feat: add MiMo speech recognition and migrate chat models to V2.6”; and 5 more.
models/mimo · medium confidence · unverified
Introduce Agent Bridge for multi-agent conversations and model fallback
The bridge layer now supports an Agent-based conversation mode alongside the existing single-bot routing. A new AgentBridge subsystem (agent\_bridge.py, agent\_initializer.py, agent\_event\_handler.py) handles agent initialization, workspace isolation, memory sync, and event streaming, including a cap on intermediate thinking messages for WeChat channels. The core Bridge class (bridge.py) has been refactored to route chat requests to either the legacy bot factory or the agent bridge, and now implements a configurable chat fallback chain that walks through multiple provider/model links when the primary model fails. Additionally, bot\_type resolution was hardened to prevent AttributeErrors when model names are parsed as non-strings, and voice-to-text providers are auto-selected based on available API keys.
bridge · high confidence
New Docker deployment with persistent data and browser support
The Docker deployment has been updated to use a new base image (python:3.10-slim-bullseye) and a new Dockerfile (Dockerfile.latest) that includes optional browser installation via Playwright for enhanced functionality. Configuration is now managed via environment variables in docker-compose.yml, with a default model set to deepseek-flash and web host bound to 127.0.0.1. Data persistence is improved by separating workspace (cow) and configuration data (cow-data) into distinct volumes, ensuring config.json and credentials survive container upgrades. The entrypoint script handles timezone configuration and ensures proper file ownership for the non-root agent user.
docker · high confidence
New desktop API client with multi-agent scoping and local auth
The desktop renderer now uses a dedicated API client (client.ts) that authenticates via a Bearer token in localStorage to bypass cross-origin cookie limitations when running from a file:// origin. The client automatically scopes requests to a specific agent by appending an agent\_id query parameter or injecting it into JSON bodies, supporting both a global active agent and per-request overrides. It also includes a specialized multipart upload method with retry logic to handle intermittent connection resets from the local backend.
desktop/src/renderer/src/api · high confidence
Offline support via vendored frontend assets
The Web Console now bundles third-party frontend assets (such as Font Awesome, Inter font, Tailwind CSS, Markdown-it, Highlight.js, and D3) locally, enabling the console to function in fully offline or air-gapped environments without relying on external CDNs. Mermaid.js remains loaded from a CDN at runtime with integrity checks, so diagrams will not render in offline mode but will fall back to plain code blocks.
channel/web/static/vendor · high confidence
OpenAI integration refactored to native HTTP client with Responses API support
The OpenAI model implementation has been refactored to remove the dependency on the official \openai\ SDK, replacing it with a custom lightweight HTTP client (\OpenAIHTTPClient\) that handles chat completions, image generation, and streaming. This change introduces support for the OpenAI Responses API (configurable via \open\_ai\_api\_type\), enabling tool calling for models like \gpt-6-astra\ that require it. The update also adds automatic attribution headers for specific gateways (OpenRouter, Vercel AI Gateway, OrcaRouter) and ensures backward compatibility through a new exception mapping layer (\openai\_compat\) that preserves existing error handling behaviors.
models/openai · high confidence
Unified multi-provider image generation with auto-fallback
The image generation skill now supports a unified script that automatically attempts image generation across multiple providers (OpenAI, Gemini, Seedream, Qwen, MiniMax, and LinkAI) in a defined priority order. It handles specific model families like gpt-image-2.5-flare, gpt-image-2, and various Gemini/Seedream models, while also implementing robust error handling by skipping empty response fields and falling back to alternative providers or formats when one fails. Also in this area: 2 commits (1 fix) modifying skills; Fix image format preservation and JSON parsing in image generation; Fixes to error handling and channel management logic; Fix crash when image-generation skill receives non-object JSON arguments.
skills/image-generation/scripts · medium confidence · unverified
Unified voice processing and multi-provider TTS/ASR support
The voice module now uses a centralized factory to instantiate a wide range of text-to-speech and speech-to-text providers (including Baidu, Google, OpenAI, Azure, ElevenLabs, Ali, Edge, Xunfei, Tencent, MiniMax, DashScope, Zhipu, Mimo, and custom OpenAI-compatible endpoints) via a single abstract interface. Audio handling is standardized through a new conversion utility that supports decoding Silk files, converting between formats (WAV, MP3, AMR), and splitting long audio segments, ensuring consistent behavior across different voice services.
voice · high confidence
Web console core refactored with atomic config saves and HTTPS support
The web console's backend logic has been reorganized into a dedicated core module (\_common.py, channel.py, providers.py, template.py, tls.py) to decouple the channel lifecycle from request handlers. This change introduces atomic file writes for config.json, ensuring that a failed save operation no longer corrupts or wipes the configuration file. It also adds support for direct HTTPS access to the web console via an opt-in TLS configuration, and updates the provider catalogue to include new model variants (such as GPT-6.1-sol, GPT-6-luna, and Claude Opus 5.5) while refining the server-side HTML assembly for better asset versioning.
channel/web/core · high confidence
Fixes
32 commits (22 fixes) fixing agent/tools/web\_fetch
A fix in agent/tools/web\_fetch — 32 commits (22 fixs), 2 files: “fix(web\_fetch): remove the partial download when the transfer fails mid-stream”; “fix(web\_fetch): preserve readable HTML page text”; “fix(web\_fetch): bound the webpage response by the same size limit as documents”; and 29 more.
_agent/tools/web\fetch · medium confidence · unverified
Agent Mail overlay becomes fail-safe with optional Node.js and CLI installation
The Agent Mail integration now installs Node.js (v22.23.3) and the @tencent-qqmail/agently-cli (v1.0.18) at runtime if they are not already present, while ensuring that any failure in this process does not prevent the main CowAgent service from starting. This change makes the Agent Mail overlay fail-safe by allowing the agent to run without the mail features if the dependencies cannot be installed, rather than blocking startup entirely.
docker/agent-mail · high confidence
ChatGPT bot migration to OpenAI-compatible layer with custom provider support
The ChatGPT bot implementation has been rewritten to use the shared OpenAI-compatible HTTP client and error-handling layer, replacing the previous direct SDK usage. This change ensures consistent error handling and allows the bot to correctly resolve API credentials from specific custom providers (identified by \custom:\<id\>\) rather than relying on global configuration, preventing 404 errors when routing to non-OpenAI endpoints. It also adds support for new model variants (including GPT-5 series and o1) with appropriate parameter adjustments and improves session token counting accuracy.
models/chatgpt · high confidence
Desktop build system overhaul and Windows shortcut fix
The desktop build process has been restructured to produce a self-contained backend bundle using PyInstaller, which now explicitly includes CLI command modules, document parsing libraries (PDF, Word, Excel, PPT), and the web console assets. To keep the initial download size small, the heavy Feishu SDK is excluded from the bundle and instead fetched on-demand as a trimmed, pure-Python package. The build also introduces macOS signing and notarization with device entitlements for microphone and camera access, and fixes a Windows issue where application updates would break desktop shortcuts by replacing the installer's staging move with an in-place removal and repair logic.
desktop/build · high confidence
Test coverage
Added development tools to validate console script load order and router behavior; Added tests for chat scroll-follow logic; Test suite isolation and stability improvements.
Dependencies
Introduces CowAgent Desktop client and standardizes Python dependency constraints
This change introduces the \cowagent-desktop\ application (v2.1.9), a new Electron-based desktop client built with React, Vite, and TypeScript, featuring built-in auto-updates via \electron-updater\ and native macOS signing/notarization. It bundles a slimmed-down Python backend (\requirements-desktop.txt\) that supports Python 3.13+ by conditionally installing \web.py\ from GitHub and \legacy-cgi\, while also adding document parsing libraries (PDF, Word, Excel, PPT) and browser automation via Playwright. The main Python project dependencies (\requirements.txt\, \pyproject.toml\) are standardized to enforce Python 3.7+ compatibility, relax \numpy\ to \>=1.21, and pin specific versions for AI SDKs (e.g., \zai-sdk\>=0.2.3\ for GLM-5.3) and IM channels.
(dependencies) · high confidence
Updated axios HTTP client library
The static axios library has been updated to version 1.1.2, replacing the previous version in the web channel's static assets.
channel/web/static · high confidence
Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.
How this codebase got here
Score
- CAI 37 → 39 (+1.2)
- Rubric changed (rubric-2026.10.5 → rubric-2026.11.4) — scores are not directly comparable.
Lenses
- Code Health 36 → 35 (-0.9)
- Architecture 85 → 84 (-1.1)
- Maturity 72 → 77 (+4.9)
- Readiness 31 → 32 (+0.8)
- Security 54 → 64 (+10.5)
- Accessibility 36 → 38 (+2.4)
- Performance 60 → 60 (+0.0)
Resolved (80)
- Change coupling: deepseek_bot.py ↔ moonshot_bot.py (models/deepseek/deepseek_bot.py)
- Change coupling: index.ts ↔ types.ts (desktop/src/main/index.ts)
- Change coupling: preload.ts ↔ types.ts (desktop/src/main/preload.ts)
- Change coupling: slack_channel.py ↔ telegram_channel.py (channel/slack/slack_channel.py)
- Change coupling: web_channel.py ↔ client.ts (channel/web/web_channel.py)
- Change-coupling hub: doubao_bot.py → deepseek_bot.py, minimax_bot.py, moonshot_bot.py (models/doubao/doubao_bot.py)
- ChromeLauncher._profile_chrome_processes (cognitive 16) (agent/tools/browser/chrome_launcher.py)
- Duplicated block (10 lines × 2) (app.py)
- Duplicated block (10 lines × 2) (voice/custom/custom_voice.py)
- Duplicated block (10 lines × 2) (voice/custom/custom_voice.py)
- Duplicated block (15 lines × 2) (agent/tools/browser/browser_tool.py)
- Duplicated block (15 lines × 4) (models/deepseek/deepseek_bot.py)
- Duplicated block (15–17 lines × 2) (channel/wechat_kf/wechat_kf_channel.py)
- Duplicated block (15–19 lines × 2) (voice/openai/openai_voice.py)
- Duplicated block (16–17 lines × 2) (bridge/agent_bridge.py)
- Duplicated block (16–17 lines × 3) (models/doubao/doubao_bot.py)
- Duplicated block (17 lines × 2) (agent/memory/embedding/provider.py)
- Duplicated block (17 lines × 2) (voice/dashscope/dashscope_voice.py)
- Duplicated block (17–19 lines × 2) (models/minimax/minimax_bot.py)
- Duplicated block (18 lines × 3) (channel/discord/discord_channel.py)
- …and 60 more
New (274)
- Boundary-crossing change coupling: const.py ↔ reasoning_capabilities.py (common/const.py)
- Boundary-crossing change coupling: models.js ↔ i18n.ts (channel/web/static/js/views/models.js)
- Change coupling: discord_channel.py ↔ slack_channel.py (channel/discord/discord_channel.py)
- Change coupling: doubao_bot.py ↔ minimax_bot.py (models/doubao/doubao_bot.py)
- Change coupling: doubao_bot.py ↔ moonshot_bot.py (models/doubao/doubao_bot.py)
- Change coupling: i18n.js ↔ models.js (channel/web/static/js/core/i18n.js)
- Change-coupling hub: types.ts → index.ts, preload.ts, primitives.tsx (desktop/src/renderer/src/types.ts)
- ChatGPTBot.reply_image (cognitive 17) (models/chatgpt/chat_gpt_bot.py)
- ChatGPTBot.reply_image (cyclomatic 17) (models/chatgpt/chat_gpt_bot.py)
- ClassTooLong: ChatPage (desktop/src/renderer/src/pages/ChatPage.tsx)
- ClaudeAPIBot._sanitize_message (cognitive 24) (models/claudeapi/claude_api_bot.py)
- ClaudeAPIBot._sanitize_message (cyclomatic 16) (models/claudeapi/claude_api_bot.py)
- Duplicated block (12–18 lines × 2) (channel/feishu/feishu_channel.py)
- Duplicated block (14–15 lines × 2) (models/doubao/doubao_bot.py)
- Duplicated block (14–20 lines × 8) (models/deepseek/deepseek_bot.py)
- Duplicated block (15–17 lines × 2) (agent/tools/web_search/web_search.py)
- Duplicated block (19–26 lines × 5) (models/deepseek/deepseek_bot.py)
- Duplicated block (21–26 lines × 7) (models/deepseek/deepseek_bot.py)
- Duplicated block (32–34 lines × 3) (models/deepseek/deepseek_bot.py)
- Duplicated block (5 lines × 2) (agent/guest/actions.py)
- …and 254 more
Changes since last survey
- 242 commits — 77 feature/other, 165 fixes
By area
- (repo) — 81 commits
- channel/web — 32 commits
- desktop/src — 22 commits
- agent/tools — 20 commits
- agent/protocol — 7 commits
- docs/ja — 7 commits
- channel/feishu — 6 commits
- docs/zh — 5 commits
- agent/guest — 4 commits
- cli/commands — 4 commits
- common/cloud_client.py — 4 commits
- models/gemini — 4 commits
- channel/wecom_bot — 3 commits
- (root) — 2 commits
- .github/workflows — 2 commits
- agent/knowledge — 2 commits
- channel/discord — 2 commits
- channel/weixin — 2 commits
- cli/update_service.py — 2 commits
- plugins/cow_cli — 2 commits
Notable commits
- fix: Fix Feishu and Lark websocket domain selection
- fix: Merge pull request #3798 from rudycelekli/fix/read-word-simple-fields
- fix: Merge pull request #3802 from rudycelekli/fix/browser-profile-prefix-20261020
- fix: Merge pull request #3807 from rudycelekli/fix/scheduler-edited-completion-20261021
- fix: Merge pull request #3813 from rudycelekli/fix/modelscope-stream-usage-trailer
- fix: Merge pull request #3818 from rudycelekli/fix/telegram-remote-file-reply
- fix: Merge pull request #3819 from rudycelekli/fix/slack-remote-file-reply
- fix: Merge pull request #3820 from rudycelekli/fix/discord-remote-file-reply
- fix: Merge pull request #3821 from rudycelekli/fix/chatgpt-vision-temp-cleanup
- fix: Merge pull request #3830 from Rofds/fix-chat-composer-focus
- fix: Merge pull request #3835 from c020627/fix-linkai-asr-text-guard
- fix: Merge pull request #3837 from c020627/fix-voice-send-temp-cleanup
- fix: Merge pull request #3838 from c020627/fix-weixin-qr-expired-refresh-guard
- fix: Merge pull request #3843 from pei711/fix/minimax-tool-result-ordering
- fix: Merge pull request #3845 from c020627/fix-wecom-image-exif-orientation
- fix: Merge pull request #3849 from xiong-wei1/fix/trajectory-eval-expectations
- fix: Merge pull request #3850 from pcbeingused333/fix/gemini-usage-thoughts-tokens
- fix: Merge pull request #3851 from NotZhu/fix/loop-hint-not-persisted
- fix: Merge pull request #3853 from tanishqbaweja/fix/syntax-gate-jsonc-helm
- fix: Merge pull request #3854 from tanishqbaweja/fix/edit-fuzzy-outer-newlines
- …and 222 more
Architecture
- Unchanged — 0 containers · 1 contexts · 0 edges
Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.
Survey your own repository
zhayujie/CowAgent was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.
About this page
- The score is its most recent published measurement, taken on 14 November 2026 at a pinned commit. It is not a live figure and does not change until the project is measured again.
- Measured at commit fd75a33d862fe7ea9dec2fa1c0a27f92f181985d — the exact code this score is about.
- Scored under rubric-2026.11.4 — the same rubric and the same method as every other entry in this index.
- Measured by Canine Code Analyzer (cca.canine.dev) using cca measure-2026.11.65+preprod-36d4cd83012f.