metrix_one – Multi-Tenant SaaS for AI Strategy Frameworks

THE IDEA

metrix_one – Multi-Tenant SaaS for AI Strategy Frameworks

Type: Software (multi-tenant SaaS, web platform) Status: Beta Tech Stack: Python 3.13, FastAPI, SQLAlchemy (async), Postgres 16 (row-level security), Alembic, Redis + RQ, MinIO (S3), Keycloak (OIDC/PKCE), Anthropic/OpenAI/Google/xAI/DeepSeek/Kimi/MiniMax/Ollama (LLM), Tavily (web search), WeasyPrint + Jinja2 (PDF), React 19, Vite 6, Tailwind CSS v4, TypeScript, Docker Compose Problem Statement Strategy work costs founders, small teams and consultants weeks of manual market research — and anyone offering it as a service needs a cleanly separated data space per client. A tool that only runs on one machine is not enough for that: what is needed is real tenant isolation, central user management and cost control through quotas. Description metrix_one turns terse inputs or a free-form founder brain dump into structured, bilingual (EN/DE) business strategy analyses across 18 frameworks by way of large language models — from market breakdown and offer design through Business Model Canvas, finance and risk register to Bullseye/Traction. Every framework produces validated JSON output, rendered as an interactive web view, a PDF and a Markdown export. A Council mode lets several AI models answer in parallel and synthesises them through a moderator; a synthetic market simulation tests a pitch against a panel of AI personas including word-of-mouth diffusion. Evidence (PDF/DOCX/XLSX) can be uploaded and is fed into generation as context through text extraction, optionally supplemented by Tavily web search. From a finished analysis, social posts, ad copy, blog outlines and a nine-slide pitch deck can be derived directly, supplemented by analysis-independent quick tools (repurposing, content calendar, hook bank). The platform is genuinely multi-tenant: data isolation is enforced at the Postgres level by row-level security, authentication runs through an external Keycloak, long jobs go to Redis/RQ workers, and costs are steered through token quotas and an optional bring-your-own-key. Not yet production-complete: there is no self-service signup (invite only) and no integrated payment handling. Use Case Founders, small teams and consultants produce well-founded, shareable strategy documents and marketing content in minutes, and can test a business idea against a simulated customer panel beforehand — instead of spending weeks on manual market research and strategy work. Link: https://github.com/rawk7000/metrix_one (private repo) The public landing page: strategy frameworks, multi-model AI and synthetic market feedback in one workspace. The framework library, grouped into Discover, Design, Validate and Activate. Pricing: the plan follows workspaces, model mix and actual usage. Contact: onboarding runs through a demo request rather than self-service signup. Signing in hands over to an external Keycloak — here the entry point.