v.0.5.0-beta

[awtkns]

🎉🤖   Beta 0.5.0   🤖🎉

Welcome to the Beta 5.0 release of AgentGPT! With this comes a barrel of features to go bananas over. Some of the highlights are:

• Migration to a FastApi based backend which has replaced our NextJS backend. The maturity of Python LLM tooling was a key factor when making this decision.
• Addition of various tools such as Code Generation, Image Generation (via Replicate) and Web Search
• Numerous styling and UI improvements

We will be releasing updates on a weekly schedule, stay tuned every Tuesday for the latest improvements to AgentGPT 🚀 If you'd like to help shape these updates join our discord server. We would love to hear your feedback. ❤️

👀 What's Changed

• ✨ Support for custom base URL by @asim-shrestha in ✨ Support for custom base URL #543
• ♻️ Refactor api by @awtkns in ♻️ Refactor api #555
• 💅 Update styling for chat window buttons by @asim-shrestha in 💅 Update styling for chat window buttons #559
• 💅 Update ChatMessage styling by @asim-shrestha in 💅 Update ChatMessage styling #560
• ✨ Update analysis by @asim-shrestha in ✨ Update analysis  #566
• 🛠️ Tools selector by @asim-shrestha in 🛠️ Tools selector #569
• ✨ Sync by @asim-shrestha in ✨ Sync #570
• 🌐 Enable web search by default by @asim-shrestha in 🌐 Enable web search by default #572
• 🐛 Fix "Conclude" task by @asim-shrestha in 🐛 Fix "Conclude" task #574
• 🛠️ Select tools in the frontend by @asim-shrestha in 🛠️ Select tools in the frontend #575
• 🐛 Fix mypy issues by @asim-shrestha in 🐛 Fix mypy issues #576
• 💻 Add code writer tool by @asim-shrestha in 💻 Add code writer tool #585

Full Changelog: v.0.4.0-beta...v.0.5.0-beta

---

This discussion was created from the release v.0.5.0-beta.

[zhongqingsong]

Title is "🎉🤖 Beta 4.0 🤖🎉" , but now it's Beta 5.0. Once Again.

[yhyu13]

How is web search paid in 0.5? Do we still need to bring our own serp key?

[asim-shrestha]

Yeah exactly, not sure if there are any other ways to reliably do it. I know theres some self hostable search APIs

[yhyu13]

I found we are still summarizing web search results by serp's given snippets, right? I tried to add a crawler with puppeter but nodejs alwasy complain that native http request package is not supported.

I think we should offload crawling to py backend

[yhyu13]

Another stupid question, how do we debug the py backend? I used VSCode auto attach debugger for npm client,