ClientSynthAI

Safely Test Your Theories

Testing a specialized marketplace with real users is expensive, risky, and ethically complex. ClientSynthAI, part of the MarketForge suite, solves this.

Licensing: ClientSynth is source-available under the Business Source License 1.1 — free to study, test, and use, including in production, on two conditions: synthetic populations keep their provenance markings and are never passed off as real participants, and offering ClientSynth as a commercial service requires a separate written license (currently granted free on request). On 2030-07-05 the license converts automatically to MIT. And you don't need ClientSynth at all: Cosolvent's synthetic-population ingest schema is open — bring your own generator.

It provides a sophisticated engine to spawn an entire "digital twin" population of users—complete with conflicting priorities, distinct geographical attributes, unique resource constraints, and varied negotiation styles. You can safely populate a Cosolvent-based marketplace framework to watch deals stall, succeed, or fall apart organically.

Core Advantages

• ✅ Market Validation: Discover structural liquidity bottlenecks without spending capital recruiting real participants.
• ✅ Investor Demonstration: Showcase the elegant physics of your market platform operating under full-load conditions.
• ✅ Plausible Attributes: Uses LLMs to dynamically generate highly realistic profile content, documents, and historical behaviors.

Data Generation Pipeline

✅	AI-Powered Field Generation	Context-aware value generation via OpenRouter with per-field prompts and batch retry logic.
✅	17+ Field Types	Name, email, phone, company, address, date, number, text, image, PDF, enum, boolean, URL, JSON, custom AI.
✅	Job Processor	Production-grade batch processing with pause/resume/cancel and progress tracking.
✅	Persona Contexts	Coherent populations where fields are internally consistent (title matches industry).
🔜	Locale-Aware Generation	Culturally appropriate names, addresses, and certifications per country/region.
🔜	Time-Series Data	Generate data that changes over time — seasonal volumes, price fluctuations.

Schema Intelligence

✅	Visual Schema Designer	Drag-and-drop schema builder with field type selection, prompt configuration, and ordering.
✅	Schema Discovery from Files	Upload CSV, JSON, or XLSX and the system automatically infers a schema.
✅	Schema Induction	Field clustering, type detection heuristics, and LLM-assisted description generation.
🔜	Pre-Built Templates	Common data shapes (e-commerce, healthcare, financial, real estate) to start from.

Image & PDF Generation

✅	Context-Aware Images	Image prompts built dynamically from record data — headshots match role and demographics.
✅	Template-Based PDFs	Generate realistic certificates, invoices, and reports from structured data.
✅	S3 Storage Integration	Generated assets automatically uploaded to S3 with URL and metadata tracking.

Data Quality & Variation

✅	Distribution Manager	Define target distributions for categorical fields with deviation analysis.
✅	Pattern Detector	Catches when the AI falls into loops or repetitive templates.
✅	Similarity Scorer	Cross-record uniqueness scoring ensures generated records are sufficiently distinct.
🔜	Population-Level Scoring	Evaluate entire populations for market realism: buyer/seller ratios, coverage, capacity.

Cosolvent Integration

✅	Cosolvent Participant Export	Dedicated export format reshaping records into Cosolvent's participants JSONB format.
✅	API Streaming	Continuous webhook-based hydration of Cosolvent test instances.
🔜	MarketDefinition Import	Import a Cosolvent MarketDefinition and auto-generate conformant schemas.
🔜	Scenario Definition Language	Define population scenarios in JSON/YAML with geographic and role distributions.

Architecture

Language	TypeScript (strict)
Framework	Next.js 14 (App Router)
Database	PostgreSQL (RLS multi-tenant)
AI (text)	OpenRouter (Gemini, GPT, Claude)
AI (images)	Fal (multi-provider)
Deployment	Vercel (serverless)