AI Matchmaker Report With Video

Estimated LLM cost (last full run)

Run Generate all (or the initial load pipeline) to estimate cost from prompt sizes, models, and thinking levels. Uses heuristic token counts and approximate public Gemini rates; dating cards add a flat allowance for video input (not measured from file duration).

LLM configurations (Gemini) ▼

Positivity score

Scores one date transcript 1–5 for how positive the person was. Used in compatibility/chemistry. Response must be a single integer 1–5.

Model:

Gemini 3 thinking level:

Sent as generation_config.thinking_config.thinking_level. 3.1 Pro sets this to low; 3 Flash to medium when you change the model.

System Prompt:

User Prompt (transcript inserted at {{TRANSCRIPT}}):

Compatibility score formula (variable placeholders)

getCoupleValue(raw):  0,2→0; 1,3→5; 4,5→3
getGiftValue(amount):  0→0; (0,500]→0.5; >500→1
posNorm = positivityFromLLM / 5   (1–5 → 0–1)

decisionScore = (daterDec + partnerDec) / 2   // each 0,3,5
giftScore      = (daterGift + partnerGift) / 2   // each 0, 0.5, 1
posScore       = (daterPosNorm + partnerPosNorm) / 2  // both from LLM (or partner→1 if no transcript)

combined = (decisionScore/5)*40 + giftScore*30 + posScore*30   // 40% + 30% + 30% → 0–100
finalScore = round(clamp(combined, 0, 100))

Intro paragraph

Model:

Gemini 3 thinking level:

Sent as generation_config.thinking_config.thinking_level. 3.1 Pro sets this to low; 3 Flash to medium when you change the model.

System Prompt:

User prompt (read-only) — text-only context sent to Gemini (no videos):

Dating card prompt

Model:

Gemini 3 thinking level:

Sent as generation_config.thinking_config.thinking_level. 3.1 Pro sets this to low; 3 Flash to medium when you change the model.

System Prompt:

User prompt (read-only) — data we send per card:

User prompt is sent automatically per date: video with LEFT/RIGHT labels, dater name, partner name, interests, bio, and decisions/ratings (daterDecision, partnerDecision, looks/personality ratings and types).

Dating insights summary

Model:

Gemini 3 thinking level:

Sent as generation_config.thinking_config.thinking_level. 3.1 Pro sets this to low; 3 Flash to medium when you change the model.

System Prompt:

User Prompt (per-date JSON with daterDecision, partnerDecision, theDynamic, myCoachingNotes, attraction, personality):

Takeaways

Model:

Gemini 3 thinking level:

Sent as generation_config.thinking_config.thinking_level. 3.1 Pro sets this to low; 3 Flash to medium when you change the model.

System Prompt:

User Prompt (per-date card fields only — same as intro):

First impressions

Model:

Gemini 3 thinking level:

Sent as generation_config.thinking_config.thinking_level. 3.1 Pro sets this to low; 3 Flash to medium when you change the model.

System Prompt:

User Prompt (ai_matchmaker_report_stats):

Note: This user prompt is dynamic: it updates with ai_matchmaker_report_stats for the currently selected event and user.

Word cloud and quotes

Model:

Gemini 3 thinking level:

Sent as generation_config.thinking_config.thinking_level. 3.1 Pro sets this to low; 3 Flash to medium when you change the model.

System Prompt:

User Prompt (confessional transcripts):

Generate

—

Intro paragraph.

First impressions

Profile views

— views

0 5

Date requests

— requests

0 10+

Couples

—%

0% 25% 50% 75% 100%

Connects

—%

0% 25% 50% 75% 100%

Dating insights

Insights placeholder.

Select an event and user above to see your dates.

Top takeaway

Ready to make the most of your next Couple event? Here's my top tip for you after observing your dates tonight.

248 people are already signed up!

AI Matchmaker Report

Select User and Session

AI models

Selected

Date Transcript

LLM prompts

LLM configurations (Gemini) ▼

Positivity score

Intro paragraph

Dating card prompt

Dating insights summary

Takeaways

First impressions

Word cloud and quotes

Generate

—

First impressions

Dating insights