Clanker – 452KB emotional scoring engine strapped to Llama-1B

Spaces
1

Built a deterministic engine that computes 7D emotional coordinates (VADUGWI) from text structure. Hooked it up to Llama-3.2-1B in a Gradio Space.

The model generates dialogue between two characters. The engine scores every line on 7 dimensions and tracks how each message shifts the other character’s emotional baseline. State carries forward through A+B=C transitions.

What the engine does that the model can’t:

  • Detects 26 structural patterns (VICTIMIZATION, SELF_NULLIFY, SARCASM_INVERSION, CHOPPER_SPLIT, etc.)
  • Tracks self-worth (W) separately from valence – blaming yourself reads differently than blaming the world
  • Tracks intent direction (I) – reaching out vs pulling away vs commanding
  • Runs at 0.15ms/sentence on CPU, ~452KB total

The Space has two tabs:

  1. Two characters argue (pick personalities, give them a topic, watch the emotional states shift)
  2. Score any text (type anything, see the raw coordinates)

What I’m working on now: training a Phi-4-mini LoRA to actually learn the VADUGWI math so it can compute emotional coordinates internally, not just generate text.

Space: https://huggingface.co/spaces/deucebucket/clanker

Would love feedback on what the engine gets right and where it falls apart. Try typing something sarcastic in the Score tab.