diff --git a/README.md b/README.md index 075eecb..381716b 100644 --- a/README.md +++ b/README.md @@ -380,6 +380,27 @@ Assigned based on your message patterns: --- +## Sentiment Classification + +Every message (subject to `LLM_SAMPLE_RATE`) is classified by Ollama into one of 10 sentiment categories. The bot reacts with an emoji when the confidence score is strong enough (|score| > 0.5). Per-user counts are tracked in the database and viewable via `!sentiment`. + +| Sentiment | Emoji | Score range | Example | +|-----------|-------|-------------|---------| +| Positive | ๐Ÿ‘ | > 0.5 | "This is awesome, great work!" | +| Excited | ๐Ÿ”ฅ | > 0.5 | "OH MY GOD I can't wait for this!!" | +| Supportive | ๐Ÿค— | > 0.5 | "You've got this, don't give up" | +| Grateful | ๐Ÿ’œ | > 0.5 | "Thank you so much for helping me" | +| Humorous | ๐Ÿ˜‚ | > 0.5 | "lmao that's the funniest thing I've seen all day" | +| Curious | ๐Ÿง | > 0.5 | "How does that work exactly?" | +| Neutral | โ€” | โ€” | "I'll be back in 10 minutes" | +| Sarcastic | ๐Ÿคจ | < -0.5 | "Oh sure, that'll definitely work" | +| Frustrated | ๐Ÿ˜ฎโ€๐Ÿ’จ | < -0.5 | "I've been trying to fix this for three hours" | +| Negative | ๐Ÿ‘Ž | < -0.5 | "This is broken and nobody cares" | + +The LLM also returns a float score (-1.0 to 1.0) for each message. These scores are averaged per user to derive an overall mood shown in `!sentiment` output and fed into `!howami` roast profiles. + +--- + ## External APIs All optional. The bot works fine without any of them, you just won't have those specific features.