Creator Showcase is a self-hosted, white-label platform for advertising managers and agencies to showcase, search and recommend creators to clients.
What started as a simple creator catalog gradually turned into a complete product that can be installed, configured and operated by a non-technical team without touching the code.
The platform includes a fully configurable creator showcase, Telegram Mini App and AI sales assistant.
Clients can search creators traditionally or describe what they need in natural language. Instead of letting an LLM invent results, the AI acts as an agent: it reads live catalog metadata, chooses appropriate search tools and turns the request into deterministic database queries.
If the AI provider fails or times out, the platform falls back to its own fuzzy-search and Levenshtein-based ranking, so basic discovery never depends entirely on an external model.
A separate AI assistant can qualify a client through conversation, understand the campaign, select real creators from the database and generate personalized pitches for each recommendation.
Almost every part of the system can be controlled from the admin panel:
creator profiles, platforms and metrics, categories, tags, pricing, media, visual settings, AI providers, models, prompts, Telegram flows, buttons, administrators, broadcasts, channel publishing and system behavior.
No source-code editing is required for normal operation.
The project also includes database management tools, initialization and maintenance utilities, migration helpers and additional scripts for repetitive administrative tasks.
Deployment was designed as part of the product.
The installation process automatically creates and configures the required isolated services — backend, Telegram bot, PostgreSQL database and reverse proxy — with persistent storage and internal networking.
The goal was to make deployment closer to installing software than manually building infrastructure.
The platform also supports multilingual versions with smart routing and redirects, allowing different localized deployments to share the same product logic without turning localization into a collection of duplicated sites.
I didn't want the public side to look like another generic SaaS dashboard.
The frontend uses custom distortion maps, SVG displacement filters, liquid-glass effects, animated transitions and interaction-heavy creator cards. A significant part of the project was spent making these effects responsive and usable rather than simply decorative.
Because the platform handles private business data and administrative access, security was treated as a system requirement.
It includes hashed credentials, JWT authentication, rate limiting, Telegram initData signature validation, generated secrets, restricted uploads, audit logging, isolated services and database networking, non-root containers and production-safe defaults.
FastAPI • PostgreSQL • SQLAlchemy • aiogram • Telegram Mini Apps • LLM tool calling • Groq • Gemini • OpenAI • OpenRouter • Docker Compose • Caddy • JavaScript • SVG
The entire system was designed so that AI handles reasoning, deterministic tools handle data, fallbacks keep important functionality alive, and operational complexity stays hidden from the end user.
I'm currently open to commercial licensing, white-label deployments, selling the source code, or acquisition of the project itself.
Built with