Skip to content

intro.mdx

Latest commit

 

History

History
158 lines (130 loc) · 5.6 KB

intro.mdx

File metadata and controls

158 lines (130 loc) · 5.6 KB
title QWED: deterministic verification for LLMs and AI agents
sidebarTitle Introduction
description QWED is a deterministic verification platform for LLMs and AI agents using formal methods, symbolic execution, and policy guards.
icon shield-check
**QWED v5.1.1** is now live — Trust boundary hardening: cache context binding, attestation fail-closed, and proof-path corrections. [See what's new →](/changelog#v5-1-1-—-trust-boundary-hardening)

What is QWED?

QWED (Query With Evidence & Determinism) is a trust boundary for AI systems:

  • LLMs can translate user intent into structured claims.
  • QWED verifies those claims with deterministic engines before execution or response.
  • You get proof-backed outcomes instead of probability-only confidence.

QWED is designed for LLM verification, AI agent security, verified tool calls, prompt injection defense, and deterministic transaction verification in high-stakes workflows.

"Do not trust generated output. Verify it."

flowchart LR
    U[User Query] --> L[LLM Translation]
    L --> Q[QWED Verification]
    Q --> R[Verified Result]

    L -. Untrusted output .-> X[Possible hallucination]
    Q -. Deterministic proof .-> Y[Accepted or rejected]

    classDef untrusted fill:#fff4e5,stroke:#f59e0b,color:#92400e;
    classDef trusted fill:#ecfeff,stroke:#06b6d4,color:#155e75;
    classDef result fill:#ecfdf5,stroke:#22c55e,color:#166534;
    class L,X untrusted;
    class Q,Y trusted;
    class R result;
Loading

Explore common verification problems

Learn how formal verification for LLMs differs from prompting, RAG, and output formatting. Add pre-execution checks, policy enforcement, and budget controls for agent actions. Harden your stack against prompt injection, exfiltration, and unsafe execution paths. Secure Model Context Protocol tools with deterministic verification and tool schema checks.

When to use QWED first

Verify equations, constraints, and logical claims before they reach users. Catch unsafe patterns, injection risks, and structural errors before execution. Inspect actions and payloads before external systems are touched. Add deterministic checkpoints for finance, legal, tax, and regulated systems.

Quick start (5 minutes)

pip install qwed
docker pull qwedai/qwed-verification:latest
# Verify a simple claim
qwed verify "Is 2+2=5?"
# -> CORRECTED: The answer is 4

# Verify logic constraints
qwed verify-logic "(AND (GT x 5) (LT y 10))"
# -> SAT: {x=6, y=9}
Follow [Installation](/getting-started/installation) and set your provider with [LLM configuration](/getting-started/llm-configuration). Use [Quick start](/getting-started/quickstart) to validate math, logic, code, and SQL. Use [Integration getting started](/integration/getting-started) and [Production deployment](/integration/production).

Verification engines at a glance

SymPy-based symbolic verification. Z3 SAT/SMT verification with models. AST and symbolic checks for risky behavior. Parser-backed SQL safety and validation. Type and shape validation for structured outputs. Data-flow tracking for untrusted inputs. Explore core, analysis, and specialized engines.

What's new in v5.1.1

Deterministic structural and semantic verification of agent state payloads with governed atomic commits. [Learn more →](/advanced/agent-state-guard) Legacy CodeExecutor blocked, unknown tools default-denied, bounded math tolerance, ambiguous expressions rejected, schema uniqueItems fail-closed, and identity sampling rejected. LOOP-004 state-aware replay protection detects agents repeating actions on unchanged world state.

Recommended learning path

  1. Core concepts
  2. Architecture overview
  3. SDKs overview
  4. API overview
  5. Integration guide