WAI Quantum job service

Whether two programs are the same circuit up to a global phase, and how the engine knows: Clifford circuits by their tableaus at any size, others on every basis state to 8 qubits, then by checked rewrites of one followed by the other's inverse to nothing (a proof at any width), then on random product states to 22 (evidence, not proof), otherwise unchecked with the reason. Its `verdict` is that decision; `verification` is the level it reached, as every tool's verdict states it. A check that would take long here is refused with the command that runs it locally.

Every exact classical route to simulating a program, priced on it: the stabilizer tableau, the cut-rank sum, an exact matrix product state at a rigorous bond bound, the light cone of a one-qubit expectation value, and the state vector. The cheapest is named for each question. A cheap route proves the program classically easy; none being cheap proves nothing about hardness. A thorough search, with a contraction tree, runs locally: wai-quantum shortcuts --thorough.

Compile an OpenQASM program through named passes (lowering, checked rewrites, phase folding, one- and two-qubit resynthesis, and every rotation off the multiples of π/4 written as a Clifford+T word within its share of the ε budget), each checked where it acts. Returns the compiled program and a certificate sealed with the service's key: each pass's counts before and after, its check and verdict (proof, exact or approximate with its distance), the budget and the bound the distances add to. verify_quantum_receipt checks it. Quote the bound; an approximate compile is not exact. A large or very fine compile is refused with the local command.

A program's measures, computed by the engine: gates by kind, two-qubit gates, depth and two-qubit depth, T count, whether it is a Clifford circuit, and a fidelity under reference error rates (a model, not a run). Also its smells, each with its lines: gates that cancel, rotations that merge or do nothing, idle qubits, and qubits acted on but never measured.

What the service would do with a job or a set, running nothing: the method and why, the memory, the estimated time and energy (each with its basis: this service's own timings and readings where it has them, otherwise the planner's nominal rate, and no energy figure without readings), the bounds that apply, where you stand against your limits, and whether the service would take it now, or the refusal it would give, with its code. Takes what run_quantum_circuit or run_quantum_set takes. It counts as one submission.

What a fault-tolerant quantum computer would need to run a program. First its logical counts: T gates, Toffolis, rotations (each costed at the T count of its own synthesised Clifford+T word, within its share of the budget), measurements, T depth and total depth, and logical qubits with the ancillas that gates under several controls need. Then, on a physical qubit, the planar model's code distance (the gate-based surface code's model fitted to this engine's own decoders), T factories, physical qubits and runtime, and the frontier of estimates trading qubits for time. Takes a program, or a logical-counts or routine document. Every estimate is sealed in the resource-estimate receipt family, binding the program's digest and the ε budget; the answer's notes state the model's assumptions. Quote the figures with the qubit and the budget, and call them an estimate. Sealed with the service's key; verify_quantum_receipt checks the seals and re-runs it. A program with more than 256 distinct rotation angles, or a budget too fine for its rotations, is refused with the local command.

The program with its smells fixed in its own text (all of them, and each alone), each candidate with its measures and the equivalence check's verdict against the program. A candidate the check finds different is never returned.

Every code the OpenQASM reader gives (each refusal and warning carries one, as `[Q103]` at the end of its message): its outcome (refused, lowered, carried, preserved or warned) and its meaning; and beside them, under job_codes, the codes of what is asked of a program (`[J011]`: the request, the method, an intent, a repair, the host). A code never changes meaning once released. Give `code` for one of either.

Show an error-correcting code and certify its distance: its checks (Z-type: they see X errors) and stabilizers (X-type), k counted by rank, a layout to draw it, the decoders it admits, and its distance certificate. In each sector the certificate gives the least weight of a logical found, a witness of that weight, and every lighter support ruled out by an exhaustive search over connected supports (a bivariate-bicycle code's translations checked as symmetries before use), re-checkable node for node. Give `certificate` to check one: its witnesses and the search run again. Sealed with the service's key. Quote the certificate's distance and its reading; a search stopped by `bound` settles nothing past it.

Decode with the engine's decoders. At code capacity, give `code` and a `syndrome` (the violated checks) or an `error` (the flipped qubits: its syndrome is taken and its residual judged, logical or not): matching (exact) and union_find for codes whose qubits sit in at most two checks; color for colour codes (restriction to matching on the three restricted lattices, lifted at each colour, completed below half the distance by an exact search); relay_bp for any code (it can stall on a degenerate error of a topological code, and then says it did not converge); search for the lightest error with the syndrome up to max_weight. At circuit level, give `memory` ({distance, rounds, p}: the rotated surface-code memory experiment), `circuit` (the circuit text: tags, Pauli-target observables and coordinates read) or `model` (detector error model text, decompositions with ^ read), with the fired `detectors` (the prediction: which observables flipped) or `shots` (sampled and decoded: the failure rate and its standard error): matching (exact), correlated (exact matching twice through the model's decomposed faults, the second answer kept only when the model's mechanisms explain it more cheaply), union_find, bp_osd, or window (a sliding window: `window` {commit, buffer} rounds; every edge set below half the distance is corrected once the buffer is at least (d + 1)/2). Sealed with the service's key.

Measure logical failure rates over distances and physical error rates: family surface at circuit level (decoders matching, correlated or union_find; the code model P(d) = a·(p/p*)^((d+1)/2) per round fitted to the points), or color, toric or surface_capacity at code capacity. Each point gives its shots, failures, rate and standard error; below_threshold says, at each p, whether every larger distance failed less. Sealed with the service's key. Report rates with their errors: a crossing measured on few shots is not a threshold.

Worked examples, each a complete job (OpenQASM program and arguments) with what it shows: a Bell pair, a rotation, QAOA, noise, a 1000-qubit GHZ state, 100-qubit dynamics, 200 qubits as a matrix product state. Pass an example's job to run_quantum_circuit as it is, or adapt it.

The methods with their qubit limits and what each answers, the accuracy classes, the per-client limits, the current load, how jobs are kept and metered, the receipt keys, and the conventions (bit order, Pauli strings).

Repair a program the reader refuses, changing only the lines its diagnostics flag. The engine makes the fixes that are mechanical (a missing `;`, an undeclared register's size, a measurement with no target, an unbounded while given the bound chosen, an angle the bit-exact method refuses rounded to the bits chosen with its rounding stated), and holds a fix you propose (`proposed`) to the flagged lines: any change to another line is rejected (J021) and nothing is changed. Returns the repaired program, each repair with its code and every line it changed, what needs a person with the reader's advice, and the verdict.

Run an OpenQASM program as it would run on a device target: compiled exactly to its gates, placed on its most reliable qubits and routes (the expected success given), checked against the circuit, and simulated under the noise its calibration gives, on the density matrix or by trajectories. Counts and expectation values with their distance from the ideal distribution, the noise derived, and a receipt that binds the target's hash. A model of the device, not a run on it. Every angle must be a dyadic multiple of 2π. Waits up to 50 s, as run_quantum_circuit does.

Run an OpenQASM program and return measurement counts, Pauli expectation values and/or amplitudes. The service chooses the method (bit-exact or double-precision state vector up to 32 qubits, single precision on a GPU to 33, stabilizer tableau for Clifford circuits up to 32768 qubits, matrix product states up to 4096, sparse Pauli dynamics for expectation values up to 1024 with an error bound, an exact density matrix or trajectories for noise), reports why and how accurate, and signs the result. Give `intent` (an exact distribution, or a reference program it must equal or whose state it must prepare) and the engine checks it with the run. Every answer carries `verdict`: the level the engine checked to (proof, evidence, executed, parsed, refused), whether the intent holds and a plain statement, sealed in the receipt. Waits up to 50 s; a longer job answers with status running and an id for get_quantum_job.

Run several OpenQASM programs as one set (`programs`: each a program's text or a job), or one program over a grid of its inputs (`qasm` with `sweep`, as {"theta": [0.1, 0.2], "n": [2, 3]}: every combination; integer inputs written out, float and angle inputs bound to the double given). The set's other fields (shots, observables, seed, intent, …) apply to every item. Each item gets its own result and receipt, each stating its verdict, and one receipt over the set binds every item through a Merkle root; verify_quantum_receipt checks it whole. At most 100 items, run one after another in one job. Waits up to 50 s; a longer set answers with status running and an id for get_quantum_job.

Whether a result's receipt (an Ed25519 signature by the service) seals exactly that request and that result. Pass the job's own `request` (as run_quantum_circuit or get_quantum_job returns it: a run here asks for its verdict, so the request it seals says so) and its result.

A job's status, and its result when it is done. The service keeps every job across its restarts for its retention (a day on the public service); a job that was running when it stopped comes back interrupted, with the reason.

The device targets the service knows by name: synthetic, made up in the shapes of published formats, each with its qubits, pairs, the gates a program compiles to, the spread of its T1 and readout errors, and its hash. A job may give its own target document instead.

A device target's qubits, pairs, native gates with the spread of their errors and durations, its calibration snapshot and its hash; with view, each qubit's place for drawing its coupling map with its readout and one-qubit errors, and each pair's two-qubit error. Give a built-in target's name or a target document.