SiYuan AI agent context canceled

Hello everyone,

I'm just trying out the new AI assistant features and I've immediately ran into a problem.

I'm using DeepSeek v4 Flash Free from OpenCode Zen (and have also tried Xiaomi Mimo 2.5 Free from the same provider), I've tried giving it a long-running task, and I've noticed that it's very easy for the agent to crash. Looking at the errors in the logs, I see these two types of errors:

• E 2026/07/03 12:30:34 agent.go:477: agent API request failed: Post "https://opencode.ai/zen/v1/chat/completions": context canceled
• E 2026/07/03 12:40:59 agent.go:493: agent stream error: context canceled

The first error happens right after I start the chat, the second happens when the response is taking long (for example, long LLM reasoning). I have noticed the second error often comes at around 60s of thinking, but often less.

This is not a problem with the API provider, I'm using it on VSCode with Cline and have not had any problem at all, even right after testing SiYuan.

Here is my configuration:

I don't have any custom MCP enabled or notes indexing.

While I'm at it, I will also report a few other minor issues:

• Some strings have not been translated to English: 智能体 MCP 服务, 嵌入模型
• We're missing the feature to edit our last message in the AI conversation and re-generate the answer
• If an error occurs in the answer generation, such as those I have reported above, SiYuan shows the error message The network is abnormal, please try again later. This is okay, however, when this happens, the message does not have a "retry" button, which would be very useful!

I know this is an early access feature, so it's okay to see a few errors. I like this new version a lot, and I hope that with time, the SiYuan team will be able to iron these wrinkles.

Please let me know if I should set something differently in the AI settings, or if it's a known bug. Thank you!