SRS is a simple, high-performance, AI-driven real-time media server, supporting RTMP, WebRTC, HLS, HTTP-FLV, SRT, and MPEG-DASH.
Joined February 2022
- Tweets366
- Following173
- Followers490
- Likes68
SRS 8.0 dev 1 was released, with the detail performance data compared to previous version, which is generated by AI obviously.
AI is now able to analyze and improve the performance of SRS. By modifying just one line of code, AI improved about 9% performance when SRS serving 12.8K viewers. That's amazing.
AI found and fixed 135 bugs in SRS 8.0. github.com/ossrs/srs/pull/47…
AI auto run for 16 hours without my help or approve, updated 226 files 56K lines of code, created 1374 integration tests for SRS, found 111 bugs. But how can I verify all the bugs? It would take days to review that bugs one by one manually. 😂
AI add qemu test for SRS/ST, convered all the CPUs and OS that we used. github.com/ossrs/state-threa…
In AI era, the PullRequests and Issues are in fact no more than ideas. We don't need code implementation, because AI can do this automatically with high quality code. But idea still make a lot of sense, and it will always be a important part of open source community.
I just published Faster and Better, How AI Maintains SRS with Higher Quality Code than Humans blog.ossrs.io/faster-and-bet…
Within 2 hours, AI wrote 11 tools for integration test tools for ST (State Threads), with these tests, AI is able to verify all the features of ST, and assure you no bug introduced when porting to native Windows platform.
github.com/ossrs/state-threa…
AI is ready to maintain the ST (state-threads) project. ST is very difficult to maintain for human, but maybe it's a different story for AI. Let's see.
Add help for AI, recommend AI to use environment variables to config SRS, which is very powerful and convenient for AI to control and manage the service.
AI finished another demo: How to publish WHIP stream by your DJI drone. DJI doesn't support WHIP in consumer drones. This demo uses DJI MSDK to get the h264 raw stream from drone, then transmux it to RTMP/WHIP by Kotlin, and publish to your media server.
github.com/ossrs/srs-dji-dem…
This is the first huge and complicated feature AI finished, WebRTC RTX feature, Fable 5.1 worked for 10h, generated 9k lines of code, 0 bugs introduced, $360 cost. Really fantastic job!
github.com/ossrs/srs/pull/47…
This is the first SRS version that is published by AI. The process was decent and fast. I suppose AI will speed up the new versions to every month, before AI era it was per year. github.com/ossrs/srs/release…
AI has been improving the code quality of SRS. I'm delighted by the tests that AI added before change any code. AI follows skills super well, which makes it a good TDD software engineer, much better than most of human engineer I ever know.
Oryx v7.15.26 was released, upgraded SRS from 5 to 7. Now AI is driving this project. github.com/ossrs/oryx/releas…
AI rewrote the official SRS website in just two days. It migrated all the documentation to the new version of Docusaurus and tested all the links, in just two days. About three years ago, I spent two weeks creating the first version of the website. github.com/ossrs/srs-docs2
