Diary on 2024-11-26

Programming

Since yesterday, I managed to extract the original pw-cli C code to list the nodes of PipeWire, which can be embedded in Go. Today, I aimed to replicate this method to get PulseAudio devices that old Linux systems rely on. However, it turned out to be unlucky. I tried several LLM models and tools, and several rounds of testing, but encountered numerous compiling errors and segment faults. During this process, I switched between different LLM models, but none of them worked; so I gave up.

At night, I tried to integrate whisper.cpp into the Go code base. It still didn’t work, but I know the direction. I will reference ollama to improve. I created a sample and will continue tomorrow.

Working

Following up on the APE issue, both Lucca and I tried building locally on our machines. Lucca attempted to deploy the application to G23 but still failed. The logs suggest that there may be an issue on the server side. She will ask her team member to investigate it.

Another issue today is that in MMT, there is a CodeQL scan (GHAS) issue. The scan has already generated reports and found the Java files, but it ultimately prompted error code 32, indicating that it didn’t find files to scan. This is really weird. I opened a support ticket with the team to see if they can help with this.

Today, Doug rolled out v2 of the iflow deployment service, with huge improvements. Now, it can import and parse YAML format requests and create jobs in a flash, dramatically reducing our workload. Well done.

儿子学习

今天继续给儿子读哲学，之前没有提到这本书的名字《苏爷爷的哲理故事 把问题提出来》

• 游历丰富的人
• 阅读了几页关于人工智能的书