5ire
Roadmap
Feedback
Changelog

Changelog

Follow up on the latest improvements and updates.

RSS

new

improved

fixed

v0.9.8

5ire-folder
Supports Folder for Conversation Organization
Folders now support model, temperature, and system message configuration. Child conversations automatically inherit parent folder settings. Additional folder features are coming in future updates.
Linux Beta Release Now Available
The current beta version has the following known issues:
  • Application icon missing
  • Requires the --no-sandbox flag during launch for proper execution
Note
If encountering "X server or $DISPLAY environment variable not detected" error, temporarily set display configuration with:
export DISPLAY=127.0.0.1:0.0
I'm not familiar with Linux, especially the GUI env yet. The Linux version will be gradually improved later.
Enhanced Update Experience
Redesigned the update interface with streamlined version checking and installation workflows.
Bug fixes
Resolved a bunch of issues to deliver improved application stability and user experience.

new

improved

fixed

v0.9.7

Support adding tools and installing them from the market.
We have created an open marketplace for MCP Servers. it empowers users to discover exceptional tools while offering a streamlined process for sharing their own MCP server creations.
Enable or disable tool support as needed.
Now, you can choose whether to enable tools. Before enabling tool support for a model, ensure the model itself supports it to avoid errors. btw, Ollama of 5ire also supports tools now.
5ire1
The circular badge before each model name visually indicates its tool compatibility status: the outer ring in a shade of gray-green shows the model's original capability, while the inner dot uses the same color gradient to reflect the current configuration setting
Standardize the API Base
Standardized the API Base across all providers. For example,
⚠️ Important: If you encounter a 404 error after the update, please check and update the API Base accordingly.
More Improvements
  • Support rendering of reasoning content.
  • Support for Anthropic's latest claude-3.7-sonnet model.
  • Support more google's experimental models.
  • And various bug fixes

new

improved

fixed

v0.9.6

  1. A new Inspector panel has been added to the chat interface to track the execution of tools.
  2. For enabled tools, you can view the definitions of the functions they support.
  3. It now supports rendering the think content of inference models such as DeepSeek.
  4. When configuring Ollama, you can choose a model.
  5. The maximum number of context messages has been increased to 99, with a default of 10.
  6. Some details have been optimized.
  7. Some bugs have been fixed.

new

improved

fixed

v0.9.4

  1. Added support for Mistral.
  2. OpenAI now supports o1, o1-mini, and o3-mini.
  3. Built-in MCP Server now includes TAVILY.
  4. Added font size adjustment feature.
  5. Fixed content loss in editor after switching conversations.
  6. Fixed tool call failure in Gemini 2.0 Flash Experimental.
  7. Optimized some details.

new

fixed

improved

v0.9.5

  • Added support for Gemini 2 Flash and Gemini 2 Flash Lite.
  • Support copying code blocks.
  • Delayed loading of MCP Servers to cut down on startup time.
  • Changed the shortcuts for conversion navigation.
  • Fixed some minor bugs.

new

improved

v0.9.3

What's Changed
  1. Added support for Deepseek-R.
  2. MCP Servers now include Apple Notes and Hacker News.
  3. Support language switching in settings.
  4. Added shortcuts support. For details, see: docs
Apple Notes and Hacker News MCP Servers require Python >= 3.12.0

new

improved

fixed

v0.9.2

What's Changed
  1. 🎉 Added model mapping feature.
  2. ✨ The built - in MCP Servers now include MySQL.
  3. ⚒️ Fixed the problem of Claude losing output content.
  4. ⚒️ Fixed the problem of Claude failing to call tools.
  5. ⚒️ Fixed the problem that the selected Prompt and Knowledge collection in the conversation are lost after the application is restarted.
What is Model Mapping?
Model mapping refers to the process of assigning a custom name to a certain model when making API requests. Since most service providers' APIs are compatible with OpenAI's, the primary difference often lies in the model names. Through model mapping, users can define their own model names. For instance, mapping
gpt-4
to
qwen2
and combining it with a custom API base allows users to leverage services from various aggregated providers seamlessly.

new

improved

fixed

v0.9.1

What's Changed
  • The built-in MCP Server now includes linear and slack.
  • Fixed: selected knowledge bases and prompts lost on app restart.
  • UI optimized to highlight supported tool models.
  • Minor bugs fixed.
Tip
5ire is currently in the early stage, so custom MCP Servers are not open yet. However, you can add servers by modifying the
{userData}/5ire/mcp.json
configuration file (It should be
~/Library/Application Support/5ire/mcp.json
on Mac). If there's any error, just delete the file, and 5ire will reset automatically. Feel free to explore!

new

improved

v0.9.0

Support Tools 🎉
Now, 5ire supports MCP-compatible tools. With tools, you can access the file system, obtain system information, interact with databases, access remote data, and more, rather than just having a simple conversation.
NOTE
Support for MCP Servers is still in its early stages, some tools perform abnormally on Windows, and some models may fail to call tools.
Others
  1. Supports Gemini 2.0 Flash Experimental model.
  2. Extensive code refactoring
  3. Upgrade Electron to v31.7.1

fixed

v0.8.4

Fix error in parsing docx、pptx and xlsx files.
Load More
→