State model
A signed system manifest links code, prompt, model, knowledge, tools, permissions, policies, tests and owner. Any material dependency change creates a new release candidate.
A guardrail can block a response yet still leave the root cause, affected journey and customer impact invisible to operational teams.
The observability layer records guardrail intent, trigger, upstream context, blocked effect and recovery route, then groups incidents by control failure rather than message similarity alone.
Organisations, systems and operating conditions are intentionally anonymised and recomposed. The design demonstrates engineering and banking-domain reasoning; it does not represent a named client estate, vendor product or measured production result.
Five operating planes separate interaction, identity, decision control, authoritative state and operating evidence.
Permitted workThe assurance layer observes and evaluates; it may trigger a stop or review but cannot redefine business policy from telemetry alone.
Consistency ruleCarry one correlation chain across synchronous and asynchronous hops and separate event occurrence from observation and processing time.
Hard boundaryThe model is not a system of record, identity provider, policy authority or proof that an external effect occurred.
Agents, prompts, tools, policies and models must be built, certified, operated, changed and retired as one reachable system.
Diagnostic data is purpose-limited and redacted before aggregation; sensitive customer content is not copied into engineering telemetry.
A deterministic outer workflow contains model-led work inside typed, observable calls. Dashed messages remain proposals until policy or a human grants authority.
A signed system manifest links code, prompt, model, knowledge, tools, permissions, policies, tests and owner. Any material dependency change creates a new release candidate.
Build and runtime boundaries use the same schemas. Registry metadata drives discovery, policy, telemetry attribution and deprecation.
The deployed artefact, registry entry and evidence pack share one immutable release identifier; partial promotion is not a valid state.
These roles are deliberately vendor-neutral. Each can be independently owned, versioned and replaced.
Evaluates identity, purpose, capability, amount, risk tier and policy version; returns allow, deny, step-up or human-review with reasons.
Appends request, versions, policy result, model proposal, approval, action receipt, readback, correction and custody events under one correlation key.
Runs component, route, trajectory, failure, harm and outcome tests against the versioned system manifest.
Classifies unknown outcomes, replays idempotent work, runs compensations and restores from the last verified checkpoint.
Durable records carry provenance, authority, effect and custody without turning a transcript into an uncontrolled memory store.
Carry one correlation chain across synchronous and asynchronous hops and separate event occurrence from observation and processing time.
The selected design is not universally superior. It is the safer fit for this boundary and failure cost.
Compile stable decision logic and retain retrieval for explanation and residual ambiguity.
Ask a model to interpret the source document for every request.
Cost acceptedRule compilation needs controlled change, but creates repeatable decisions, regression tests and clear exceptions.
Use append-only events plus a rebuildable current-state projection.
Overwrite the case row with its latest status.
Cost acceptedReplay and storage are more complex, but point-in-time reconstruction and correction lineage remain possible.
Test prompts, models, tools, knowledge, policies, state transitions and human paths together.
Use a static answer-quality benchmark as the release gate.
Cost acceptedSystem evaluation takes longer and needs synthetic environments, but detects authority and recovery failures that answer scoring misses.
Use local commits, idempotent steps, compensations and an unknown-outcome state.
Attempt one atomic transaction across independent banking systems.
Cost acceptedSagas expose temporary inconsistency and require recovery logic, but fit systems that cannot share one transaction boundary.
Retries are bounded by knowledge of business effect; unknown outcome remains visible, owned and independently reconciled.
Actual thresholds belong to accountable service owners. The design exposes the equations and observables that those owners must baseline.
gate_runtime = test_cases x routes x dependency_versionsrelease_capacity = available_environments / average_gate_durationoperating_cost = model + tools + platform + human_review + incidentsCollect the minimum diagnostic metadata, tokenise subjects and keep raw prompts or case evidence behind stricter access and retention.
A design is production-ready only when teams can prove what happened, recover it and change it safely.
trace and outcome completeness
control-bypass and false-negative review
cost and human-effort attribution
kill-switch and recovery exercise
Control version, trigger class, context fingerprint, blocked segment or action, recovery result, impact assessment and owner response.
Control and service owners classify severity, tune the route and approve any rule change.
Automate evidence collection before automating approval. Expand self-service only when the paved route proves safer and faster than bespoke delivery.
Service, control, finance, model, data and operations owners interpret evidence and decide intervention.