Infrastructure

Prax runs on a stack of containerized services: a Flask app, an optional Docker sandbox, a Playwright browser, and an optional observability pipeline.

Contents

• Sandbox — how to provide Prax a sandbox (local or remote), and how Prax uses it
• MCP Server — expose a curated, bearer-gated subset of Prax tools to other agents over the Model Context Protocol
• Content Publishing — Notes, courses, news — Hugo vs TeamWork API, editing, versioning
• Observability — Traces, metrics, Grafana LGTM stack, execution graphs
• Memory — Two-layer memory system (STM + LTM), vector store, knowledge graph, hybrid retrieval, decay, embedding providers
• Context Management — Token budgeting, tool result clearing, LLM compaction, hard truncation
• Health Monitoring — Telemetry events, anomaly detection, self-repair advisories, health API
• Docker — Docker Compose setup, dev mode, standalone deployment