Verwandte Artikel zu Reliable AI Agents: Guardrails, Evals, and the Deterministic...
Reliable AI Agents: Guardrails, Evals, and the Deterministic Core — Engineering Trust. Red-teaming, invariants, injection defense, observability, and ... real credit age (The AI Agent Series, Band 5) - Softcover
Buch 5 von 7: The AI Agent Series
Inhaltsangabe
Your agent works in the demo. Now make it safe enough to put money — and a regulator — behind.
When a wrong answer is an awkward email, "usually works" is fine. When it is money that does not come back, a limit breached, a regulator's question, "usually works" is a liability with good demos. This book engineers the difference. Its running system is Harbor Credit — an agent that extends trade credit to a small business's wholesale buyers — and across ten chapters it wraps that agent in the full assurance pipeline that turns an impressive prototype into a system you can certify.
Through ten chapters, each moving from a reproduced failure to tested code, you will build:
- A red-team harness and a six-type failure taxonomy — because agent failures are systematic, not bad luck
- Evals that report quality and safety as separate numbers, with a calibrated LLM-judge and a golden set as executable spec
- Input and output guardrails — deterministic, fail-safe rails at the boundary, the model free in the middle
- The deterministic core: business invariants in plain Python that have the last word over whatever the model proposes
- Business invariants elicited from catastrophe, sorted from preferences, and pinned to acceptance cases
- A threat model and the strongest security an agent can have — safe by construction, where even an obeyed injection cannot breach a limit
- Observability with byte-for-byte trajectory replay — you debug trajectories, not lines
- Graceful degradation — timeouts, circuit breakers, and a safe mode that never auto-approves
- Property-based and adversarial testing that certifies distributions, not single cases
- The assurance case — the auditable, evidenced argument, with disclosed residual risk, that the agent is safe to deploy
The governing idea, held for ten chapters: put the LLM at the boundary and the invariants in the core, so safety holds regardless of model quality — perfect, degraded, or jailbroken. Each chapter adds a research lineage (FMEA and fault trees to reference monitors, QuickCheck to the safety case), a failure catalog, and exercises. The companion repo has one dependency, thirty-three tests, and runs entirely offline.
Who it's for: senior engineers, and anyone deploying agents in high-stakes or regulated domains, who have built and dissected an agent (Books 2–3 or equivalent).
The AI Agent Series takes readers from first contact to professional mastery across seven volumes. This is Book 5: trust, engineered.
Die Inhaltsangabe kann sich auf eine andere Ausgabe dieses Titels beziehen.
Suchergebnisse für Reliable AI Agents: Guardrails, Evals, and the Deterministic...
Beispielbild für diese ISBN
Reliable AI Agents (Paperback)
Print-on-DemandAnbieter: Grand Eagle Retail, Bensenville, IL, USA
Verkäuferbewertung 5 von 5 Sternen
Paperback. Zustand: new. Paperback. Your agent works in the demo. Now make it safe enough to put money - and a regulator - behind.When a wrong answer is an awkward email, "usually works" is fine. When it is money that does not come back, a limit breached, a regulator's question, "usually works" is a liability with good demos. This book engineers the difference. Its running system is Harbor Credit - an agent that extends trade credit to a small business's wholesale buyers - and across ten chapters it wraps that agent in the full assurance pipeline that turns an impressive prototype into a system you can certify.Through ten chapters, each moving from a reproduced failure to tested code, you will build: A red-team harness and a six-type failure taxonomy - because agent failures are systematic, not bad luckEvals that report quality and safety as separate numbers, with a calibrated LLM-judge and a golden set as executable specInput and output guardrails - deterministic, fail-safe rails at the boundary, the model free in the middleThe deterministic core: business invariants in plain Python that have the last word over whatever the model proposesBusiness invariants elicited from catastrophe, sorted from preferences, and pinned to acceptance casesA threat model and the strongest security an agent can have - safe by construction, where even an obeyed injection cannot breach a limitObservability with byte-for-byte trajectory replay - you debug trajectories, not linesGraceful degradation - timeouts, circuit breakers, and a safe mode that never auto-approvesProperty-based and adversarial testing that certifies distributions, not single casesThe assurance case - the auditable, evidenced argument, with disclosed residual risk, that the agent is safe to deploy The governing idea, held for ten chapters: put the LLM at the boundary and the invariants in the core, so safety holds regardless of model quality - perfect, degraded, or jailbroken. Each chapter adds a research lineage (FMEA and fault trees to reference monitors, QuickCheck to the safety case), a failure catalog, and exercises. The companion repo has one dependency, thirty-three tests, and runs entirely offline.Who it's for: senior engineers, and anyone deploying agents in high-stakes or regulated domains, who have built and dissected an agent (Books 2-3 or equivalent).The AI Agent Series takes readers from first contact to professional mastery across seven volumes. This is Book 5: trust, engineered. This item is printed on demand. Shipping may be from multiple locations in the US or from the UK, depending on stock availability. Bestandsnummer des Verkäufers 9798171349462
Beispielbild für diese ISBN
Reliable AI Agents
Print-on-DemandAnbieter: PBShop.store US, Wood Dale, IL, USA
Verkäuferbewertung 5 von 5 Sternen
PAP. Zustand: New. New Book. Shipped from UK. THIS BOOK IS PRINTED ON DEMAND. Established seller since 2000. Bestandsnummer des Verkäufers L0-9798171349462
Neu kaufen
EUR 17,22
Versand gratis
Versand innerhalb von USA
Versand innerhalb von USA
Anzahl: Mehr als 20 verfügbar
Beispielbild für diese ISBN
Reliable AI Agents
Print-on-DemandAnbieter: PBShop.store UK, Fairford, GLOS, Vereinigtes Königreich
Verkäuferbewertung 5 von 5 Sternen
PAP. Zustand: New. New Book. Delivered from our UK warehouse in 4 to 14 business days. THIS BOOK IS PRINTED ON DEMAND. Established seller since 2000. Bestandsnummer des Verkäufers L0-9798171349462
Neu kaufen
EUR 15,87
EUR 3,94 Versand
Versand von Vereinigtes Königreich nach USA
Versand von Vereinigtes Königreich nach USA
Anzahl: Mehr als 20 verfügbar
Beispielbild für diese ISBN
Reliable AI Agents : Guardrails, Evals, and the Deterministic Core - Engineering Trust. Red-teaming, invariants, injection defense, observability, and the assurance case, built around a real credit age
Anbieter: AHA-BUCH GmbH, Einbeck, Deutschland
Verkäuferbewertung 5 von 5 Sternen
Taschenbuch. Zustand: Neu. Neuware. Bestandsnummer des Verkäufers 9798171349462
Neu kaufen
EUR 27,95
EUR 35,00 Versand
Versand von Deutschland nach USA
Versand von Deutschland nach USA
Anzahl: 2 verfügbar
Beispielbild für diese ISBN
Reliable AI Agents (Paperback)
Print-on-DemandAnbieter: CitiRetail, Stevenage, Vereinigtes Königreich
Verkäuferbewertung 5 von 5 Sternen
Paperback. Zustand: new. Paperback. Your agent works in the demo. Now make it safe enough to put money - and a regulator - behind.When a wrong answer is an awkward email, "usually works" is fine. When it is money that does not come back, a limit breached, a regulator's question, "usually works" is a liability with good demos. This book engineers the difference. Its running system is Harbor Credit - an agent that extends trade credit to a small business's wholesale buyers - and across ten chapters it wraps that agent in the full assurance pipeline that turns an impressive prototype into a system you can certify.Through ten chapters, each moving from a reproduced failure to tested code, you will build: A red-team harness and a six-type failure taxonomy - because agent failures are systematic, not bad luckEvals that report quality and safety as separate numbers, with a calibrated LLM-judge and a golden set as executable specInput and output guardrails - deterministic, fail-safe rails at the boundary, the model free in the middleThe deterministic core: business invariants in plain Python that have the last word over whatever the model proposesBusiness invariants elicited from catastrophe, sorted from preferences, and pinned to acceptance casesA threat model and the strongest security an agent can have - safe by construction, where even an obeyed injection cannot breach a limitObservability with byte-for-byte trajectory replay - you debug trajectories, not linesGraceful degradation - timeouts, circuit breakers, and a safe mode that never auto-approvesProperty-based and adversarial testing that certifies distributions, not single casesThe assurance case - the auditable, evidenced argument, with disclosed residual risk, that the agent is safe to deploy The governing idea, held for ten chapters: put the LLM at the boundary and the invariants in the core, so safety holds regardless of model quality - perfect, degraded, or jailbroken. Each chapter adds a research lineage (FMEA and fault trees to reference monitors, QuickCheck to the safety case), a failure catalog, and exercises. The companion repo has one dependency, thirty-three tests, and runs entirely offline.Who it's for: senior engineers, and anyone deploying agents in high-stakes or regulated domains, who have built and dissected an agent (Books 2-3 or equivalent).The AI Agent Series takes readers from first contact to professional mastery across seven volumes. This is Book 5: trust, engineered. This item is printed on demand. Shipping may be from our UK warehouse or from our Australian or US warehouses, depending on stock availability. Bestandsnummer des Verkäufers 9798171349462
Neu kaufen
EUR 19,73
EUR 44,32 Versand
Versand von Vereinigtes Königreich nach USA
Versand von Vereinigtes Königreich nach USA
Anzahl: 1 verfügbar