Tutorials
Practical, text-first guides for ValGuard.ai — validation layer setup, multi-step playbooks, and production operations. New to the platform? Start with Getting started or jump into a topic below.
Recommended path: Validate LLM output → First playbook → Deploy → Monitor flows
Playbooks & workflows
Chain assistants, branches, sandbox testing, and deployment.
How to build your first playbook
Create a draft playbook from a template, assign assistants, publish, and call it with X-VG-Flow in under 15 minutes.
~9 min read
How to chain agents
Link multiple AI assistants in sequence or with branches so each step passes validated output to the next.
~7 min read
How to add fallback logic
Route failed validations to repair steps, alternate assistants, or safe default responses in your playbook.
~8 min read
How to test branches in sandbox mode
Preview playbook paths locally with Simulate, try edge-case messages, and validate routing before going live.
~8 min read
How to run parallel steps with fan-out and merge
Model genuinely parallel work with fan-out and merge nodes — run independent extraction and check steps concurrently in one playbook.
~10 min read
Validation & safety
Deterministic rules, RAG checks, and hallucination guardrails.
How to validate LLM output
Attach deterministic validators to an agent, test in the playground, and enforce JSON schema or compliance rules on every completion.
~7 min read
How to prevent hallucinations
Use grounded-answer validators, citation checks, and field-presence rules to block invented facts before users see them.
~8 min read
How to build a RAG flow with validation
Combine retrieval, answer generation, and post-generation checks so citations and claims match your knowledge base.
~8 min read
How to set up a prompt injection firewall
Enable inbound input validation to block prompt injection and jailbreak attempts before they ever reach the LLM.
~9 min read
How to tune reask loops
Configure on_fail: reask and max_reasks so self-correctable failures retry automatically instead of blocking outright.
~10 min read
How to validate streaming responses
Understand the tradeoff between token-by-token streaming and blocking validation, and configure agents for the right one.
~10 min read
How to test validators in CI/CD
Test validator rules automatically in CI using the internal test API, catching regressions before they reach production.
~11 min read
Operations
Budgets, monitoring, and production rollout.
How to use budget alerts
Set per-agent USD spend caps, choose block or Growth+ degrade modes, configure alerts, and avoid surprise LLM bills.
~9 min read
How to set monthly token limits
Set optional organization and per-agent token caps on Growth+, leave usage uncapped by default, and avoid HTTP 402 when limits are reached.
~7 min read
How to deploy your first agent
Move from shadow mode to enforcement, rotate API keys, and point production traffic through the validation layer.
~7 min read
How to monitor your flows
Track playbook runs, agent health score and league rank, validation failures, score-drop alerts, and weekly digest review in the dashboard.
~10 min read
How to roll out shadow mode safely
Roll validators from shadow mode into full enforcement with a safe, reversible checklist — read would-block metrics before you ever block real traffic.
~9 min read
How to verify guardrail proof
Verify HMAC-signed evidence envelopes in the Proof panel, export an auditor window, and run valguard-cli verify-proof offline without storing raw prompts.
~8 min read
How to configure validation webhooks
Configure validation webhooks, verify HMAC signatures, and get real-time alerts on blocks, shadow events, and orchestration steps.
~10 min read
How to set up provider vault and BYOK
Store provider API keys in Provider Vault or pass them per-request with BYOK — plus how to set up custom model endpoints.
~9 min read
How to configure RBAC and team permissions
Invite teammates, assign owner/admin/member roles, and understand exactly what each role can and cannot do.
~9 min read
How to use Cost control
Use the Cost hub recommendations, spend-by-agent strip, budget-aware playbook phrases, model-price estimates, and optional cheaper-model auto-pick.
~10 min read
How to use self-healing without re-ask
Enable automatic self-healing before validation to reduce retry calls, lower token spend, and keep strict quality gates for critical fields.
~8 min read
Reference docs
- Playbooks (63) — headers, billing, SDK examples
- Validators catalog — all built-in rules
- Quickstart — curl and language SDKs