Welcome • User Guide
ToS • Privacy • Canary
Donate • Bugs • License

©2026 Poal.co

1.2K

I've been working on my own architecture, and im renaming Tay to Tater. But upon looking at the code and requirements, Tay is not that hard to code, especially in python. It's essentially a graph reflector with some levenshtein and heuristics, but nothing out of this world. Heck you can even add a random node walker to create fun conversations or use weighted descendant trees instead. Point is, expect the code base link by next month

I've been working on my own architecture, and im renaming Tay to Tater. But upon looking at the code and requirements, Tay is not that hard to code, especially in python. It's essentially a graph reflector with some levenshtein and heuristics, but nothing out of this world. Heck you can even add a random node walker to create fun conversations or use weighted descendant trees instead. Point is, expect the code base link by next month

(post is archived)

You are currently inside a comment thread.
Click here to see all the comments (10).
[–] • 0 pt

possibly, im using a graph where the bot traverses to next child based off the input. Essentially a massive decision tree with tons of nodes. Im adding the ability to tag certain nodes with the stems so that the bot can also do traversals by topic. But for the most part, its a basic graph with a random traversal for now. So hey would have nodes for "hey", "how are you", "what are you up to", and it would simply pick one of those as a prediction. For the alpha, the ui will let you force a weight on the node, since the bot wont know until later which speech is most popular

[–] • 1 pt

Check out Chatterbot (you can use external corpus for training)

https://chatterbot.readthedocs.io/en/stable/

[–] • 1 pt

That would be super helpful actually, thanks

[–] • 2 pts

Pauline, in our chat, uses a modified version of that code.

https://chat.poal.co