Product / UX / Safety & Guardrails

Osprey

An open, high-performance safety rules engine for real-time abuse mitigation, originally built at Discord and donated to ROOST. It evaluates events and their properties in real time using a Python-based rules DSL (SML), ingesting signals — including ML classifier outputs — to automatically detect and action spam, abuse, and policy violations at scale. Trust & Safety operational infrastructure rather than an AI guardrail model.

Osprey safety rules engine, Apache-2.0, donated by Discord to ROOST; ~446 GitHub stars (July 2026). In production at Discord (400M daily actions, 2.3M rules/sec), Bluesky, and Matrix. Verified live July 2026.

Openness

5 high confidence

5.0

license: Apache-2.0(OSI)
source: public
self-host: yes
service: none
core-gated: ungated

Fully open source: Apache-2.0 real-time safety rules engine, self-hostable, no proprietary tier. Donated by Discord to ROOST.

https://github.com/roostorg/osprey recorded 2026-07-01
Apache-2.0; Rust coordinator + stateless Python workers; SML rules DSL

Adoption

3 medium confidence

3.0

446 GitHub stars / 52 forks (July 2026) is a modest community signal, but Osprey is in production at Discord (400M daily actions, 2.3M rules/sec), Bluesky, and Matrix. Adoption rests on marquee deployments rather than breadth; open-sourced under ROOST in Sep 2025.

https://roost.tools/blog/introducing-osprey-v1-0-open-source-infrastructure-for-real-time-abuse-mitigation/ recorded 2026-07-01
v1.0 release; production users (Discord, Bluesky, Matrix)

Capability

4 medium confidence

4.0

Battle-tested at Discord scale; single-purpose Trust & Safety rules engine, not an AI-native guardrail.

https://github.com/roostorg/osprey recorded 2026-07-01
Rust coordinator, stateless Python workers, SML DSL

Unchanged since 2026-08-01 (last edited, not re-checked)