analyze_pure_state_entanglement
ChatGPTFree API analysis of exactly four complex amplitudes in 00,01,10,11 order for a normalized two-qubit pure state. real is required; imag defaults to zero in the API. The API checks squared norm within 1e-9 and delegates concurrence, Schmidt coefficients and entropy to rqm-entanglement. No circuit, QASM, density matrix, tomography, account selector or options. Preserves API provenance and limitations. Mathematical analysis only; not hardware measurement, experimental certification or quantum advantage evidence. No provider/simulator execution, paid job, saved artifact or account mutation. Shared API limits apply; 15-second timeout, bounded response and no automatic retries. Invalid input is not silently repaired; unavailable analysis has no heuristic fallback.
analyze_quantum_coupling
ChatGPTAnalyze ideal circuit coupling via the existing free public API. Supply legacy CircuitIR 0.2 with 1–8 qubits and at most 512 unitary operations; fixed gate arities, distinct wires and finite rotation angles within +/-1e6 radians. No measurements, symbolic parameters, account selectors or state/options overrides. Starts from the API's default ideal input state. Qualitative fallback is retained. Newer entanglement responses preserve mode, provenance, nullable is_entangled and limitations together; older API responses omit all four and remain unchanged. Missing evidence or null is_entangled is unknown, not false; never infer measured state from gate presence or separability from an empty pair list. Ideal mathematical or qualitative gate analysis only. API 'measured' means computed ideal metrics, not physical hardware measurements. Pair structure and a single input-state fidelity are not circuit equivalence, certification, or quantum advantage evidence. No paid job, saved artifact or account mutation. Shared API limits apply; no automatic retries.
analyze_quantum_entanglement
ChatGPTAnalyze ideal circuit entanglement via the existing free public API. Supply legacy CircuitIR 0.2 with 1–8 qubits and at most 512 unitary operations; fixed gate arities, distinct wires and finite rotation angles within +/-1e6 radians. No measurements, symbolic parameters, account selectors or state/options overrides. Starts from the API's default ideal input state. Qualitative fallback is retained. Newer entanglement responses preserve mode, provenance, nullable is_entangled and limitations together; older API responses omit all four and remain unchanged. Missing evidence or null is_entangled is unknown, not false; never infer measured state from gate presence or separability from an empty pair list. Ideal mathematical or qualitative gate analysis only. API 'measured' means computed ideal metrics, not physical hardware measurements. Pair structure and a single input-state fidelity are not circuit equivalence, certification, or quantum advantage evidence. No paid job, saved artifact or account mutation. Shared API limits apply; no automatic retries.
approve_variational_budget
ChatGPTRefund eligibility follows original frozen customer_terms; failure or cancellation alone does not guarantee a refund. Report completion only from the original receipt/settlement evidence. Bounded hosted hardware QAOA MaxCut or product-Pauli VQE, 1–8 qubits, 1–4 layers, 100 shots per separately billable measurement group. Planning is not execution or spending approval. Preserve the user's exact device. Plan the immutable workflow, review all groups and required_maximum_children, then obtain a fresh quote_hardware_execution for the exact next circuit and device. quote_variational_budget binds that first quote, child_maximum_charge, workflow hash, cumulative USD ceiling and child count; the cap must cover every allowed child at that frozen ceiling. Show all terms and obtain explicit customer approval before approve_variational_budget. Each submit_variational_child requires its fresh exact-device quote and original next_iteration_key; Core checks the immutable circuit, target, terms, child ceiling, complete-cohort capacity and saved admission gates. Never change the budget, substitute devices, recycle spent authority or create replacement children to recover an uncertain response. Use get_variational_budget and get_hardware_execution for original-child recovery; bind execution IDs in group order, then recover_hardware_variational_workflow to checkpoint verified evidence and the next trial. Pending children or billing are not completed/settled. Stopping prevents new trials but does not cancel accepted children or promise refunds; cancel_hardware_execution is separate. No convergence, ground-state accuracy or quantum advantage is promised. Wallet-backed derived scientific workflows remain unavailable until their child-delivery path is qualified; ordinary verified child results use the original quote and delivery receipt. Deployment and account/device eligibility are required; unavailable server policy fails closed.
bind_hardware_variational_iteration
ChatGPTRefund eligibility follows original frozen customer_terms; failure or cancellation alone does not guarantee a refund. Report completion only from the original receipt/settlement evidence. Bounded hosted hardware QAOA MaxCut or product-Pauli VQE, 1–8 qubits, 1–4 layers, 100 shots per separately billable measurement group. Planning is not execution or spending approval. Preserve the user's exact device. Plan the immutable workflow, review all groups and required_maximum_children, then obtain a fresh quote_hardware_execution for the exact next circuit and device. quote_variational_budget binds that first quote, child_maximum_charge, workflow hash, cumulative USD ceiling and child count; the cap must cover every allowed child at that frozen ceiling. Show all terms and obtain explicit customer approval before approve_variational_budget. Each submit_variational_child requires its fresh exact-device quote and original next_iteration_key; Core checks the immutable circuit, target, terms, child ceiling, complete-cohort capacity and saved admission gates. Never change the budget, substitute devices, recycle spent authority or create replacement children to recover an uncertain response. Use get_variational_budget and get_hardware_execution for original-child recovery; bind execution IDs in group order, then recover_hardware_variational_workflow to checkpoint verified evidence and the next trial. Pending children or billing are not completed/settled. Stopping prevents new trials but does not cancel accepted children or promise refunds; cancel_hardware_execution is separate. No convergence, ground-state accuracy or quantum advantage is promised. Wallet-backed derived scientific workflows remain unavailable until their child-delivery path is qualified; ordinary verified child results use the original quote and delivery receipt. Deployment and account/device eligibility are required; unavailable server policy fails closed.
cancel_hardware_execution
ChatGPTWhen the selected QPU's execution window is closed or its wait does not meet the user's needs, proactively offer a fresh comparison of eligible alternatives for the same circuit and shots. ONLINE does not mean the execution window is open now. Highlight a currently open alternative only with fresh, timestamped provider status, execution-window and live-queue evidence for that exact device; current account/target eligibility, physical-location and workload requirements still apply. For catalog availability, require freshness=fresh, observed_at and age_seconds within max_age_seconds; otherwise treat current availability as unknown. Read window.open_now separately from status and never interpret next_open_at as the task start time. Show the active or next window in UTC and the user's timezone, queue depth and observation time/age. An open window or a shorter queue does not establish a faster finish; disclose unknown start/completion ETA and never estimate runtime from queue depth. Missing, null or stale availability evidence cannot establish an available-now or faster alternative; say live availability is unavailable rather than guessing from catalog generation/review times or ONLINE. Compare itemized provider estimate, RQM fee and fixed customer total for each offered option, marking unquoted totals unknown. Read current available wallet funds and disclose any funding gap; reserved funds are not available to fund another task. Ask for explicit device selection, then obtain a fresh quote and fresh approval of its exact charge before any new submission; comparing alternatives grants no spending or funding approval. For an already queued execution, keep monitoring the original ID/task and leave it untouched: never auto-cancel, replace, resubmit or reuse its approval for another device. dispatch_diagnostics contains historical observations about individual admission requests or worker invocations, not the execution's final outcome. api_admission_refused records a verified application refusal; api_transport_ambiguous records an unconfirmed API delivery; worker_post_claim_not_dispatched records one claimed worker stopping before SDK create. Later verified task/result evidence remains authoritative. These observations never authorize resubmission or hold release and never imply zero provider cost. For customer quotes, copy target_snapshot_id and target_review_sha256 from the selected device customer_quote in a fresh list_quantum_hardware result. For direct quotes and planning, pass the selected device_arn as target_device_arn. For guided quoting, pass it as selected_device_arn and plan_request.target_device_arn; pass target_snapshot_id and target_review_sha256 at the top level. Preserve these values unchanged; renewable catalog generations can advance during a conversation. Never invent or substitute a review digest. An expired review requires a fresh catalog and quote, followed by fresh spending approval. Report expires_at exactly; never guess that a quote expired from perceived conversational delay. Compare with a trustworthy current timestamp, or let Core enforce expiry on an explicitly approved submission. Never bypass expiry or alter approval bindings. Quotes carrying quote_only=true are price quotes only: submission/reservation/dispatch are unavailable while the security stop remains. They cannot be used to submit later; obtain a fresh execution-enabled quote after separate activation. For these quotes lifetime_ceiling_usd=0 grants no provider-spending authority. For legacy execution quotes, lifetime_ceiling_usd is cumulative provider-exposure authority, separate from the customer charge and any overall infrastructure/accounting cap. Issuing a quote does not reserve funds. Customer quotes are capped at $100 total per job, including the service fee and rounding. This charge limit does not cap the account balance. For gate_circuit_100_v1 and gate_circuit_v1, target_review.native_action_sha256 is the legacy reference action; quote.native_action_sha256 binds this selected circuit and must match its frozen prepared artifact. The…
compare_quantum_coupling
ChatGPTCompare ideal circuit coupling via the existing free public API. Supply legacy CircuitIR 0.2 with 1–8 qubits and at most 512 unitary operations; fixed gate arities, distinct wires and finite rotation angles within +/-1e6 radians. No measurements, symbolic parameters, account selectors or state/options overrides. Starts from the API's default ideal input state. Qualitative fallback is retained. Newer entanglement responses preserve mode, provenance, nullable is_entangled and limitations together; older API responses omit all four and remain unchanged. Missing evidence or null is_entangled is unknown, not false; never infer measured state from gate presence or separability from an empty pair list. Ideal mathematical or qualitative gate analysis only. API 'measured' means computed ideal metrics, not physical hardware measurements. Pair structure and a single input-state fidelity are not circuit equivalence, certification, or quantum advantage evidence. No paid job, saved artifact or account mutation. Shared API limits apply; no automatic retries.
describe_hardware_execution
ChatGPTRead the signed-in RQM account's configured default Braket device and hardware readiness. Takes no arguments. Returns the device ARN and applicable quote, quote-only, and execution gates. Use before planning or pricing a hardware experiment. Readiness does not confirm circuit compatibility, available funds, a binding quote, or permission to spend. Use list_quantum_hardware for current execution windows, queues, and eligible alternatives. Verify a refund only from the original receipt or settlement evidence. This read does not create a quote, reserve funds, acknowledge a result, or run a task.
describe_quantum_capabilities
ChatGPTFree, anonymous API-owned scientific capability contracts and domain suggestions. Start here before choosing or funding a quantum job. Inspect separate source, installed, hosted, simulator, live-QPU and public coverage; unknown means unverified. Read schemas, examples and limits; unavailable chemistry, cryptography or hybrid workflows must not be replaced with Bell or promised as execution. This finite registry is not the complete roadmap. Workflow tool references do not promise availability on this endpoint or eligibility for the caller. No account, balance, job, private policy, quote, reservation or hardware execution is accessed. Existing authenticated tools own those operations.
explain_quantum_circuit
ChatGPTExplain supplied OpenQASM 3 using the existing free, public, non-persistent API. Returns summary, standard and technical explanations of standard-compatible quaternion/SU(2), Bloch and supported two-qubit geometry with limits and provenance. Up to 65536 UTF-8 source bytes, 8 qubits, 64 classical bits and 512 operations; API parser enforces circuit limits. Partial reports remain partial. No account, funding, purchase, saved artifact, simulator sample or hardware run is created. Rate limits apply; do not automatically retry. Do not use as certification or quantum advantage evidence.
find_learning_resources
ChatGPTWhen users ask where or how to learn, find relevant reviewed public Storefront materials. Use short topic keywords and optional difficulty. Suggest one to three matching links with prerequisites and evidence category; keep standard quantum mechanics, mathematical background, RQM methods and research distinct. The catalog is intentionally small: do not invent results or treat an empty match as absence of material. Summaries are discovery metadata, not grounded lesson excerpts. Keep Storefront as the classroom; use learn_quantum_circuit for beginner circuit demonstrations. Free, read-only, no account, no paid jobs or hardware.
get_account_balance
ChatGPTRead the connected account's existing prepaid USD balance. Use this to check available funds without starting a job or changing the account. A positive balance does not establish saved spending-policy readiness or permission for a particular purchase. Does not add funds, reserve credits, purchase services or authorize spending.
get_account_job
ChatGPTRead this agent's saved account-funded job by its returned job ID to check status and recover available result and receipt. Use after submission or when following up on an existing purchase; acceptance is not completion. Report the saved state and only available results, with their limitations. If the job ID was lost, use list_account_jobs with the original idempotency key first. Never authorizes another purchase.
get_hardware_execution
ChatGPTFetch the saved status and verified result, when available, for one existing RQM Braket execution by execution_id. Use the original ID to check progress or retrieve the result. Not read-only: if a verified result is available for a reserved prepaid execution, this call acknowledges its delivery and can settle the original frozen customer charge exactly once. If no result is available, it reports the saved state without acknowledging delivery. A result that fails verification is rejected. This call does not create a quote, submit a task, or replace the original execution. Use list_hardware_executions for a summary without delivery acknowledgment. For software circuit-assurance jobs, use list_account_jobs with the original idempotency key, then get_account_job. If a queued task's wait does not meet the user's needs, use list_quantum_hardware to compare eligible alternatives for the same circuit and shots. Never auto-cancel, replace, resubmit or reuse its approval. Report refund eligibility from the original frozen customer_terms and refund completion only from its receipt/settlement evidence. Failure or cancellation alone does not guarantee a refund. Only quotes explicitly carrying full_refund_when_verified_failed or full_refund_when_verified_cancelled promise a full refund for that verified condition; do not assume those terms for legacy or owner quotes. A promised refund is not a completed refund: report pending, held, charged, released or refunded amounts exactly as the original receipt records. Unknown outcomes retain holds; never infer zero cost or refund completion from a missing provider task.
get_hardware_receipt
ChatGPTFetch the saved billing and settlement state for one existing RQM Braket execution by execution_id. Use the original ID to check the customer charge, delivery settlement, and any separate provider cost evidence. Unreconciled provider cost is not a final invoice. Not read-only: if a verified result is available for a reserved prepaid execution, this call acknowledges its delivery and can settle the original frozen customer charge exactly once. If no result is available, it reports the saved state without acknowledging delivery. A result that fails verification is rejected. This call does not create a quote, submit a task, or replace the original execution. Use list_hardware_executions for a summary without delivery acknowledgment. For software circuit-assurance jobs, use list_account_jobs with the original idempotency key, then get_account_job. Report refund eligibility from the original frozen customer_terms and refund completion only from its receipt/settlement evidence. Failure or cancellation alone does not guarantee a refund. Only quotes explicitly carrying full_refund_when_verified_failed or full_refund_when_verified_cancelled promise a full refund for that verified condition; do not assume those terms for legacy or owner quotes. A promised refund is not a completed refund: report pending, held, charged, released or refunded amounts exactly as the original receipt records. Unknown outcomes retain holds; never infer zero cost or refund completion from a missing provider task.
get_hardware_variational_workflow
ChatGPTRead the saved checkpoint of an existing RQM Braket variational workflow by run_id. Returns the exact device, saved iteration and evaluation counts, workflow state, and bound child execution IDs. Use it to resume reviewing the original workflow. This account-scoped read does not poll the provider, advance or stop the workflow, submit a child, or settle charges. Saved state does not establish fresh hardware availability.
get_quantum_experiment
ChatGPTWhen the selected QPU's execution window is closed or its wait does not meet the user's needs, proactively offer a fresh comparison of eligible alternatives for the same circuit and shots. ONLINE does not mean the execution window is open now. Highlight a currently open alternative only with fresh, timestamped provider status, execution-window and live-queue evidence for that exact device; current account/target eligibility, physical-location and workload requirements still apply. For catalog availability, require freshness=fresh, observed_at and age_seconds within max_age_seconds; otherwise treat current availability as unknown. Read window.open_now separately from status and never interpret next_open_at as the task start time. Show the active or next window in UTC and the user's timezone, queue depth and observation time/age. An open window or a shorter queue does not establish a faster finish; disclose unknown start/completion ETA and never estimate runtime from queue depth. Missing, null or stale availability evidence cannot establish an available-now or faster alternative; say live availability is unavailable rather than guessing from catalog generation/review times or ONLINE. Compare itemized provider estimate, RQM fee and fixed customer total for each offered option, marking unquoted totals unknown. Read current available wallet funds and disclose any funding gap; reserved funds are not available to fund another task. Ask for explicit device selection, then obtain a fresh quote and fresh approval of its exact charge before any new submission; comparing alternatives grants no spending or funding approval. For an already queued execution, keep monitoring the original ID/task and leave it untouched: never auto-cancel, replace, resubmit or reuse its approval for another device. For hardware comparisons, present every listed device with its exact ARN, provider, region, physical-location evidence, observation times and unknowns. Listed means inventory only; quotable means the current customer_quote/review and quote gates permit a price review for the exact device, circuit and shots; executable additionally requires current execution admission, compatibility, account policy, funds and explicit quote approval. Use describe_hardware_execution to identify the configured default ARN. For that ARN, use its own quote_enabled (falling back to execution_enabled only when quote_enabled is absent) and quote_only flags. For non-default ARNs, read configured_target_readiness_by_device for that exact target's quote_enabled/quote_only; missing non-default readiness is not established. Never borrow another device's readiness. Report each state separately, including blockers; never infer availability from tool visibility, provider online status, a price, past runs or another device's readiness. Compare the same supplied circuit and shot count; disclose differing scope or unsupported inputs. Planning validates the API contract and installed translation offline, not live device compatibility: device_capability_validated=false remains unverified. Width, gates, measurements and shot bounds must fit the selected target's current review and provider capabilities; the worker checks live compatibility before dispatch. Analog hardware is not a gate-circuit substitute. Catalog or planning provider-price estimates are not binding customer quotes or final provider invoices. Show source and observation/expiry times, currency, shots and unknown fee/total fields; never invent a service fee or derive it from a spending ceiling. Only the returned Core quote binds the exact device, circuit, shots, itemized provider estimate, fee, fixed customer total, maximum charge, expiry and terms. quote_only=true cannot reserve funds or execute; later execution requires a fresh execution-enabled quote and fresh approval. Unknown or null queue times mean unknown, not zero wait or immediate availability. Historical elapsed time and polling intervals are not queue/runtime estimates or promised completion times. Absent API turnaround advice, present all choices without c…
get_settled_hardware_result
ChatGPTStrictly read-only recovery of an already-delivered, settled gate-circuit hardware result and its existing receipt. Use in a new conversation with the original execution_id; if it was lost, list_hardware_executions with held_only=false first. Core checks the authenticated account and original OAuth principal, the saved receipt, result artifact and zero held amount. Requires only jobs.read. Never acknowledges delivery, settles, refreshes, polls providers, quotes, reserves or submits. Pending, held, missing, unsupported or inconsistent results fail closed; do not fall back to get_hardware_execution, get_hardware_receipt, get_quantum_experiment or another settlement-capable route for a read-only request. Report the saved charge as historical, not a new charge, and keep unreconciled provider costs distinct from customer settlement. No new spending is authorized.
get_variational_budget
ChatGPTRefund eligibility follows original frozen customer_terms; failure or cancellation alone does not guarantee a refund. Report completion only from the original receipt/settlement evidence. Bounded hosted hardware QAOA MaxCut or product-Pauli VQE, 1–8 qubits, 1–4 layers, 100 shots per separately billable measurement group. Planning is not execution or spending approval. Preserve the user's exact device. Plan the immutable workflow, review all groups and required_maximum_children, then obtain a fresh quote_hardware_execution for the exact next circuit and device. quote_variational_budget binds that first quote, child_maximum_charge, workflow hash, cumulative USD ceiling and child count; the cap must cover every allowed child at that frozen ceiling. Show all terms and obtain explicit customer approval before approve_variational_budget. Each submit_variational_child requires its fresh exact-device quote and original next_iteration_key; Core checks the immutable circuit, target, terms, child ceiling, complete-cohort capacity and saved admission gates. Never change the budget, substitute devices, recycle spent authority or create replacement children to recover an uncertain response. Use get_variational_budget and get_hardware_execution for original-child recovery; bind execution IDs in group order, then recover_hardware_variational_workflow to checkpoint verified evidence and the next trial. Pending children or billing are not completed/settled. Stopping prevents new trials but does not cancel accepted children or promise refunds; cancel_hardware_execution is separate. No convergence, ground-state accuracy or quantum advantage is promised. Wallet-backed derived scientific workflows remain unavailable until their child-delivery path is qualified; ordinary verified child results use the original quote and delivery receipt. Deployment and account/device eligibility are required; unavailable server policy fails closed.
learn_quantum_circuit
ChatGPTTeach beginners qubits, superposition, measurement, interference or entanglement through built-in OpenQASM 3 circuits. First omit predictions and ask the learner to predict the shown outcomes. Then compare their predictions with free ideal statevector probabilities computed without sampled shots. Describe what the demonstration shows and its limits. This tool does not accept arbitrary source, use an account, start a paid job, or run hardware. Keep established quantum mechanics separate from RQM-specific research interpretations.
list_account_jobs
ChatGPTList up to 100 recent jobs for this agent, or find the original account-funded purchase using its idempotency key. Use for purchase history and recovery before retrying an uncertain submission; use list_buyer_jobs or search_buyer_jobs for available services. Omit the key for recent history, or pass the exact original key to recover one purchase. Use a returned job ID with get_account_job for saved status, result and receipt. An empty response does not prove that a timed-out submission cannot still be accepted; preserve the original key and inputs on any retry. This read never purchases or reruns a job.
list_buyer_jobs
ChatGPTList supported quantum software services, their input requirements, evidence limits, readiness and separately exposed tool names. Free descriptive read; performs no computation, submission, quote or spending authorization.
list_hardware_executions
ChatGPTWhen the selected QPU's execution window is closed or its wait does not meet the user's needs, proactively offer a fresh comparison of eligible alternatives for the same circuit and shots. ONLINE does not mean the execution window is open now. Highlight a currently open alternative only with fresh, timestamped provider status, execution-window and live-queue evidence for that exact device; current account/target eligibility, physical-location and workload requirements still apply. For catalog availability, require freshness=fresh, observed_at and age_seconds within max_age_seconds; otherwise treat current availability as unknown. Read window.open_now separately from status and never interpret next_open_at as the task start time. Show the active or next window in UTC and the user's timezone, queue depth and observation time/age. An open window or a shorter queue does not establish a faster finish; disclose unknown start/completion ETA and never estimate runtime from queue depth. Missing, null or stale availability evidence cannot establish an available-now or faster alternative; say live availability is unavailable rather than guessing from catalog generation/review times or ONLINE. Compare itemized provider estimate, RQM fee and fixed customer total for each offered option, marking unquoted totals unknown. Read current available wallet funds and disclose any funding gap; reserved funds are not available to fund another task. Ask for explicit device selection, then obtain a fresh quote and fresh approval of its exact charge before any new submission; comparing alternatives grants no spending or funding approval. For an already queued execution, keep monitoring the original ID/task and leave it untouched: never auto-cancel, replace, resubmit or reuse its approval for another device. Report refund eligibility from the original frozen customer_terms and refund completion only from its receipt/settlement evidence. Failure or cancellation alone does not guarantee a refund. Only quotes explicitly carrying full_refund_when_verified_failed or full_refund_when_verified_cancelled promise a full refund for that verified condition; do not assume those terms for legacy or owner quotes. A promised refund is not a completed refund: report pending, held, charged, released or refunded amounts exactly as the original receipt records. Unknown outcomes retain holds; never infer zero cost or refund completion from a missing provider task. Recover an existing hardware execution without another paid submission. Core scopes every call to the authenticated account and principal; no account, principal, email or chat/session selector is accepted. A new chat or token for the original approved OAuth client retains its durable owner identity. Other approved clients have different pairwise subjects/principals and cannot recover these jobs without a separately reviewed owner-linking contract. Matching email, wallet or audience alone grants no access. Use list_hardware_executions when execution IDs or submission keys were lost. It defaults to held_only=true; use held_only=false for settled history. Follow next_cursor unchanged with the same held_only until null; pages are ordered by execution ID, not recency. held_total covers all this caller's held executions, not just the page. Listing does not poll providers, acknowledge results, release holds or submit work. Use a returned execution_id with get_hardware_execution or get_hardware_receipt for verified status, results and receipt. Those retrieval tools acknowledge a verified available result under its original frozen billing terms exactly once; listing's result_available is only a summary, not verified result evidence. If recovery.can_refresh is true, refresh_hardware_execution can reconcile original provider status without creating or dispatching a task. Unknown acceptance retains its hold until evidence resolves it; final_billing_required needs operator final-allocation evidence and cannot be cleared by resubmitting, canceling or switching payment rails. A missing job on one page or an empty list does…
list_quantum_hardware
ChatGPTWhen the selected QPU's execution window is closed or its wait does not meet the user's needs, proactively offer a fresh comparison of eligible alternatives for the same circuit and shots. ONLINE does not mean the execution window is open now. Highlight a currently open alternative only with fresh, timestamped provider status, execution-window and live-queue evidence for that exact device; current account/target eligibility, physical-location and workload requirements still apply. For catalog availability, require freshness=fresh, observed_at and age_seconds within max_age_seconds; otherwise treat current availability as unknown. Read window.open_now separately from status and never interpret next_open_at as the task start time. Show the active or next window in UTC and the user's timezone, queue depth and observation time/age. An open window or a shorter queue does not establish a faster finish; disclose unknown start/completion ETA and never estimate runtime from queue depth. Missing, null or stale availability evidence cannot establish an available-now or faster alternative; say live availability is unavailable rather than guessing from catalog generation/review times or ONLINE. Compare itemized provider estimate, RQM fee and fixed customer total for each offered option, marking unquoted totals unknown. Read current available wallet funds and disclose any funding gap; reserved funds are not available to fund another task. Ask for explicit device selection, then obtain a fresh quote and fresh approval of its exact charge before any new submission; comparing alternatives grants no spending or funding approval. For an already queued execution, keep monitoring the original ID/task and leave it untouched: never auto-cancel, replace, resubmit or reuse its approval for another device. For hardware comparisons, present every listed device with its exact ARN, provider, region, physical-location evidence, observation times and unknowns. Listed means inventory only; quotable means the current customer_quote/review and quote gates permit a price review for the exact device, circuit and shots; executable additionally requires current execution admission, compatibility, account policy, funds and explicit quote approval. Use describe_hardware_execution to identify the configured default ARN. For that ARN, use its own quote_enabled (falling back to execution_enabled only when quote_enabled is absent) and quote_only flags. For non-default ARNs, read configured_target_readiness_by_device for that exact target's quote_enabled/quote_only; missing non-default readiness is not established. Never borrow another device's readiness. Report each state separately, including blockers; never infer availability from tool visibility, provider online status, a price, past runs or another device's readiness. Compare the same supplied circuit and shot count; disclose differing scope or unsupported inputs. Planning validates the API contract and installed translation offline, not live device compatibility: device_capability_validated=false remains unverified. Width, gates, measurements and shot bounds must fit the selected target's current review and provider capabilities; the worker checks live compatibility before dispatch. Analog hardware is not a gate-circuit substitute. Catalog or planning provider-price estimates are not binding customer quotes or final provider invoices. Show source and observation/expiry times, currency, shots and unknown fee/total fields; never invent a service fee or derive it from a spending ceiling. Only the returned Core quote binds the exact device, circuit, shots, itemized provider estimate, fee, fixed customer total, maximum charge, expiry and terms. quote_only=true cannot reserve funds or execute; later execution requires a fresh execution-enabled quote and fresh approval. Unknown or null queue times mean unknown, not zero wait or immediate availability. Historical elapsed time and polling intervals are not queue/runtime estimates or promised completion times. Absent API turnaround advice, present all choices without c…
optimize_quantum_circuit
ChatGPTPurchase verified-circuit-optimization-v1 through existing Account Core authorization, saved spending policy and durable results. Select compiler or single_qubit_fusion (bounded Qiskit adjacent single-qubit fusion); returns the original unless preservation is verified. Inspect the service contract first. Caller must approve max_total_price in USD. Preserve the same idempotency key and exact request on retry; recover uncertain submissions with list_account_jobs, then get_account_job. Acceptance is not completion. This does not submit hardware or establish device compatibility.
plan_hardware_variational_workflow
ChatGPTRefund eligibility follows original frozen customer_terms; failure or cancellation alone does not guarantee a refund. Report completion only from the original receipt/settlement evidence. Bounded hosted hardware QAOA MaxCut or product-Pauli VQE, 1–8 qubits, 1–4 layers, 100 shots per separately billable measurement group. Planning is not execution or spending approval. Preserve the user's exact device. Plan the immutable workflow, review all groups and required_maximum_children, then obtain a fresh quote_hardware_execution for the exact next circuit and device. quote_variational_budget binds that first quote, child_maximum_charge, workflow hash, cumulative USD ceiling and child count; the cap must cover every allowed child at that frozen ceiling. Show all terms and obtain explicit customer approval before approve_variational_budget. Each submit_variational_child requires its fresh exact-device quote and original next_iteration_key; Core checks the immutable circuit, target, terms, child ceiling, complete-cohort capacity and saved admission gates. Never change the budget, substitute devices, recycle spent authority or create replacement children to recover an uncertain response. Use get_variational_budget and get_hardware_execution for original-child recovery; bind execution IDs in group order, then recover_hardware_variational_workflow to checkpoint verified evidence and the next trial. Pending children or billing are not completed/settled. Stopping prevents new trials but does not cancel accepted children or promise refunds; cancel_hardware_execution is separate. No convergence, ground-state accuracy or quantum advantage is promised. Wallet-backed derived scientific workflows remain unavailable until their child-delivery path is qualified; ordinary verified child results use the original quote and delivery receipt. Deployment and account/device eligibility are required; unavailable server policy fails closed.
plan_quantum_experiment
ChatGPTWhen the selected QPU's execution window is closed or its wait does not meet the user's needs, proactively offer a fresh comparison of eligible alternatives for the same circuit and shots. ONLINE does not mean the execution window is open now. Highlight a currently open alternative only with fresh, timestamped provider status, execution-window and live-queue evidence for that exact device; current account/target eligibility, physical-location and workload requirements still apply. For catalog availability, require freshness=fresh, observed_at and age_seconds within max_age_seconds; otherwise treat current availability as unknown. Read window.open_now separately from status and never interpret next_open_at as the task start time. Show the active or next window in UTC and the user's timezone, queue depth and observation time/age. An open window or a shorter queue does not establish a faster finish; disclose unknown start/completion ETA and never estimate runtime from queue depth. Missing, null or stale availability evidence cannot establish an available-now or faster alternative; say live availability is unavailable rather than guessing from catalog generation/review times or ONLINE. Compare itemized provider estimate, RQM fee and fixed customer total for each offered option, marking unquoted totals unknown. Read current available wallet funds and disclose any funding gap; reserved funds are not available to fund another task. Ask for explicit device selection, then obtain a fresh quote and fresh approval of its exact charge before any new submission; comparing alternatives grants no spending or funding approval. For an already queued execution, keep monitoring the original ID/task and leave it untouched: never auto-cancel, replace, resubmit or reuse its approval for another device. For hardware comparisons, present every listed device with its exact ARN, provider, region, physical-location evidence, observation times and unknowns. Listed means inventory only; quotable means the current customer_quote/review and quote gates permit a price review for the exact device, circuit and shots; executable additionally requires current execution admission, compatibility, account policy, funds and explicit quote approval. Use describe_hardware_execution to identify the configured default ARN. For that ARN, use its own quote_enabled (falling back to execution_enabled only when quote_enabled is absent) and quote_only flags. For non-default ARNs, read configured_target_readiness_by_device for that exact target's quote_enabled/quote_only; missing non-default readiness is not established. Never borrow another device's readiness. Report each state separately, including blockers; never infer availability from tool visibility, provider online status, a price, past runs or another device's readiness. Compare the same supplied circuit and shot count; disclose differing scope or unsupported inputs. Planning validates the API contract and installed translation offline, not live device compatibility: device_capability_validated=false remains unverified. Width, gates, measurements and shot bounds must fit the selected target's current review and provider capabilities; the worker checks live compatibility before dispatch. Analog hardware is not a gate-circuit substitute. Catalog or planning provider-price estimates are not binding customer quotes or final provider invoices. Show source and observation/expiry times, currency, shots and unknown fee/total fields; never invent a service fee or derive it from a spending ceiling. Only the returned Core quote binds the exact device, circuit, shots, itemized provider estimate, fee, fixed customer total, maximum charge, expiry and terms. quote_only=true cannot reserve funds or execute; later execution requires a fresh execution-enabled quote and fresh approval. Unknown or null queue times mean unknown, not zero wait or immediate availability. Historical elapsed time and polling intervals are not queue/runtime estimates or promised completion times. Absent API turnaround advice, present all choices without c…
quantum_circuit_assurance_report_v1
ChatGPTProblem: Validate, compile, and report bounded assurance evidence for OpenQASM 3. Input: JSON with source. Result: assurance outcome, warnings, normalized source and evidence. Limits: source UTF-8 bytes: 65536, verification qubits: 3. Deterministic bounded local software evidence; not hardware evidence, a formal proof, certification, signature, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_cross_provider_portability_v1
ChatGPTProblem: Translate this supported OpenQASM 3 circuit to the supplied local toolchain format without silent gate substitution. Input: JSON with source, target format. Result: translated target artifact, explicit gate mapping, source and artifact digests. Limits: required gates: 32, source UTF-8 bytes: 65536. Deterministic bounded local software evidence only; no QPU, provider acceptance, certification, formal-proof, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_execution_plan_select_v1
ChatGPTProblem: Select a supported execution plan against the supplied cost, depth, duration limits, and weights. Input: JSON with plans, weights, maximum cost usd, maximum depth, maximum duration seconds. Result: ranked feasible plans, objective scores, selected plan identity. Limits: maximum operations: 4096, request bytes: 65536. Deterministic bounded local software evidence only; no QPU, live-provider, physical, formal-proof, certification, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_failed_workflow_diagnosis_v1
ChatGPTProblem: Diagnose why this bounded quantum workflow failed and return supported corrective actions. Input: JSON with stage, error text. Result: failure class, evidence terms, bounded corrective actions. Limits: request bytes: 65536. Deterministic bounded local software evidence only; no QPU, provider, physical, formal-proof, certification, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_gate_basis_decomposition_v1
ChatGPTProblem: Decompose this supported circuit operation into the supplied bounded gate basis and verify the declared relation. Input: JSON with operation, target basis, phase context. Result: decomposition or withheld candidate, gate list, verification status. Limits: request bytes: 65536. Deterministic bounded local software evidence only; no QPU, provider, physical, formal-proof, certification, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_managed_simulator_v1
ChatGPTProblem: Execute a bounded RQM Circuit IR circuit on the managed local statevector simulator. Input: JSON with circuit ir. Result: typed counts, probabilities, simulation configuration evidence. Limits: operations: 250, qubits: 16, shots: 10000. Deterministic bounded local software evidence; not hardware evidence, a formal proof, certification, signature, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_openqasm3_preflight_v1
ChatGPTProblem: Parse, normalize, and assess OpenQASM 3 before downstream execution. Input: JSON with source. Result: go, no-go, or unsupported verdict, normalized source, warnings and evidence. Limits: source UTF-8 bytes: 65536, verification qubits: 3. Deterministic bounded local software evidence; not hardware evidence, a formal proof, certification, signature, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_openqasm3_repair_v1
ChatGPTProblem: Repair this bounded OpenQASM 3 program only with the supplied conservative repair policy. Input: JSON with source, allowed repairs. Result: repaired or withheld source, exact edit list, post-repair validation verdict. Limits: repair kinds: 4, source UTF-8 bytes: 65536. Deterministic bounded local software evidence only; no QPU, provider acceptance, certification, formal-proof, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_optimization_preservation_audit_v1
ChatGPTProblem: Audit candidate optimization equivalence and qualitative coupling preservation. Input: JSON with original circuit, candidate circuit. Result: preserved or candidate-withheld outcome, input and output hashes, metric and semantic evidence. Limits: request bytes: 65536, verification qubits: 3. Deterministic bounded local software evidence; not hardware evidence, a formal proof, certification, signature, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_phase_aware_rewrite_v1
ChatGPTProblem: Rewrite this supported circuit under the supplied phase-equivalence and target constraints. Input: JSON with operation, target gate, phase context. Result: rewritten operation or original, candidate withholding, verification status. Limits: request bytes: 65536. Deterministic bounded local software evidence only; no QPU, provider, physical, formal-proof, certification, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_phase_sensitive_inspection_v1
ChatGPTProblem: Diagnose the phase, rotation, ordering, or convention difference between these supported operations. Input: JSON with reference, candidate, phase context. Result: difference kind, angle delta, correction hint. Limits: request bytes: 65536. Deterministic bounded local software evidence only; no QPU, provider, physical, formal-proof, certification, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_provider_readiness_check_v1
ChatGPTProblem: Accept or reject this bounded circuit for the supplied pinned provider target and readiness rules. Input: JSON with qubit count, operations, target, rules. Result: readiness verdict, rule results and blockers, target provenance. Limits: maximum operations: 4096, request bytes: 65536. Deterministic bounded local software evidence only; no QPU, live-provider, physical, formal-proof, certification, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_provider_result_compare_v1
ChatGPTProblem: Compare these bounded provider results using the supplied normalization and difference rules. Input: JSON with left counts, right counts, normalization, rules. Result: normalized deltas, probability metrics, caller-rule verdict. Limits: maximum operations: 4096, request bytes: 65536. Deterministic bounded local software evidence only; no QPU, live-provider, physical, formal-proof, certification, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_provider_target_compile_v1
ChatGPTProblem: Compile this bounded circuit to the supplied pinned provider target and return compatibility evidence. Input: JSON with qubit count, operations, target. Result: compiled artifact, gate mapping, pinned target snapshot and blockers. Limits: maximum operations: 4096, request bytes: 65536. Deterministic bounded local software evidence only; no QPU, live-provider, physical, formal-proof, certification, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_simulation_anomaly_diagnostic_v1
ChatGPTProblem: Compare this bounded simulation result with the supplied expectation and anomaly rules. Input: JSON with observed probabilities, expected probabilities, maximum l1 error, maximum bin error. Result: L1 and bin deltas, violating outcomes, anomaly verdict. Limits: request bytes: 65536. Deterministic bounded local software evidence only; no QPU, provider, physical, formal-proof, certification, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_simulation_result_compare_v1
ChatGPTProblem: Compare these bounded simulation results using the supplied deterministic probability rules. Input: JSON with left probabilities, right probabilities, rules. Result: probability deltas, distance metrics, threshold verdict. Limits: maximum operations: 4096, request bytes: 65536. Deterministic bounded local software evidence only; no QPU, live-provider, physical, formal-proof, certification, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_supported_circuit_equivalence_v1
ChatGPTProblem: Compare two RQM Circuit IR circuits within the supported verifier scope. Input: JSON with original circuit, candidate circuit. Result: equivalent, not-equivalent, or unsupported verdict, verification evidence, input hashes. Limits: request bytes: 65536, verification qubits: 3. Deterministic bounded local software evidence; not hardware evidence, a formal proof, certification, signature, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_two_qubit_local_equivalence_v1
ChatGPTProblem: Classify this supported two-qubit operation by its bounded local-equivalence class and phase context. Input: JSON with operation, phase context. Result: local-equivalence class, canonical coordinates, explicit phase treatment. Limits: qubits: 2, supported operations: 5. Deterministic bounded local software evidence only; no QPU, provider acceptance, certification, formal-proof, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_two_qubit_target_synthesis_v1
ChatGPTProblem: Choose and verify a bounded target-basis implementation for this supported two-qubit operation under supplied objectives. Input: JSON with operation, target basis, phase context, objectives. Result: verified circuit or original-retained outcome, gate count and depth, class and phase verification. Limits: max two qubit gates: 8, qubits: 2. Deterministic bounded local software evidence only; no QPU, provider acceptance, certification, formal-proof, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quantum_verified_circuit_optimization_v1
ChatGPTProblem: Optimize RQM Circuit IR and return the original unless preservation is verified. Input: JSON with circuit. Result: verified candidate or original fallback, metric deltas, preservation evidence. Limits: request bytes: 65536, verification qubits: 3. Deterministic bounded local software evidence; not hardware evidence, a formal proof, certification, signature, or quantum-advantage claim. Uses prepaid RQM credits, not a new USDC wallet payment. Catalog tariff identifiers may describe the separate direct buyer/x402 payment route; they do not select or change this connector payment route. Confirm this account-funded operation price and the approved USD ceiling; do not infer account eligibility or switch payment rails from a catalog tariff. The owner's existing saved spending policy still applies. Supply the user's exact inputs, approved USD max_total_price and one stable idempotency_key. A funded balance is not permission to spend. Acceptance is not completion: retrieve saved status, result and receipt with get_account_job. Recover an uncertain purchase with list_account_jobs and the original key before any retry; preserve identical inputs and ceiling. Never submit a new purchase or switch payment rails to recover. This operation does not activate or submit hardware.
quote_hardware_execution
ChatGPTWhen the selected QPU's execution window is closed or its wait does not meet the user's needs, proactively offer a fresh comparison of eligible alternatives for the same circuit and shots. ONLINE does not mean the execution window is open now. Highlight a currently open alternative only with fresh, timestamped provider status, execution-window and live-queue evidence for that exact device; current account/target eligibility, physical-location and workload requirements still apply. For catalog availability, require freshness=fresh, observed_at and age_seconds within max_age_seconds; otherwise treat current availability as unknown. Read window.open_now separately from status and never interpret next_open_at as the task start time. Show the active or next window in UTC and the user's timezone, queue depth and observation time/age. An open window or a shorter queue does not establish a faster finish; disclose unknown start/completion ETA and never estimate runtime from queue depth. Missing, null or stale availability evidence cannot establish an available-now or faster alternative; say live availability is unavailable rather than guessing from catalog generation/review times or ONLINE. Compare itemized provider estimate, RQM fee and fixed customer total for each offered option, marking unquoted totals unknown. Read current available wallet funds and disclose any funding gap; reserved funds are not available to fund another task. Ask for explicit device selection, then obtain a fresh quote and fresh approval of its exact charge before any new submission; comparing alternatives grants no spending or funding approval. For an already queued execution, keep monitoring the original ID/task and leave it untouched: never auto-cancel, replace, resubmit or reuse its approval for another device. dispatch_diagnostics contains historical observations about individual admission requests or worker invocations, not the execution's final outcome. api_admission_refused records a verified application refusal; api_transport_ambiguous records an unconfirmed API delivery; worker_post_claim_not_dispatched records one claimed worker stopping before SDK create. Later verified task/result evidence remains authoritative. These observations never authorize resubmission or hold release and never imply zero provider cost. For customer quotes, copy target_snapshot_id and target_review_sha256 from the selected device customer_quote in a fresh list_quantum_hardware result. For direct quotes and planning, pass the selected device_arn as target_device_arn. For guided quoting, pass it as selected_device_arn and plan_request.target_device_arn; pass target_snapshot_id and target_review_sha256 at the top level. Preserve these values unchanged; renewable catalog generations can advance during a conversation. Never invent or substitute a review digest. An expired review requires a fresh catalog and quote, followed by fresh spending approval. Report expires_at exactly; never guess that a quote expired from perceived conversational delay. Compare with a trustworthy current timestamp, or let Core enforce expiry on an explicitly approved submission. Never bypass expiry or alter approval bindings. Quotes carrying quote_only=true are price quotes only: submission/reservation/dispatch are unavailable while the security stop remains. They cannot be used to submit later; obtain a fresh execution-enabled quote after separate activation. For these quotes lifetime_ceiling_usd=0 grants no provider-spending authority. For legacy execution quotes, lifetime_ceiling_usd is cumulative provider-exposure authority, separate from the customer charge and any overall infrastructure/accounting cap. Issuing a quote does not reserve funds. Customer quotes are capped at $100 total per job, including the service fee and rounding. This charge limit does not cap the account balance. For gate_circuit_100_v1 and gate_circuit_v1, target_review.native_action_sha256 is the legacy reference action; quote.native_action_sha256 binds this selected circuit and must match its frozen prepared artifact. The…
quote_variational_budget
ChatGPTRefund eligibility follows original frozen customer_terms; failure or cancellation alone does not guarantee a refund. Report completion only from the original receipt/settlement evidence. Bounded hosted hardware QAOA MaxCut or product-Pauli VQE, 1–8 qubits, 1–4 layers, 100 shots per separately billable measurement group. Planning is not execution or spending approval. Preserve the user's exact device. Plan the immutable workflow, review all groups and required_maximum_children, then obtain a fresh quote_hardware_execution for the exact next circuit and device. quote_variational_budget binds that first quote, child_maximum_charge, workflow hash, cumulative USD ceiling and child count; the cap must cover every allowed child at that frozen ceiling. Show all terms and obtain explicit customer approval before approve_variational_budget. Each submit_variational_child requires its fresh exact-device quote and original next_iteration_key; Core checks the immutable circuit, target, terms, child ceiling, complete-cohort capacity and saved admission gates. Never change the budget, substitute devices, recycle spent authority or create replacement children to recover an uncertain response. Use get_variational_budget and get_hardware_execution for original-child recovery; bind execution IDs in group order, then recover_hardware_variational_workflow to checkpoint verified evidence and the next trial. Pending children or billing are not completed/settled. Stopping prevents new trials but does not cancel accepted children or promise refunds; cancel_hardware_execution is separate. No convergence, ground-state accuracy or quantum advantage is promised. Wallet-backed derived scientific workflows remain unavailable until their child-delivery path is qualified; ordinary verified child results use the original quote and delivery receipt. Deployment and account/device eligibility are required; unavailable server policy fails closed.
recommend_quantum_device
ChatGPTFree API-owned device advice based on the goal and documented modality. Candidates are not validated for qubit capacity, connectivity, availability, fidelity or cost. Preserve the user's exact choice; never substitute or hide expensive choices. This creates no quote or authorization. Before paid submission verify current exact-device compatibility and account eligibility, obtain a fresh binding quote for the unchanged workload, and receive explicit customer approval. Hardware variational execution requires the coordinated Core budget contract and exact-device admission; simulator advice is not hardware evidence.
recommend_quantum_routes
ChatGPTRequest advisory hardware turnaround routes for the same supplied circuit and shots using API-owned exact-device preparation, timestamped availability/windows/queues and comparable history. Return every ranked_routes option, including incompatible, unavailable, retired and unknown choices, with reasons, freshness and confidence. Preserve exact device_arn or workload.target_device_arn chosen by the user; show selected_device_arn separately from suggested_device_arn. The suggestion never changes the user's selection or an admitted job. This is readiness advice, not a verified fastest finish: null estimated_time_to_delivered_result_seconds means unknown, never zero. Queue count and next_open_at are not durations or delivery ETAs; historical median is observed journal-to-result retrieval, not a forecast. All routes remain runnable=false and execution_authorized=false pending the existing exact-device checks and fresh quote approval. Read-only: never quote, reserve funds, submit, switch, cancel or activate anything in response to this tool. Requires jobs.read, not spending permission.
recover_hardware_variational_workflow
ChatGPTRefund eligibility follows original frozen customer_terms; failure or cancellation alone does not guarantee a refund. Report completion only from the original receipt/settlement evidence. Bounded hosted hardware QAOA MaxCut or product-Pauli VQE, 1–8 qubits, 1–4 layers, 100 shots per separately billable measurement group. Planning is not execution or spending approval. Preserve the user's exact device. Plan the immutable workflow, review all groups and required_maximum_children, then obtain a fresh quote_hardware_execution for the exact next circuit and device. quote_variational_budget binds that first quote, child_maximum_charge, workflow hash, cumulative USD ceiling and child count; the cap must cover every allowed child at that frozen ceiling. Show all terms and obtain explicit customer approval before approve_variational_budget. Each submit_variational_child requires its fresh exact-device quote and original next_iteration_key; Core checks the immutable circuit, target, terms, child ceiling, complete-cohort capacity and saved admission gates. Never change the budget, substitute devices, recycle spent authority or create replacement children to recover an uncertain response. Use get_variational_budget and get_hardware_execution for original-child recovery; bind execution IDs in group order, then recover_hardware_variational_workflow to checkpoint verified evidence and the next trial. Pending children or billing are not completed/settled. Stopping prevents new trials but does not cancel accepted children or promise refunds; cancel_hardware_execution is separate. No convergence, ground-state accuracy or quantum advantage is promised. Wallet-backed derived scientific workflows remain unavailable until their child-delivery path is qualified; ordinary verified child results use the original quote and delivery receipt. Deployment and account/device eligibility are required; unavailable server policy fails closed.
recover_quantum_experiment
ChatGPTWhen the selected QPU's execution window is closed or its wait does not meet the user's needs, proactively offer a fresh comparison of eligible alternatives for the same circuit and shots. ONLINE does not mean the execution window is open now. Highlight a currently open alternative only with fresh, timestamped provider status, execution-window and live-queue evidence for that exact device; current account/target eligibility, physical-location and workload requirements still apply. For catalog availability, require freshness=fresh, observed_at and age_seconds within max_age_seconds; otherwise treat current availability as unknown. Read window.open_now separately from status and never interpret next_open_at as the task start time. Show the active or next window in UTC and the user's timezone, queue depth and observation time/age. An open window or a shorter queue does not establish a faster finish; disclose unknown start/completion ETA and never estimate runtime from queue depth. Missing, null or stale availability evidence cannot establish an available-now or faster alternative; say live availability is unavailable rather than guessing from catalog generation/review times or ONLINE. Compare itemized provider estimate, RQM fee and fixed customer total for each offered option, marking unquoted totals unknown. Read current available wallet funds and disclose any funding gap; reserved funds are not available to fund another task. Ask for explicit device selection, then obtain a fresh quote and fresh approval of its exact charge before any new submission; comparing alternatives grants no spending or funding approval. For an already queued execution, keep monitoring the original ID/task and leave it untouched: never auto-cancel, replace, resubmit or reuse its approval for another device. For hardware comparisons, present every listed device with its exact ARN, provider, region, physical-location evidence, observation times and unknowns. Listed means inventory only; quotable means the current customer_quote/review and quote gates permit a price review for the exact device, circuit and shots; executable additionally requires current execution admission, compatibility, account policy, funds and explicit quote approval. Use describe_hardware_execution to identify the configured default ARN. For that ARN, use its own quote_enabled (falling back to execution_enabled only when quote_enabled is absent) and quote_only flags. For non-default ARNs, read configured_target_readiness_by_device for that exact target's quote_enabled/quote_only; missing non-default readiness is not established. Never borrow another device's readiness. Report each state separately, including blockers; never infer availability from tool visibility, provider online status, a price, past runs or another device's readiness. Compare the same supplied circuit and shot count; disclose differing scope or unsupported inputs. Planning validates the API contract and installed translation offline, not live device compatibility: device_capability_validated=false remains unverified. Width, gates, measurements and shot bounds must fit the selected target's current review and provider capabilities; the worker checks live compatibility before dispatch. Analog hardware is not a gate-circuit substitute. Catalog or planning provider-price estimates are not binding customer quotes or final provider invoices. Show source and observation/expiry times, currency, shots and unknown fee/total fields; never invent a service fee or derive it from a spending ceiling. Only the returned Core quote binds the exact device, circuit, shots, itemized provider estimate, fee, fixed customer total, maximum charge, expiry and terms. quote_only=true cannot reserve funds or execute; later execution requires a fresh execution-enabled quote and fresh approval. Unknown or null queue times mean unknown, not zero wait or immediate availability. Historical elapsed time and polling intervals are not queue/runtime estimates or promised completion times. Absent API turnaround advice, present all choices without c…
refresh_hardware_execution
ChatGPTWhen the selected QPU's execution window is closed or its wait does not meet the user's needs, proactively offer a fresh comparison of eligible alternatives for the same circuit and shots. ONLINE does not mean the execution window is open now. Highlight a currently open alternative only with fresh, timestamped provider status, execution-window and live-queue evidence for that exact device; current account/target eligibility, physical-location and workload requirements still apply. For catalog availability, require freshness=fresh, observed_at and age_seconds within max_age_seconds; otherwise treat current availability as unknown. Read window.open_now separately from status and never interpret next_open_at as the task start time. Show the active or next window in UTC and the user's timezone, queue depth and observation time/age. An open window or a shorter queue does not establish a faster finish; disclose unknown start/completion ETA and never estimate runtime from queue depth. Missing, null or stale availability evidence cannot establish an available-now or faster alternative; say live availability is unavailable rather than guessing from catalog generation/review times or ONLINE. Compare itemized provider estimate, RQM fee and fixed customer total for each offered option, marking unquoted totals unknown. Read current available wallet funds and disclose any funding gap; reserved funds are not available to fund another task. Ask for explicit device selection, then obtain a fresh quote and fresh approval of its exact charge before any new submission; comparing alternatives grants no spending or funding approval. For an already queued execution, keep monitoring the original ID/task and leave it untouched: never auto-cancel, replace, resubmit or reuse its approval for another device. Report refund eligibility from the original frozen customer_terms and refund completion only from its receipt/settlement evidence. Failure or cancellation alone does not guarantee a refund. Only quotes explicitly carrying full_refund_when_verified_failed or full_refund_when_verified_cancelled promise a full refund for that verified condition; do not assume those terms for legacy or owner quotes. A promised refund is not a completed refund: report pending, held, charged, released or refunded amounts exactly as the original receipt records. Unknown outcomes retain holds; never infer zero cost or refund completion from a missing provider task. Recover an existing hardware execution without another paid submission. Core scopes every call to the authenticated account and principal; no account, principal, email or chat/session selector is accepted. A new chat or token for the original approved OAuth client retains its durable owner identity. Other approved clients have different pairwise subjects/principals and cannot recover these jobs without a separately reviewed owner-linking contract. Matching email, wallet or audience alone grants no access. Use list_hardware_executions when execution IDs or submission keys were lost. It defaults to held_only=true; use held_only=false for settled history. Follow next_cursor unchanged with the same held_only until null; pages are ordered by execution ID, not recency. held_total covers all this caller's held executions, not just the page. Listing does not poll providers, acknowledge results, release holds or submit work. Use a returned execution_id with get_hardware_execution or get_hardware_receipt for verified status, results and receipt. Those retrieval tools acknowledge a verified available result under its original frozen billing terms exactly once; listing's result_available is only a summary, not verified result evidence. If recovery.can_refresh is true, refresh_hardware_execution can reconcile original provider status without creating or dispatching a task. Unknown acceptance retains its hold until evidence resolves it; final_billing_required needs operator final-allocation evidence and cannot be cleared by resubmitting, canceling or switching payment rails. A missing job on one page or an empty list does…
review_quantum_experiment_price
ChatGPTWhen the selected QPU's execution window is closed or its wait does not meet the user's needs, proactively offer a fresh comparison of eligible alternatives for the same circuit and shots. ONLINE does not mean the execution window is open now. Highlight a currently open alternative only with fresh, timestamped provider status, execution-window and live-queue evidence for that exact device; current account/target eligibility, physical-location and workload requirements still apply. For catalog availability, require freshness=fresh, observed_at and age_seconds within max_age_seconds; otherwise treat current availability as unknown. Read window.open_now separately from status and never interpret next_open_at as the task start time. Show the active or next window in UTC and the user's timezone, queue depth and observation time/age. An open window or a shorter queue does not establish a faster finish; disclose unknown start/completion ETA and never estimate runtime from queue depth. Missing, null or stale availability evidence cannot establish an available-now or faster alternative; say live availability is unavailable rather than guessing from catalog generation/review times or ONLINE. Compare itemized provider estimate, RQM fee and fixed customer total for each offered option, marking unquoted totals unknown. Read current available wallet funds and disclose any funding gap; reserved funds are not available to fund another task. Ask for explicit device selection, then obtain a fresh quote and fresh approval of its exact charge before any new submission; comparing alternatives grants no spending or funding approval. For an already queued execution, keep monitoring the original ID/task and leave it untouched: never auto-cancel, replace, resubmit or reuse its approval for another device. For hardware comparisons, present every listed device with its exact ARN, provider, region, physical-location evidence, observation times and unknowns. Listed means inventory only; quotable means the current customer_quote/review and quote gates permit a price review for the exact device, circuit and shots; executable additionally requires current execution admission, compatibility, account policy, funds and explicit quote approval. Use describe_hardware_execution to identify the configured default ARN. For that ARN, use its own quote_enabled (falling back to execution_enabled only when quote_enabled is absent) and quote_only flags. For non-default ARNs, read configured_target_readiness_by_device for that exact target's quote_enabled/quote_only; missing non-default readiness is not established. Never borrow another device's readiness. Report each state separately, including blockers; never infer availability from tool visibility, provider online status, a price, past runs or another device's readiness. Compare the same supplied circuit and shot count; disclose differing scope or unsupported inputs. Planning validates the API contract and installed translation offline, not live device compatibility: device_capability_validated=false remains unverified. Width, gates, measurements and shot bounds must fit the selected target's current review and provider capabilities; the worker checks live compatibility before dispatch. Analog hardware is not a gate-circuit substitute. Catalog or planning provider-price estimates are not binding customer quotes or final provider invoices. Show source and observation/expiry times, currency, shots and unknown fee/total fields; never invent a service fee or derive it from a spending ceiling. Only the returned Core quote binds the exact device, circuit, shots, itemized provider estimate, fee, fixed customer total, maximum charge, expiry and terms. quote_only=true cannot reserve funds or execute; later execution requires a fresh execution-enabled quote and fresh approval. Unknown or null queue times mean unknown, not zero wait or immediate availability. Historical elapsed time and polling intervals are not queue/runtime estimates or promised completion times. Absent API turnaround advice, present all choices without c…
search_buyer_jobs
ChatGPTSearch supported quantum software services by a concrete objective. Returns candidate descriptions, input requirements, readiness, evidence limits and references to the separately exposed tools. Free descriptive read; performs no computation, purchase or spending authorization.
stop_hardware_variational_workflow
ChatGPTRefund eligibility follows original frozen customer_terms; failure or cancellation alone does not guarantee a refund. Report completion only from the original receipt/settlement evidence. Bounded hosted hardware QAOA MaxCut or product-Pauli VQE, 1–8 qubits, 1–4 layers, 100 shots per separately billable measurement group. Planning is not execution or spending approval. Preserve the user's exact device. Plan the immutable workflow, review all groups and required_maximum_children, then obtain a fresh quote_hardware_execution for the exact next circuit and device. quote_variational_budget binds that first quote, child_maximum_charge, workflow hash, cumulative USD ceiling and child count; the cap must cover every allowed child at that frozen ceiling. Show all terms and obtain explicit customer approval before approve_variational_budget. Each submit_variational_child requires its fresh exact-device quote and original next_iteration_key; Core checks the immutable circuit, target, terms, child ceiling, complete-cohort capacity and saved admission gates. Never change the budget, substitute devices, recycle spent authority or create replacement children to recover an uncertain response. Use get_variational_budget and get_hardware_execution for original-child recovery; bind execution IDs in group order, then recover_hardware_variational_workflow to checkpoint verified evidence and the next trial. Pending children or billing are not completed/settled. Stopping prevents new trials but does not cancel accepted children or promise refunds; cancel_hardware_execution is separate. No convergence, ground-state accuracy or quantum advantage is promised. Wallet-backed derived scientific workflows remain unavailable until their child-delivery path is qualified; ordinary verified child results use the original quote and delivery receipt. Deployment and account/device eligibility are required; unavailable server policy fails closed.
submit_hardware_execution
ChatGPTWhen the selected QPU's execution window is closed or its wait does not meet the user's needs, proactively offer a fresh comparison of eligible alternatives for the same circuit and shots. ONLINE does not mean the execution window is open now. Highlight a currently open alternative only with fresh, timestamped provider status, execution-window and live-queue evidence for that exact device; current account/target eligibility, physical-location and workload requirements still apply. For catalog availability, require freshness=fresh, observed_at and age_seconds within max_age_seconds; otherwise treat current availability as unknown. Read window.open_now separately from status and never interpret next_open_at as the task start time. Show the active or next window in UTC and the user's timezone, queue depth and observation time/age. An open window or a shorter queue does not establish a faster finish; disclose unknown start/completion ETA and never estimate runtime from queue depth. Missing, null or stale availability evidence cannot establish an available-now or faster alternative; say live availability is unavailable rather than guessing from catalog generation/review times or ONLINE. Compare itemized provider estimate, RQM fee and fixed customer total for each offered option, marking unquoted totals unknown. Read current available wallet funds and disclose any funding gap; reserved funds are not available to fund another task. Ask for explicit device selection, then obtain a fresh quote and fresh approval of its exact charge before any new submission; comparing alternatives grants no spending or funding approval. For an already queued execution, keep monitoring the original ID/task and leave it untouched: never auto-cancel, replace, resubmit or reuse its approval for another device. dispatch_diagnostics contains historical observations about individual admission requests or worker invocations, not the execution's final outcome. api_admission_refused records a verified application refusal; api_transport_ambiguous records an unconfirmed API delivery; worker_post_claim_not_dispatched records one claimed worker stopping before SDK create. Later verified task/result evidence remains authoritative. These observations never authorize resubmission or hold release and never imply zero provider cost. For customer quotes, copy target_snapshot_id and target_review_sha256 from the selected device customer_quote in a fresh list_quantum_hardware result. For direct quotes and planning, pass the selected device_arn as target_device_arn. For guided quoting, pass it as selected_device_arn and plan_request.target_device_arn; pass target_snapshot_id and target_review_sha256 at the top level. Preserve these values unchanged; renewable catalog generations can advance during a conversation. Never invent or substitute a review digest. An expired review requires a fresh catalog and quote, followed by fresh spending approval. Report expires_at exactly; never guess that a quote expired from perceived conversational delay. Compare with a trustworthy current timestamp, or let Core enforce expiry on an explicitly approved submission. Never bypass expiry or alter approval bindings. Quotes carrying quote_only=true are price quotes only: submission/reservation/dispatch are unavailable while the security stop remains. They cannot be used to submit later; obtain a fresh execution-enabled quote after separate activation. For these quotes lifetime_ceiling_usd=0 grants no provider-spending authority. For legacy execution quotes, lifetime_ceiling_usd is cumulative provider-exposure authority, separate from the customer charge and any overall infrastructure/accounting cap. Issuing a quote does not reserve funds. Customer quotes are capped at $100 total per job, including the service fee and rounding. This charge limit does not cap the account balance. For gate_circuit_100_v1 and gate_circuit_v1, target_review.native_action_sha256 is the legacy reference action; quote.native_action_sha256 binds this selected circuit and must match its frozen prepared artifact. The…
submit_quantum_experiment
ChatGPTWhen the selected QPU's execution window is closed or its wait does not meet the user's needs, proactively offer a fresh comparison of eligible alternatives for the same circuit and shots. ONLINE does not mean the execution window is open now. Highlight a currently open alternative only with fresh, timestamped provider status, execution-window and live-queue evidence for that exact device; current account/target eligibility, physical-location and workload requirements still apply. For catalog availability, require freshness=fresh, observed_at and age_seconds within max_age_seconds; otherwise treat current availability as unknown. Read window.open_now separately from status and never interpret next_open_at as the task start time. Show the active or next window in UTC and the user's timezone, queue depth and observation time/age. An open window or a shorter queue does not establish a faster finish; disclose unknown start/completion ETA and never estimate runtime from queue depth. Missing, null or stale availability evidence cannot establish an available-now or faster alternative; say live availability is unavailable rather than guessing from catalog generation/review times or ONLINE. Compare itemized provider estimate, RQM fee and fixed customer total for each offered option, marking unquoted totals unknown. Read current available wallet funds and disclose any funding gap; reserved funds are not available to fund another task. Ask for explicit device selection, then obtain a fresh quote and fresh approval of its exact charge before any new submission; comparing alternatives grants no spending or funding approval. For an already queued execution, keep monitoring the original ID/task and leave it untouched: never auto-cancel, replace, resubmit or reuse its approval for another device. For hardware comparisons, present every listed device with its exact ARN, provider, region, physical-location evidence, observation times and unknowns. Listed means inventory only; quotable means the current customer_quote/review and quote gates permit a price review for the exact device, circuit and shots; executable additionally requires current execution admission, compatibility, account policy, funds and explicit quote approval. Use describe_hardware_execution to identify the configured default ARN. For that ARN, use its own quote_enabled (falling back to execution_enabled only when quote_enabled is absent) and quote_only flags. For non-default ARNs, read configured_target_readiness_by_device for that exact target's quote_enabled/quote_only; missing non-default readiness is not established. Never borrow another device's readiness. Report each state separately, including blockers; never infer availability from tool visibility, provider online status, a price, past runs or another device's readiness. Compare the same supplied circuit and shot count; disclose differing scope or unsupported inputs. Planning validates the API contract and installed translation offline, not live device compatibility: device_capability_validated=false remains unverified. Width, gates, measurements and shot bounds must fit the selected target's current review and provider capabilities; the worker checks live compatibility before dispatch. Analog hardware is not a gate-circuit substitute. Catalog or planning provider-price estimates are not binding customer quotes or final provider invoices. Show source and observation/expiry times, currency, shots and unknown fee/total fields; never invent a service fee or derive it from a spending ceiling. Only the returned Core quote binds the exact device, circuit, shots, itemized provider estimate, fee, fixed customer total, maximum charge, expiry and terms. quote_only=true cannot reserve funds or execute; later execution requires a fresh execution-enabled quote and fresh approval. Unknown or null queue times mean unknown, not zero wait or immediate availability. Historical elapsed time and polling intervals are not queue/runtime estimates or promised completion times. Absent API turnaround advice, present all choices without c…
submit_variational_child
ChatGPTRefund eligibility follows original frozen customer_terms; failure or cancellation alone does not guarantee a refund. Report completion only from the original receipt/settlement evidence. Bounded hosted hardware QAOA MaxCut or product-Pauli VQE, 1–8 qubits, 1–4 layers, 100 shots per separately billable measurement group. Planning is not execution or spending approval. Preserve the user's exact device. Plan the immutable workflow, review all groups and required_maximum_children, then obtain a fresh quote_hardware_execution for the exact next circuit and device. quote_variational_budget binds that first quote, child_maximum_charge, workflow hash, cumulative USD ceiling and child count; the cap must cover every allowed child at that frozen ceiling. Show all terms and obtain explicit customer approval before approve_variational_budget. Each submit_variational_child requires its fresh exact-device quote and original next_iteration_key; Core checks the immutable circuit, target, terms, child ceiling, complete-cohort capacity and saved admission gates. Never change the budget, substitute devices, recycle spent authority or create replacement children to recover an uncertain response. Use get_variational_budget and get_hardware_execution for original-child recovery; bind execution IDs in group order, then recover_hardware_variational_workflow to checkpoint verified evidence and the next trial. Pending children or billing are not completed/settled. Stopping prevents new trials but does not cancel accepted children or promise refunds; cancel_hardware_execution is separate. No convergence, ground-state accuracy or quantum advantage is promised. Wallet-backed derived scientific workflows remain unavailable until their child-delivery path is qualified; ordinary verified child results use the original quote and delivery receipt. Deployment and account/device eligibility are required; unavailable server policy fails closed.