|
Continuum C++ API
Unified runtime for token + tensor execution
|
#include <semantic_cache.hpp>
Public Member Functions | |
| virtual | ~EmbeddingProvider ()=default |
| virtual std::vector< float > | embed (const std::string &text) const =0 |
Embed text; the result must have dimension() elements. | |
| virtual std::size_t | dimension () const =0 |
| virtual std::string | identity () const =0 |
Source of the vectors the semantic tier compares.
Implement it in C++ or subclass continuum._native.EmbeddingProvider in Python (a local model, a hosted endpoint, precomputed vectors, ...).
|
virtualdefault |
|
pure virtual |
Implemented in continuum::runtime::BruteForceEmbeddingProvider.
|
pure virtual |
Embed text; the result must have dimension() elements.
Implemented in continuum::runtime::BruteForceEmbeddingProvider.
|
pure virtual |
Stable identity of the embedding space (model name + version + any setting that changes vectors). It is part of the semantic cache key, so vectors from different embedders never compare against each other.
Implemented in continuum::runtime::BruteForceEmbeddingProvider.