Search the recipe library using natural language. Set source to knowledge (default) to query the knowledge hub, or graph to query the LightRAG graph index. Optionally filter results by chain or token.
Request payload for recipe search across graph-based or knowledge-base sources.
Search source: 'graph' for graph-based search, 'knowledge' for knowledge base
graph, knowledge Search mode: local, global, or mix
Entities (local/mix) or relationships (global/mix) count
Text chunks from vector search
Token budget for entity context
Token budget for relation context
Overall token ceiling; chunks get whatever remains after KG context
Max results for knowledge search
Filter recipes by chain symbol (e.g. 'ethereum', 'solana', 'bitcoin')
Filter recipes by token symbol (e.g. 'ETH', 'USDT', 'BTC')
List of keywords for filtering documents during keyword search
Search type: 'vector', 'keyword', or 'hybrid'
vector, keyword, hybrid Successful Response
Unified search response envelope.
Pagination metadata for list responses.
Supports both legacy offset-based and cursor-based pagination.
Cursor fields (has_more, after, before) are populated for
cursor-paginated endpoints. Legacy fields (offset, limit) are
populated when the caller uses the deprecated offset parameter.