Project 04: Structured Output Contract

Build a schema-validated extractor that guarantees structured, safe output for downstream tool use.

Quick Reference

Attribute Value
Difficulty Level 3
Time Estimate 1 week
Main Programming Language Python
Alternative Programming Languages JavaScript, TypeScript
Coolness Level 3
Business Potential 4
Prerequisites JSON, validation concepts
Key Topics schema validation, output correction

1. Learning Objectives

By completing this project, you will:

  1. Define a strict output schema
  2. Validate outputs with a guardrails framework
  3. Implement repair and fallback policies
  4. Measure schema failure rates

2. All Theory Needed (Per-Concept Breakdown)

Guardrails Control Plane Fundamentals

Fundamentals A guardrails control plane is the set of policies, detectors, validators, and decision rules that sit around an LLM agent to ensure it behaves safely and predictably. Unlike traditional input validation, guardrails must handle probabilistic outputs, ambiguous intent, and adversarial prompts. The control plane therefore spans the entire lifecycle of an interaction: input filtering, context validation (including RAG sources), output moderation, and tool-use permissioning. Frameworks such as Guardrails AI and NeMo Guardrails provide structured validation and dialogue control, while models like Prompt Guard or Llama Guard provide detection signals that must be interpreted by policy.  The core insight is that no single framework enforces safety end-to-end; you must compose multiple controls and define how they interact.

Deep Dive into the concept A control plane begins with policy: a formal definition of what is allowed and why. Governance frameworks like the NIST AI RMF and ISO/IEC 42001 provide the organizational structure for this policy layer, while OWASP LLM Top 10 provides a security taxonomy for risks.  Policies must be translated into guardrail rules that are actionable: detect prompt injection in untrusted data, validate output schemas before tools execute, and block unsafe categories. This translation is non-trivial because LLM outputs are probabilistic and context-sensitive. A policy such as “never leak secrets” must be expressed as a chain of checks: input scanning for malicious prompts, context segmentation for untrusted data, output moderation to catch leakage, and tool gating to prevent exfiltration. Each check introduces uncertainty and cost, which means policy must include thresholds, confidence levels, and escalation paths.

Detectors such as Prompt Guard, Lakera Guard, or Rebuff provide risk scores and categories for injection attempts.  These detectors are probabilistic and therefore require calibration. The control plane must normalize detector outputs into a shared risk scale, define what “block” vs “review” means, and log the decision context for later auditing. Output guardrails such as Llama Guard or OpenAI moderation detect unsafe content in generated responses.  These checks must be aligned to your own taxonomy; the model’s categories may not map exactly to your policy. This is why evaluation and red-teaming are crucial: without testing, you do not know if your thresholds or taxonomy mappings are effective.

Structured output validation adds determinism to a probabilistic system. Guardrails AI uses validators and schema checks to ensure outputs conform to an expected structure, enabling safer tool calls and data extraction.  NeMo Guardrails extends this by introducing Colang, a flow language that constrains dialogue paths and allows explicit safety steps, such as mandatory disclaimers or confirmation prompts.  These frameworks provide building blocks, but they do not decide how to integrate them into a business context. For example, a schema validator can ensure a tool call is syntactically correct, but only policy can decide if that tool call should be allowed at all. This is why tool permissioning and sandboxing are critical complement pieces that most guardrails frameworks do not provide natively.

Evaluation is the evidence layer. Tools like garak and OpenAI Evals allow you to run red-team tests and custom evaluation suites to measure whether guardrails are actually working.  Without these tests, guardrails may create a false sense of security. Monitoring and telemetry are the final layer: you must log guardrail decisions, measure false positives and negatives, and track drift over time. Guardrails AI supports observability integration via OpenTelemetry, which can feed monitoring dashboards for guardrail KPIs.  The control plane is therefore a loop: policies drive controls, controls generate evidence, and evidence updates policies. This loop is the only sustainable way to manage guardrails in production.

How this fit on projects

  • You will apply this control-plane model directly in §5.4 and §5.11 and validate it in §6.

Definitions & key terms

  • Control plane: The policy-driven layer that decides what an agent may do.
  • Detector: A model or rule that assigns risk categories or scores. 
  • Validator: A structured check that enforces schema or constraints. 
  • Tool gating: Permissions and constraints for tool execution.
  • Evaluation suite: A set of tests that measure guardrails effectiveness. 

Mental model diagram

Policy -> Detectors -> Validators -> Tool Gate -> Output
 ^ | | | |
 | v v v v
Evidence <- Logs <- Thresholds <- Decisions <- Monitoring

How it works (step-by-step)

  1. Define policy risks and acceptable thresholds.
  2. Select detectors and validators aligned to those risks.
  3. Normalize detector outputs and enforce schema rules.
  4. Apply tool permissions based on risk and context.
  5. Log decisions and run evaluation suites continuously.

Minimal concrete example

Guardrail Decision Record
- input_source: retrieved_doc
- detector: prompt_injection
- score: 0.84
- policy_action: block
- tool_gate: deny
- audit_id: 2026-01-03-0001

Common misconceptions

  • “A single framework solves guardrails end-to-end.”
  • “Moderation is enough to prevent prompt injection.”
  • “Validation guarantees correctness without policy.”

Check-your-understanding questions

  1. Why is a policy layer required in addition to detectors?
  2. How do validators reduce tool misuse risk?
  3. Why is evaluation necessary even if detectors exist?

Check-your-understanding answers

  1. Detectors provide signals, but policy decides actions and thresholds.
  2. Validators ensure structured, safe tool inputs before execution.
  3. Detectors can fail or drift; evaluation reveals blind spots.

Real-world applications

  • Enterprise assistants with access to sensitive data
  • RAG systems ingesting third-party documents
  • Autonomous workflows with high-impact tools

Where you’ll apply it

  • See §5.4 and §6 in this file.
  • Also used in: P02-prompt-injection-firewall.md, P03-content-safety-gate.md, P08-policy-router-orchestrator.md.

References

  • NIST AI RMF 1.0. 
  • ISO/IEC 42001:2023 AI Management Systems. 
  • OWASP LLM Top 10 v1.1. 
  • Guardrails AI framework. 
  • NeMo Guardrails and Colang. 
  • Prompt Guard model card. 
  • Llama Guard documentation. 
  • garak LLM scanner. 
  • OpenAI Evals. 

Key insights Guardrails are a control plane, not a single model or API.

Summary A layered control plane combines policy, detection, validation, and evaluation into a continuous safety loop.

Homework/Exercises to practice the concept

  • Draft a policy map with three risks and a detector for each.
  • Define a monitoring dashboard with three guardrail KPIs.

Solutions to the homework/exercises

  • Example risks: injection, data leakage, tool misuse; KPIs: block rate, false positives, tool denial rate.

3. Project Specification

3.1 What You Will Build

A schema validation layer that accepts outputs only when they conform to a contract.

3.2 Functional Requirements

  1. Define a schema with required fields
  2. Validate model output with Guardrails AI 
  3. Retry or repair invalid outputs
  4. Log validation failures

3.3 Non-Functional Requirements

  • Performance: Validation under 500ms for normal outputs
  • Reliability: Deterministic validation results
  • Usability: Clear error reporting for invalid outputs

3.4 Example Usage / Output

$ structured-extract run --schema invoice
Validated Output: {vendor: Acme, total: 199.50}

3.5 Data Formats / Schemas / Protocols

Schema definition: fields, types, constraints, required flags

3.6 Edge Cases

  • Missing required fields
  • Invalid numeric formats
  • Extra unexpected fields

3.7 Real World Outcome

This section is the golden reference. You will compare your output against it.

3.7.1 How to Run (Copy/Paste)

  • Load schema file
  • Run extractor CLI on sample input
  • Inspect validation report

3.7.2 Golden Path Demo (Deterministic)

Use a fixed input text and expected structured output to verify deterministic validation.

3.7.3 If CLI: Exact Terminal Transcript (Success)

$ structured-extract run --schema invoice
Status: VALID
Fields: vendor=Acme total=199.50 currency=USD

3.7.4 Failure Demo (Deterministic)

$ structured-extract run --schema invoice
Status: INVALID
Error: missing field 'currency'

Exit codes: 0 on valid, 2 on schema failure

4. Solution Architecture

A validator sits between LLM output and downstream consumers, enforcing schema contracts.

4.1 High-Level Design

┌─────────────┐ ┌─────────────┐ ┌─────────────┐
│ Input │────▶│ Policy │────▶│ Output │
│ Handler │ │ Engine │ │ Reporter │
└─────────────┘ └─────────────┘ └─────────────┘

4.2 Key Components

Component Responsibility Key Decisions
Schema Loader Reads schema definitions Versioned schemas
Validator Checks output fields Deterministic rules
Repair Loop Attempts correction Max retries

4.4 Data Structures (No Full Code)

Validation report: status, errors, retries, final_output

4.4 Algorithm Overview

  1. Parse schema
  2. Validate output
  3. Repair if invalid
  4. Emit report

5. Implementation Guide

5.1 Development Environment Setup

mkdir structured-output && cd structured-output

5.2 Project Structure

project/
├── schemas/
├── validator/
├── logs/
└── tests/

5.3 The Core Question You’re Answering

“How do I guarantee the model’s output matches a strict schema before I trust it?”

Schema validation reduces unsafe tool calls and data corruption.

5.4 Concepts You Must Understand First

  1. Schema validation
  2. Repair strategies

5.5 Questions to Guide Your Design

  1. Schema strictness
    • Which fields are required?
    • Which can be optional?
  2. Repair policy
    • How many retries?
    • When to fail fast?

5.6 Thinking Exercise

Schema vs Semantics

List cases where output is valid format but wrong meaning.

Questions to answer:

  • Which validators catch semantic errors?
  • Which require human review?

5.7 The Interview Questions They’ll Ask

  1. “Why is schema validation necessary for tool calls?”
  2. “How do you handle repeated validation failures?”
  3. “What is the trade-off between strict and lenient schemas?”
  4. “How do you test a validator?”
  5. “What is a repair loop?”

5.8 Hints in Layers

Hint 1: Start small Use 3 required fields.

Hint 2: Add semantic checks Validate relationships between fields.

Hint 3: Limit retries Avoid infinite loops.

Hint 4: Log failures Collect examples for tuning.

5.9 Books That Will Help

Topic Book Chapter
Validation Guardrails AI docs Validators 

5.10 Implementation Phases

Phase 1: Schema Design (1-2 days)

Goals: define schema. Tasks: list fields and constraints. Checkpoint: schema complete.

Phase 2: Validation Engine (2-3 days)

Goals: enforce schema. Tasks: validator integration. Checkpoint: invalid outputs detected.

Phase 3: Repair Loop (2 days)

Goals: handle errors. Tasks: retry logic, fallback. Checkpoint: deterministic outcomes.

5.11 Key Implementation Decisions

Decision Options Recommendation Rationale
Retry count 1 vs 3 2 retries balance cost
Fail action block vs fallback fallback better UX

6. Testing Strategy

6.1 Test Categories

Category Purpose Examples
Unit Tests Validate core logic Input classification, schema checks
Integration Tests Validate end-to-end flow Full guardrail pipeline
Edge Case Tests Validate unusual inputs Long prompts, empty outputs

6.2 Critical Test Cases

  1. Missing field: must fail
  2. Invalid type: must fail
  3. Valid output: must pass

6.3 Test Data

Sample outputs with missing and valid fields

7. Common Pitfalls & Debugging

7.1 Frequent Mistakes

Pitfall Symptom Solution
Overly strict schema Many failures Loosen constraints
Too many retries High cost Limit retries

7.2 Debugging Strategies

  • Inspect decision logs to see which rule triggered a block.
  • Replay deterministic test cases to reproduce failures.

7.3 Performance Traps

Avoid repeated validation on the same output without caching.

8. Extensions & Challenges

8.1 Beginner Extensions

  • Add optional fields
  • Add simple semantic validators

8.2 Intermediate Extensions

  • Integrate with tool gating
  • Version schemas

8.3 Advanced Extensions

  • Auto-generate schema from examples
  • Build a schema registry

9. Real-World Connections

9.1 Industry Applications

  • Data extraction pipelines: Guarantees structured outputs
  • Agent tool calls: Ensures safe parameters
  • Guardrails AI: Validator framework 
  • NeMo Guardrails: Flow control complement 

9.3 Interview Relevance

  • Schema design: Contract design reasoning
  • Validation loops: Handling invalid outputs

10. Resources

10.1 Essential Reading

  • Guardrails AI docs. 
  • NeMo Guardrails docs. 

10.2 Video Resources

  • Structured output design talks
  • LLM schema validation demos

10.3 Tools & Documentation

  • Guardrails AI docs. 
  • NeMo Guardrails docs. 
  • Project 6: Tool-Use Permissioning
  • Project 8: Policy Router Orchestrator

11. Self-Assessment Checklist

11.1 Understanding

  • I can explain the control plane layers without notes
  • I can justify every policy threshold used
  • I understand the main failure modes of this guardrail

11.2 Implementation

  • All functional requirements are met
  • All critical test cases pass
  • Edge cases are handled

11.3 Growth

  • I documented lessons learned
  • I can explain this project in an interview

12. Submission / Completion Criteria

Minimum Viable Completion:

  • Schema validation implemented
  • Repair policy defined
  • Validation failures logged

Full Completion:

  • All minimum criteria plus:
  • Schema failure rate measured
  • Semantic validators added

Excellence (Going Above & Beyond):

  • Schema registry built
  • Automated schema testing