Generate a CSS gradient from sorted action entries

GPU postsolve fast paths. Both reuse the device buffers already populated by iterate()+finalize() (averaged strategy lives in current_strategy after f

Sum of all host-side estimates (matchup + CPU state + strategy-tree EV cache + JSON). We exclude GPU state — that lives in VRAM.

One attempt at running the solver subprocess. Used by run_solver for the initial call and for the CPU fallback retry.

--------------------------------------------------------------------------- Sanity check: baseline and candidates must agree to within FP rounding on

* Map an aggression score [0, 1] to a cool→hot color gradient * (grey → orange → red). Distinct from the EV diverging palette so the * two modes don

Sum of aggressive-action frequencies in a combo strategy. Returns 0..1.

* Strip the raw solver output down to what the frontend needs to render * a SpotLibrary spot. PRESOLVE_BUNDLE_PLAN.md §6.2 BundledSpot schema.

Build a concrete 5-card hand from a rank-multiset (list of (rank, count)). Occurrences of a rank get suits 0,1,2,3; this never makes a 5-flush (a pair

* Build the canonical spot key used by `scripts/bulk-presolve.mjs` and * `scripts/compact-presolved.mjs`. Format: * `m<matchupIdx>_b<boardIdx>_<si

Bytes per call streamed through the kernel from each candidate's matrix.

Cancel the currently-running solve, if any. Returns true if a process was killed. Called by the `cancel_solve` Tauri command bound to the frontend Sto

Extract the bundled-folder portion of a chart id ("cash/6max_100bb" / * "mtt/vs_open_3b"). The chart library data update made `scenario_type` * se

Combine a chart's non-fold actions into a single PioSolver-style range * string for use as an IP/OOP default range. Frequencies sum across all * n

Recursively walk `gto_output/` collecting all .json files. Skips files whose name doesn't match the chart pattern (e.g. progress files, .range artifac

OOP-perspective per-combo EV at root, before the 1/total_ip_weight scaling applied by Solver::compute_combo_evs(). Empty on backends that don't suppor

Helper: compute one (matchup_ev, matchup_valid) pair for a specific board. `board_cards` is the FULL board to evaluate at this terminal (config flop +

Reference backend's only intra-iter parallelism is the OOP||IP `parallel sections` block, which fans out to at most 2 OMP threads regardless of `omp_g

Build a stable filename for the CSV, e.g. `runout_AsKsQs_Check_Bet_33_Call.csv`

Run the engine with `--gpu-info` to detect CUDA GPU availability. Short-lived subprocess (~100ms). Safe to call at startup and on demand.

* Map a normalized value in [0, 1] to a diverging red→grey→green color. * 0 = saturated red, 0.5 = neutral grey, 1 = saturated green. * Used for the

Trigger a CSV download in-browser. No Tauri filesystem needed since * output is sub-100KB.

Diagnostic log file location. Writes CLI args and engine stderr on each run so we can diagnose user-reported crashes without reproducing state.

* @brief Enumerate canonical next-card deals for the current board. * * This is the PioSolver / GTO+ style runout iso: given the current full board

============================================================================ v1.2.2: estimate_only — pre-solve cost preview without running iterations

Returns rough ops/second throughput for the named backend. - backend_label_lc is the lowercased backend name as it appears in the SolverResult.backend

Exchange Google ID token for DEEPFOLD JWT

Same but for a Record<string, ...> object literal (BET_SIZINGS).

Parse the C++ engine's stderr for a JSON error payload and extract `.message`. Returns None if no error JSON found.

Search the usual binary locations. Returns None when the engine hasn't been built yet — the test then prints a skip notice instead of failing CI on ma

Like `find_gto_root()` but additionally requires that the resolved `gto_output/` contains a `presolved/` subdirectory. Reason: in dev mode, Tauri may

Format seconds as M:SS

Pick two specific cards for the hero combo display

Get star count (1-5) from percentage score

Get the top N actions from a global strategy, sorted by percentage descending

Trigger Google popup sign-in in a real browser (dev mode only)

* Run Google OAuth via the system browser. Returns the Google id_token once * the user completes sign-in. Resolves/rejects based on Tauri event liste