Skip to content

CHANGELOG.md

Latest commit

 

History

History
164 lines (132 loc) · 5.76 KB

CHANGELOG.md

File metadata and controls

164 lines (132 loc) · 5.76 KB
Raw

Change Log

All notable changes to the "vscode-ollama" extension will be documented in this file.

Check Keep a Changelog for recommendations on how to structure this file.

[1.3.0] - 2026-08-19

Added

  • Coding agent mode: the model can use tools (read / write / edit / grep / find / ls / bash) instead of only direct chat
    • Read/search/list tools run automatically; file edits and command execution are sandboxed to the workspace
    • Multi-turn context persistence so the agent keeps the full session history
    • "vibe" / "go on" continue signal: the agent autonomously keeps working until the task is done, no repeated prompts needed
  • Agent status bar showing the model's tool activity (thinking / tool call / result)
  • Skill system: the agent auto-loads matching skills from skills/*.md (workspace first, then extension-bundled)
  • Plan mode: prefix a request with /plan (or open-ended codebase questions) to propose a plan first, then execute automatically
  • Web search as a tool (web_search), decided by the LLM per-question; codebase questions prefer grep / find / ls / read
  • scripts/verify-webview.js checks webview script syntax and CSP safety; runs as part of npm run compile

Changed

  • Switched from Ollama's raw /api/chat to the OpenAI-compatible /v1/chat/completions endpoint (standard tool_calls pairing, SSE streaming, max_tokens)
  • Unified build to esbuild (outputs dist/extension.js); marked.min.js is bundled locally for the webview
  • Added Content Security Policy with nonce-based scripts to both webviews
  • Rewrote the agent system prompt (Cline-style) with hard rules: codebase questions MUST use tools, real-time questions MUST use web_search, never ask the user to run commands
  • Intent judgment now fully handled by the LLM; removed all regex-based intent detection
  • grep tool falls back to POSIX grep when ripgrep is not installed

Fixed

  • Message history ordering bug (system prompt was appended to the end of the context)
  • Streaming response parsing when a JSON line is split across chunks
  • Webview script syntax errors and CSP-blocked handlers that broke chat sending and the model dropdown
  • XSS risk: LLM/user content is now HTML-escaped before rendering
  • Tool calls placed inside message.content (e.g. qwen2.5-coder) are now parsed via <tool_call> tags or bare JSON
  • Removed stale src/extension.js that conflicted with extension.ts

[1.2.1] - 2025-08-21

Changed

  • Translated chat homepage UI to English (welcome message, menu, input placeholder)

[1.2.0] - 2025-02-26

Added

  • Enhanced web search functionality
    • Use Bing as default search engine
    • Added DuckDuckGo as fallback search engine
    • Improved search reliability and stability
  • Added 'Thinking' time for DeepSeek model

Changed

  • Refined chat view interface
    • UI/UX improvements
    • Enhanced visual presentation
    • Optimized response handling

Fixed

  • Fixed web search functionality issues
    • Improved error handling for search engines
    • Enhanced page content fetching mechanism
  • Fixed various stability issues
  • Fixed /reset not working properly

[1.1.0] - 2025-02-24

Changed

  • Moved connection test functionality to settings page
  • Improved output message formatting
  • Enhanced overall user interface organization

[1.0.4] - 2025-02-21

Fixed

  • Fixed an issue where new chat responses were displayed in the wrong position after stopping generation
  • Fixed incorrect message appending in multiple conversation threads
  • Optimized streaming response handling logic for chat messages

[1.0.3] - 2025-02-20

Added

  • Support for web search
  • Support deepseek-r1 reasoning model

Changed

  • Optimized text display for better readability
  • Improved message rendering performance and stability

Fixed

  • Fixed incomplete display of English text
  • Fixed message formatting issues
  • Fixed newline character display issues
  • Fixed response styling issues

[1.0.2] - 2025-02-20

Added

  • Added "/reset" command to reset the conversation context
  • Added system prompt configuration support

Fixed

  • Fixed chat message context management based on maxTokens setting
  • Improved message history token limit handling

[1.0.1] - 2025-02-18

Added

  • Additional features in the ...menu
    • Theme switching functionality (Coming Soon!!!)
    • Chat history clear

Changed

  • Improved chat interface styling
    • Enhanced chat history and input box borders
    • Unified border colors and thickness
    • Better overall visual experience

Fixed

  • Fixed welcome message loss after clearing chat
  • Fixed theme state synchronization issues
  • Fixed various TypeScript type checking issues

[1.0.0] - 2025-02-17

Initial Release

  • Local Ollama service integration
  • Multiple LLM model configuration support
  • Web search capability
  • Intuitive chat interface
  • Model configuration and performance mode settings

Added

  • Initial release with core functionalities
  • Local LLM support with Ollama integration
    • Multiple model support and easy switching
    • Low-latency responses
    • Secure and private environment
  • Web search capability(Coming Soon!!!)
    • Real-time information integration
    • Smart search results synthesis
    • Accurate information citation
  • Streaming chat interface
    • Markdown support
    • Code-aware context
    • Syntax highlighting
  • Chat history preservation
    • Local storage
    • Session management
    • Export functionality
  • Thought process visualization
    • Step-by-step reasoning
    • Intermediate results display
  • Customizable settings
    • Server configuration
    • Model selection
    • Performance modes
    • Token management
    • Keep-alive settings

Technical

  • VS Code API integration
  • WebView implementation
  • Ollama API integration
  • Error handling and recovery
  • Performance optimizations