Felipe97/llama-cpp-compiled

1#pragma once2 3#include "server-http.h"4 5#include <atomic>6#include <cstddef>7#include <functional>8#include <memory>9#include <string>10 11// streaming buffer for one generation, survives HTTP disconnect. the producer appends SSE bytes,12// readers drain from any offset via read_from. keyed by conversation_id, one conv = one live session13 14struct stream_session;15 16using stream_session_ptr = std::shared_ptr<stream_session>;17 18// base of the producer/consumer pipe ends. virtual dtor so each runs its own teardown:19// the producer finalizes the session, the consumer leaves it untouched20struct stream_pipe {21    virtual ~stream_pipe() = default;22 23    bool is_cancelled() const;24 25protected:26    explicit stream_pipe(stream_session_ptr session);27 28    stream_session_ptr session_;29};30 31// producer end: writes chunks into the ring buffer and owns the session lifetime, finalizing it32// on destruction.33struct stream_pipe_producer : stream_pipe {34    ~stream_pipe_producer() override;35 36    bool write(const char * data, size_t len);37 38    static stream_pipe_producer * create(stream_session_ptr session);39 40private:41    explicit stream_pipe_producer(stream_session_ptr session);42};43 44void server_stream_session_manager_start();45void server_stream_session_manager_stop();46 47// route handler factories wired under /v1/stream/* by server.cpp48// child-side handlers for the resumable stream routes. the conv id travels in the conv_id49// query string because it can embed a model name containing slashes (org/repo), which the50// decoded path would split before the param is captured51server_http_context::handler_t server_stream_make_get_handler();52// POST /v1/streams/lookup with body {"conversation_ids": [...]}: only answers for ids the53// caller already owns (the WebUI passes the convs visible in its sidebar), the server never54// lists ids it has not been asked about, so a random caller cannot enumerate live sessions55server_http_context::handler_t server_stream_make_lookup_handler();56server_http_context::handler_t server_stream_make_delete_handler();57 58// extract the X-Conversation-Id header value (case-insensitive), empty when absent59std::string server_stream_conv_id_from_headers(const std::map<std::string, std::string> & headers);60 61// implement tee-style pipe (spipe) for "stream replay" functionality62struct server_res_spipe : server_http_res {63private:64    // if set, the stream survives a client disconnect:65    // connection kept alive, output is forwarded to spipe and reuse later66    std::unique_ptr<stream_pipe_producer> spipe;67    // if spipe is set, use this next_orig to implement tee-style pipe68    std::function<bool(std::string &)> next_orig;69    const server_http_req * req = nullptr;70    // set once next_orig reports no more data, so on_complete() doesn't re-drain a finished stream71    bool next_finished = false;72 73public:74    void set_req(const server_http_req * req);75    bool conn_alive();76    bool should_stop();77    void on_complete() override;78    void set_next(std::function<bool(std::string &)> next_fn);79};80