Search across ALL indexed public sources in a single query.
Searches across:
include_huggingface_datasets=true)This endpoint performs a two-phase hybrid search:
Results are combined using Reciprocal Rank Fusion (RRF) for optimal ranking.
Use this when you want to discover relevant content without knowing which specific sources to search.
Performance: Typically returns results in 1-3 seconds for thousands of indexed sources.
API key must be provided in the Authorization header
Natural language search query
"How does authentication work in FastAPI?"
Total number of results to return
1 <= x <= 100Include repository sources in search
Include documentation sources in search
Include HuggingFace dataset sources in search (excluded by default to prevent search pollution)
Weight for vector search (1-alpha for BM25). Higher values favor semantic similarity.
0 <= x <= 1Use AI to compress results into a concise answer with citations
Maximum tokens in response. Results truncated when budget reached.
100 <= x <= 100000Source type boosts (override to customize or set {} to disable)
Programming languages to boost (e.g., ['python', 'typescript'])
Boost multiplier for preferred languages
1 <= x <= 5Use TurboPuffer FTS v2 native Sum/Product boosting
Search completed successfully
Number of source namespaces searched
Total query time in milliseconds
Any errors encountered during search
AI-compressed answer (only when compress_output=true)
Token budget metadata for content responses