Welcome to gpt-flow! This is a demo application for making AI and MLP tools using a flowchart interface. Create a flow that can make large language models generate code, talk through issues or create text datasets. Watch the video demo before getting started.
Leave a star or drop an issue on the github link. For business inquiries or to check out other projects from shmage.xyz, click the icon in the top right.
Try gpt-flow- Due to how aws instances work, your first request to the bot may take longer than subsequent requests. Give your first request 15 seconds or so to warm up the instance.
- You can delete nodes by clicking them and pressing the backspace key.
- Cyclic flow graphs do work but will take a while and timeout the loading wheel. Try to be patient on responses from cyclic graphs.
- Try to make prompts to individual nodes varied, similar node prompts will lead to similar answers, even if they are receiving output from a different node.
- You can have one node direct multiple other nodes by having one node output to multiple other nodes. This can be a great way to have the bot generate a system context for other nodes automatically.
- You can upload a custom json that you've generated yourself by clicking the "Upload flow JSON button" and then clicking "Load Flow JSON". This can be a good way to generate your own flows programtically before running.
- gpt-flow is great for trying dataset generation. You can download a history of all text outputted to the bot by clicking the "Download History" button in the top right.
-UI improvements
-Model selector
-Interoperability with image-to-text, text-to-image models
-Better rendering on mobile
- UI tweaks, slight visual upgrade on light mode
-publish gpt-flow
-add character limit to cyclic flow graphs
-increase idle response time to 10 minutes