Search patterns by id, name, alias, or intent. Returns top matches.

Ranks substring matches by where they hit (id > name > alias > intent), case-insensitive, with the most specific match first.

Return the full body of a pattern by id.

List patterns, optionally filtered by category (e.g. 'tool-use-environment').

Reverse-index view of a pattern: who implements it (frameworks), who uses it (recipes, methodologies), and what anti-patterns oppose it.

Get code examples for a pattern, optionally filtered to one framework (e.g. 'langchain', 'dspy', 'claude-agent-sdk', 'pseudo').

Given an observed symptom (e.g. 'agent loops forever', 'tool calls reference functions that do not exist'), return likely anti-patterns and the positive patterns that fix them.

Note: until structured symptoms fields land on anti-patterns (catalog task #3), this falls back to keyword search across each anti-pattern's name, intent, and problem text.

List anti-patterns. The catalog's anti-patterns category holds all of them; the optional category argument lets a caller pass a topical keyword (matched against name/intent) for narrowing.

Return a recipe (abstract composition) by id.

Return a framework (concrete composition) by id.

List frameworks. Optional category filters by composition category (orchestration-framework, agent-sdk, coding-agent, voice-conversational, ...).

Return a methodology entry by id.

Recommend recipes and frameworks for a use case + constraints.

Heuristic v1: maps use_case to a composition category (coding-agent, rag, voice-conversational, browser-computer-use, research-agent, agent-sdk, orchestration-framework, conversational-bot) and returns top recipes + a few frameworks. A proper facet-based recommender is catalog task #7.