ServeHandlers.hpp File Reference

Pure request → response functions for the serve subcommand. More...

#include "models/Model.hpp"
#include "stats/Metrics.hpp"
#include "utils/Types.hpp"
#include <nlohmann/json.hpp>
#include <cstddef>
#include <deque>
#include <map>
#include <mutex>
#include <optional>
#include <string>
#include <unordered_map>

Go to the source code of this file.

Classes
struct	ppforest2::cli::serve::LoadedModel
	Immutable model snapshot captured once at server startup. More...
struct	ppforest2::cli::serve::PredictionResult
	Result of running a prediction request — kept around for the predictions store so the server can serve cached HTML on shareable URLs. More...
class	ppforest2::cli::serve::PredictionStore
	In-memory bounded store of recent prediction results, keyed by opaque IDs that survive a page refresh. Capacity-limited LRU; evicts the oldest entry once capacity is reached. More...
struct	ppforest2::cli::serve::Response
	Status, body, MIME content type, and extra response headers. More...

Namespaces
namespace	ppforest2
	Binarization strategies for multiclass-to-binary reduction.
namespace	ppforest2::cli
	Command-line interface: argument parsing, subcommands, and benchmark/evaluation orchestration.
namespace	ppforest2::cli::serve

Functions
Response	ppforest2::cli::serve::handle_health (std::string const &version)
	GET /health → 200 with status + version.
Response	ppforest2::cli::serve::handle_predict (LoadedModel const &loaded, PredictionStore &store, std::string const &csv_body)
	POST /predict (Accept: application/json) → 200 with {predictions, [proportions], [metrics], id}.
Response	ppforest2::cli::serve::handle_predict_html (LoadedModel const &loaded, PredictionStore &store, std::string const &csv_body)
	POST /predict (Accept: text/html) → 200 with the predict page rendered against the new result. Content-Location header points to the canonical /predict?id=<id> URL so the browser can update its address bar and share/refresh works.
Response	ppforest2::cli::serve::handle_predict_view (LoadedModel const &loaded, PredictionStore const &store, std::string const &id_query)
	GET /predict → 200 with the predict page (always showing the upload form). When id_query refers to an entry still in the store, the results section is also populated.
Response	ppforest2::cli::serve::handle_summarize (LoadedModel const &loaded)
	GET / (Accept: application/json) → 200 with the loaded model JSON, minus the heavy model field.
Response	ppforest2::cli::serve::handle_summary_html (LoadedModel const &loaded)
	GET / (Accept: text/html) → 200 with a self-contained HTML page rendering meta, training metrics, and variable importance.

Detailed Description

Pure request → response functions for the serve subcommand.

Handlers are pure functions of (cached model snapshot, request body) so they can be unit-tested without bringing up an HTTP server.