# SekerAI (סקר AI)

> AI-powered synthetic market research SaaS platform calibrated by the Israeli Central Bureau of Statistics (CBS / למ"ס) demographic models, powered by Google Gemini, featuring real-time cross-tabulation, internal monologue inspection, and 1-on-1 depth interviews.

SekerAI is part of the Kesher Ecosystem: https://kesher.me
Parent Hub LLM Index: https://kesher.me/llms.txt
Canonical Domain: https://seker.kesher.me

## Core Mission & Architecture
SekerAI solves the bottleneck of traditional market research in Israel (which often takes 2–3 weeks and costs 15,000–35,000 ILS per survey) by providing representative synthetic samples within 3 minutes at a fraction of the cost.

### 1. Demographic Distribution Calibration (CBS / למ"ס)
- Calibrated to representative Israeli demographic distributions across 6 key axes:
  - Age groups (18–29, 30–44, 45–59, 60–75)
  - Gender parity (49% Male, 51% Female)
  - Districts (Center & Gush Dan, Jerusalem, North, Haifa, South, Judea & Samaria)
  - Sectors (Jewish, Arab, Druze / Other)
  - Religiosity spectrum (Secular, Traditional, Religious, Ultra-Orthodox / Haredi)
  - Income levels (Below average, Average, Above average)
- Enforces natural authentic demographic correlations (e.g. city-to-district mappings, sector-specific occupational nuances, family sizes, and housing status).

### 2. Google Gemini Reasoning Pipeline
- Primary Engine: Google Gemini Flash (`gemini-2.5-flash` / `gemini-3.5-flash` with structured outputs `responseSchema`).
- Chunked structured execution: processes personas in optimized batches, achieving over 70% token savings while drastically reducing latency.
- Strict anti-hallucination guardrails: no fabricated synthetic mock fallbacks; live exponential backoff with multi-region fallback routing.

### 3. Internal Monologue Inspection (`internalMonologue`)
- For every question answered, the synthetic persona generates a confidential first-person thought process (`internalMonologue`) before locking in their quantitative choice.
- Researchers can inspect what personas secretly thought regarding pricing, brand loyalty, skepticism, and daily Israeli lifestyle constraints.

### 4. Zero-Latency Cross-Tabulated Analytics (0ms)
- Instant client-side recalculation across demographic splits (e.g. Center vs Periphery, Secular vs Religious, High vs Low Income, Women vs Men).
- Deep option drilldowns: clicking on any chart bar immediately pops open the individual personas who chose that option along with their thoughts.

### 5. 1-on-1 Depth Interviews (Dynamic Chat)
- Researchers can open a live 1-on-1 chat with any persona at any time.
- The persona remembers its demographic persona skeleton, survey answers, and internal monologues, responding with authentic conversational Hebrew reflecting its exact background.

### 6. Public Shared Interactive Research Dashboards (`/share/[id]`)
- Owners can publish read-only interactive research dashboards accessible to the public without requiring login.
- Public viewers can freely filter, drill down, and conduct 1-on-1 interviews with personas, while the survey questions and data remain securely protected from mutation.

## Key Links & Endpoints
- Homepage & Interactive Simulation: https://seker.kesher.me
- Public Survey Sharing: https://seker.kesher.me/share/{surveyId}
- Login / Registration: https://seker.kesher.me/login
- Full LLM Specification & Prompts: https://seker.kesher.me/llms-full.txt
- Kesher Ecosystem Hub: https://kesher.me