From be858855c8ec1cf646c472b271f87112709f2335 Mon Sep 17 00:00:00 2001 From: aodulov Date: Mon, 13 Oct 2025 14:02:09 +0300 Subject: [PATCH] Higher specificity for matching "Afraid to Ask" --- backend/src/services/LLMService.ts | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/backend/src/services/LLMService.ts b/backend/src/services/LLMService.ts index d963a29..d34f792 100644 --- a/backend/src/services/LLMService.ts +++ b/backend/src/services/LLMService.ts @@ -40,7 +40,7 @@ export class LLMService { - "noGoes": Synthesize a text describing what at least ONE participant does not want. This should clearly state the collective exclusions. Use the more broad opinions summarizing all the specific options if they do not contradict each other drastically. - "needsDiscussion": Synthesize a text describing where there is a direct conflict (e.g., one participant wants it, another does not want it). This should highlight areas requiring further negotiation. Do not include 'afraidToAsk' in this category. - 'AfraidToAsk' ideas that do NOT semantically match any other participant's 'wants' or 'accepts' should remain private and NOT be included in any of the synthesized categories. + 'AfraidToAsk' ideas that do NOT semantically match any other participant's 'wants' or 'accepts' very closely should remain private and NOT be included in any of the synthesized categories. Matching must use minimal level of generalization. Formulate common ideas from the point of 'us', e.g. "We are going to...", or "We want to...", or "We think...", or "We do not...".