Continuum C++ API
Unified runtime for token + tensor execution
Loading...
Searching...
No Matches
continuum::runtime::Session Class Reference

#include <session.hpp>

Public Member Functions

 Session (const std::string &id, backend::BackendRegistry &backends, std::size_t max_cache_entries=8192)
 
 Session (const std::string &id, backend::BackendRegistry &backends, KVCacheIndex &external_cache)
 
std::vector< continuum::Value > run (const ir::Graph &g, const std::unordered_map< ir::NodeId, continuum::Value > &inputs)
 
const ReusePolicy & policy () const
 
void set_policy (const ReusePolicy &policy)
 
void set_memo_table (MemoTable *memo)
 
void set_semantic_cache (SemanticCacheIndex *sc)
 
void set_embedding_provider (EmbeddingProvider *ep)
 
void set_layer_cache (LayerKVCacheIndex *lc)
 
void set_memory_graph (MemoryGraphStore *mg)
 
void set_cache_namespace (std::string ns)
 
const std::string & cache_namespace () const
 
void set_observer (ReuseObserver *observer)
 
ReuseObserver * observer () const
 
const ReuseMetrics & metrics () const
 
void reset_metrics ()
 
KVCacheIndex & cache ()
 
const KVCacheIndex & cache () const
 
MemoTable * memo_table () const
 
SemanticCacheIndex * semantic_cache () const
 
std::vector< CacheTierStats > cache_stats () const
 Occupancy of every attached tier; detached tiers are omitted.
 
const std::string & id () const
 
std::int64_t run_count () const
 
std::size_t cache_size () const
 
bool save_cache_metadata (const std::string &path) const
 
bool load_cache_metadata (const std::string &path)
 

Constructor & Destructor Documentation

◆ Session() [1/2]

continuum::runtime::Session::Session ( const std::string &  id,
backend::BackendRegistry &  backends,
std::size_t  max_cache_entries = 8192 
)
explicit

◆ Session() [2/2]

continuum::runtime::Session::Session ( const std::string &  id,
backend::BackendRegistry &  backends,
KVCacheIndex &  external_cache 
)
explicit

Member Function Documentation

◆ cache() [1/2]

KVCacheIndex & continuum::runtime::Session::cache ( )
inline

◆ cache() [2/2]

const KVCacheIndex & continuum::runtime::Session::cache ( ) const
inline

◆ cache_namespace()

const std::string & continuum::runtime::Session::cache_namespace ( ) const
inline

◆ cache_size()

std::size_t continuum::runtime::Session::cache_size ( ) const
inline

◆ cache_stats()

std::vector< CacheTierStats > continuum::runtime::Session::cache_stats ( ) const

Occupancy of every attached tier; detached tiers are omitted.

◆ id()

const std::string & continuum::runtime::Session::id ( ) const
inline

◆ load_cache_metadata()

bool continuum::runtime::Session::load_cache_metadata ( const std::string &  path)

◆ memo_table()

MemoTable * continuum::runtime::Session::memo_table ( ) const
inline

◆ metrics()

const ReuseMetrics & continuum::runtime::Session::metrics ( ) const
inline

◆ observer()

ReuseObserver * continuum::runtime::Session::observer ( ) const
inline

◆ policy()

const ReusePolicy & continuum::runtime::Session::policy ( ) const
inline

◆ reset_metrics()

void continuum::runtime::Session::reset_metrics ( )
inline

◆ run()

std::vector< continuum::Value > continuum::runtime::Session::run ( const ir::Graph &  g,
const std::unordered_map< ir::NodeId, continuum::Value > &  inputs 
)

◆ run_count()

std::int64_t continuum::runtime::Session::run_count ( ) const
inline

◆ save_cache_metadata()

bool continuum::runtime::Session::save_cache_metadata ( const std::string &  path) const

◆ semantic_cache()

SemanticCacheIndex * continuum::runtime::Session::semantic_cache ( ) const
inline

◆ set_cache_namespace()

void continuum::runtime::Session::set_cache_namespace ( std::string  ns)
inline

◆ set_embedding_provider()

void continuum::runtime::Session::set_embedding_provider ( EmbeddingProvider *  ep)
inline

◆ set_layer_cache()

void continuum::runtime::Session::set_layer_cache ( LayerKVCacheIndex *  lc)
inline

◆ set_memo_table()

void continuum::runtime::Session::set_memo_table ( MemoTable *  memo)
inline

◆ set_memory_graph()

void continuum::runtime::Session::set_memory_graph ( MemoryGraphStore *  mg)
inline

◆ set_observer()

void continuum::runtime::Session::set_observer ( ReuseObserver *  observer)
inline

Forward every tier-lookup and node-execution ReuseEvent to observer (nullptr, the default, disables emission).

◆ set_policy()

void continuum::runtime::Session::set_policy ( const ReusePolicy &  policy)
inline

◆ set_semantic_cache()

void continuum::runtime::Session::set_semantic_cache ( SemanticCacheIndex *  sc)
inline

The documentation for this class was generated from the following file: