Hey everyone! My name is Adi, I am the founder of FlexInference.
I built this to reduce AI spend for all my side projects. I then turned it into a proxy for Anthropic, OpenAI, and Gemini. I also then built a python and typescript SDK with strict types and better error messages. This way I can do agentic coding with a bit less wrangling.
Now I am making it available to the public.
The way it works is you set a SLA in the "start_within" parameter. If it's default or auto then it gets sent as is to the provider. If you set a time between 5 seconds and 10 minutes then it first gets requested as a flex request. If that starts getting processed then great you get 50% off. If not then the router escalates it to a default request. This way you always get a response. But if possible in the SLA you provide, we will get you cheaper inference.
I made the router free because it just never made sense to me why the ones in the market charge 5.5%. It also supports Interactions, Messages, Chat Completions and Responses. Its deployed through Cloudflare workers so it's pretty fast. And you can see all your metrics from within the dashboard.
I would really love it if you all could try it out. It's free and it doesn't require a credit card until you need a "flex race". So if anything you can just save the fee you pay per request by using this instead for that.
Let me know if you have any suggestions, bugs, feature requests, or just wanna yell at me.
Hey everyone! My name is Adi, I am the founder of FlexInference.
I built this to reduce AI spend for all my side projects. I then turned it into a proxy for Anthropic, OpenAI, and Gemini. I also then built a python and typescript SDK with strict types and better error messages. This way I can do agentic coding with a bit less wrangling.
Now I am making it available to the public.
The way it works is you set a SLA in the "start_within" parameter. If it's default or auto then it gets sent as is to the provider. If you set a time between 5 seconds and 10 minutes then it first gets requested as a flex request. If that starts getting processed then great you get 50% off. If not then the router escalates it to a default request. This way you always get a response. But if possible in the SLA you provide, we will get you cheaper inference.
I made the router free because it just never made sense to me why the ones in the market charge 5.5%. It also supports Interactions, Messages, Chat Completions and Responses. Its deployed through Cloudflare workers so it's pretty fast. And you can see all your metrics from within the dashboard.
I would really love it if you all could try it out. It's free and it doesn't require a credit card until you need a "flex race". So if anything you can just save the fee you pay per request by using this instead for that.
Let me know if you have any suggestions, bugs, feature requests, or just wanna yell at me.
Thanks,
Adi