features

Features overview

SlopSearX provides three cross-cutting feature areas that extend beyond the core architecture: engine implementations that connect to 12 different search backends, output formatters that map the internal SearchResult model to wire formats, and the adapter interface that makes new engines pluggable without orchestrator changes.

Feature areas

Feature	Description	Key files
Engine implementations	12 built-in adapters covering Brave, Wikipedia, DuckDuckGo, Google, GitHub, HuggingFace, Internet Archive, OpenAlex, Stack Exchange, arXiv, Hacker News, and Semantic Scholar	engines/*.py
Output formatters	Two output formats: SearXNG-compatible JSON with all 23 MainResult fields, and YAML+Markdown for AI agent consumption	slopsearx/formatter.py
Adapter interface	Abstract base classes EngineAdapter and ScrapeAdapter, @register_engine decorator, and global engine registry	slopsearx/adapter.py

Cross-cutting design principles

• Every engine is one file. Adding a new engine requires zero changes to the orchestrator. Each adapter is registered via @register_engine in its own file under engines/.
• Adapters never raise exceptions. All error states are classified into EngineStatus (OK, RATE_LIMITED, BLOCKED, ERROR, TIMEOUT) and returned in AdapterResponse.status. The orchestrator never sees an unhandled exception from any adapter.
• Internal schema is decoupled from wire format. The SearchResult dataclass is the internal model. Output formatters handle the conversion to SearXNG JSON or YAML+Markdown.
• Rate limiting is distributed. All engines share a Valkey-backed sliding window rate limiter that is correct even at 50+ replicas.

See also

• Core primitives — foundational data types used by all features
• System architecture — request flow and system design
• Glossary — project-specific terms