Changelog

System Config DB — all configuration moved from .env files to PostgreSQL. Manageable via admin panel.

Model Catalog — 57 models across 27 providers with rate limits, pricing, and context windows

3 Free Runs — every new user gets 3 free builds using admin API keys

Brave Search integration — web_search tool uses Brave API when available

Nexus AI Providers panel — 27 AI providers with setup guides, test connection, set as default

AI Provider expansion: 10 → 27 providers (Groq, xAI, Fireworks, SambaNova, NVIDIA, GitHub Models, Cloudflare, Cohere, Perplexity, and more)

Nexus System Config page — manage all settings from admin panel, no .env needed

LLM fallback chain — automatic provider switching on quota/rate-limit errors

User Model Settings: 13 → 26 providers in quick-add grid

Model Settings: clickable model cards with pricing and rate limit info

Free Models page: 10 → 17 providers with free tier guides

Integrations page: 10 → 27 AI providers listed

Database password authentication — PostgreSQL connection working with correct credentials

Config migration — env vars now properly persisted to system_config table on startup

Webhook Continuation — external webhooks can trigger follow-up runs for indefinite autonomous loops

Deploy Executor — actual deployment to Vercel, Railway, Fly.io, Docker via CLI (deploy_to_platform tool)

Composable Modules page — select and integrate pre-built modules (auth, payments, email, upload, websocket, admin, search, rate-limiting)

Module Dependency Resolution — auto-detect and add missing dependencies, conflict detection

Marketplace Ratings — rate templates 1-5 stars, average rating display

Autonomous Delegation — Digital Clone makes framework/language/database decisions when you're away

Code Style Inference — detects your primary language, framework, testing tool, naming conventions

Benchmark Regression Tracking — track score trends over time

ROI Analytics Dashboard — hours saved, money saved, ROI%, model performance

Header navigation — 5 links: Features, Autonomy, Pricing, Docs, Compare (4 languages)

/autonomy dedicated marketing page (7-step flow, real example, competitive comparison)

/features comprehensive tour (9 categories, 38 features with badges)

/compare competitive comparison (21 features vs Cursor/Copilot/Devin/Replit)

/trust Trust Center (security status table, 13 items)

MCP expanded: Jira, Confluence, Figma, Vercel, Supabase, Twilio (11 servers)

Decision Tree UI in run detail Thinking tab

Progress UI: elapsed time, ETA, gradient shimmer, clickable pipeline nodes

Money Mode — 8 revenue-generating templates (SaaS, E-Commerce, Marketplace, API Service, Lead Gen, Newsletter, Booking, AI Wrapper)

Digital Clone — AI learns your tech stack, coding patterns, and preferences over time

Agent Hierarchy — 7 roles (CEO, CTO, Backend, Frontend, QA, DevOps, Growth) with 3 organizational structures

AI Code Review — Submit PR URLs or diffs for structured review with risk assessment and rollback plans

Multi-Run Chaining — Automatic continuation with configurable triggers (on_success, on_failure, spec_pass)

Learning Engine — Error pattern tracking with automatic lesson injection into future agents

Performance Insights — Run metrics, model comparison, trend analysis, and optimization suggestions

Policy & Guardrails — Cost limits, runtime caps, tool allowlists, command blocklists, PR approval gates

Deploy Integration — Config generator for Vercel, Railway, Fly.io, Docker, and Render

SSH & Email Tools — Remote server management and SMTP email sending

Stripe Tools — Create products, prices, and checkout sessions from agent runs

Security & Trust Page — Data flow architecture, encryption details, compliance roadmap

Homepage — 15 feature cards + 5 flagship showcase sections in 4 languages

Documentation — Expanded from 7 to 21 sections with 47+ API endpoints and 26 agent tools

Dashboard — 6 new pages (Policies, Review, Performance, Money, Deploy + Onboarding Wizard)

Toast notifications replaced all browser alerts across 9 dashboard pages

Runs page — Search, status filtering, and pagination

Run detail — Wave pipeline visualization, Thinking tab, enhanced retry visibility

Privacy Policy upgraded to enterprise-grade (12 sections)

Benchmark renamed from "SWE-Bench Lite" to "CesaFlow Benchmark" (honest methodology)

BYOM messaging — accurate proxy architecture description

Brand fix — "Orkestra" → "CesaFlow" on invite page

OAuth buttons removed (were showing disabled "Coming soon")

Password policy standardized to 8 characters minimum

docs.cesaflow.ai 502 reference fixed