@marty_kausas DevFlow, our internal tool open to beta testing. Choose the LLM you want for every agent (planner, implementer, verifier, summarizer, etc..).
Integrate with sentry, get notified on telegram, teams or mattermost.
Just bring your open router key and choose which model you want.
@GergelyOrosz Doing exactly that. An orchestrator runnable from slack where you can customize the model. Currently in private beta testing. Hopefully will see the broader lights one day
@jpschroeder Agreed. Iโm testing my orchestration tool and ds4 flash for the implementation part is incredible.
Sure you canโt use it for complex reasoning but once you have detailed instructions it behaves pretty well
@Adidotdev Only in benchmarks. I have a tool where I ask an LLM to output a plan in json and GLM is the only one constantly failing into following the instructions of returning a json. Minimax or Kimi never miss a shot
@antirez DeepSeek v4 Flash under the right prompts is great!
Right now I'm testing my own harness planning with Qwen 3.7 Max and implementing with DS Flash.
Comparing it with full Opus (plan + implementation) is right now 100% worth it.
@OpenAI you have a couple of bugs in your Function Call Docs Page for javascript.
1) not sending back the response.output containing the function calls
2) input_list should be input
Cheers.
https://t.co/bsUE8IJrqx