You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
feat: speaker resolution, peer reactions, market indices, and solo-agent fallback
- Fix speaker name resolution: normalize endpoint URIs to canonical agent names
(Finn/Lucky/Prudence) in both template and utterance handlers to enable
proper peer-to-peer interactions with name-prefixed responses
Signed-off-by Deborah Dahl <dahl@conversational-technologies.com>
- Fix peer reaction counter: reset _peer_reaction_count on each fresh user
question so agents continue reacting to each other through conversational turns
- Add out-of-domain fallback: agents alone on floor now respond with
"that's outside of my expertise" for queries outside their domain expertise
- Add market indices support via ETF proxies: users can now query index data
(Dow, S&P 500, Nasdaq-100, Russell 2000) on Finnhub free tier by mapping
^DJI→DIA, ^GSPC→SPY, ^IXIC→QQQ, ^RUT→IWM with implied index levels in responses
- Tune agent personalities: Prudence disapproves of reckless/speculative advice;
Lucky disapproves of overly cautious, fear-driven investing
- Improve web-floor UI: increase conversation history font size to clamp(1rem,
2.1vw, 1.15rem); make utterance textbox full panel width; allow resize handle
up to 80vh max height
Files modified:
- lucky/utterance_handler.py (canonical names, peer reset, personality)
- lucky/template_agent.py (speaker resolution, solo-agent detection, fallback)
- prudence/utterance_handler.py (canonical names, peer reset, personality)
- prudence/template_agent.py (speaker resolution, solo-agent detection, fallback)
- financial/utterance_handler.py (indices, proxy ETFs, formatted responses)
- financial/template_agent.py (speaker resolution, fallback text standardization)
- floor-implementations/.../styles.css (font, width, height tuning)
0 commit comments