significant-gravitas/automoderate

AutoModerate is a web application for reviewing and filtering user-submitted content automatically. It is aimed at teams or products that need to decide whether text (such as comments, posts, or messages) should be approved, rejected, or held for a human to look at, without manually reading every submission. The platform is organized around projects. Each project can have its own team members, its own set of rules, and its own API keys for outside applications to connect to. When an application sends content to AutoModerate through the API, the system runs it through a series of checks in priority order. Simple keyword or pattern-based rules are checked first because they are fast. If those do not produce a clear answer, the content is sent to an AI model (OpenAI's GPT) to analyze it according to a custom prompt. The first rule that matches determines the final decision: approve, reject, or flag. Results with low confidence are automatically set aside for a human reviewer to examine. Results appear in the web interface in real time, meaning reviewers do not need to refresh the page to see new decisions come in. The interface also includes dashboards with statistics, a queue for manual review, and pages for managing team members and API keys. The application is built with Flask, a Python web framework. It uses a database to store content and decisions, supports SQLite for local development and PostgreSQL for production deployments, and includes Docker configuration to make deployment straightforward. It caches AI responses so repeated submissions of the same content do not trigger a new API call each time. The README is thorough and covers installation, configuration options, API usage examples, database structure, performance characteristics, and deployment guidance.