Build an AI Moderation Classifier in Python

Short summary

A Python Flask tutorial demonstrating a two-stage AI moderation pipeline using Telnyx AI Inference. Stage one uses embeddings similarity against a known-bad blocklist for obvious spam; stage two sends ambiguous content to an LLM for contextual judgment. The example includes batch moderation, structured response fields, and production recommendations like audit logs and human review queues.

• •Two-stage moderation: embeddings pre-filter for known-bad content, LLM judgment for ambiguous cases
• •Flask app with endpoints for indexing blocklists, single and batch moderation, stats, and health checks
• •Default models: Kimi-K2.6 for chat completions, gte-large for embeddings via Telnyx AI Inference

Generated with AI, which can make mistakes.