Smarter Guardrail
=====================

Overview
--------

The Smarter Guardrail app is responsible for enforcing constraints on the interaction
between an application and a large language model. Rather than relying solely on
prompt instructions such as “do not reveal sensitive information” or “respond
only in JSON,” guardrails validate inputs, outputs, and execution state before
information is accepted or returned. They may sanitize user input, reject prompt
injection attempts, enforce schema validation, filter unsafe or out-of-scope responses,
verify citations, limit tool access, or require structured output that conforms to
predefined contracts. Guardrails operate as deterministic
software controls surrounding the probabilistic behavior of the model, ensuring that
business rules and security requirements are enforced independently of the LLM’s reasoning.

Pre-execution guardrails inspect user requests for malicious content, policy violations, or
malformed data before the prompt reaches the model. Runtime guardrails constrain
the model’s available tools, memory, and permissions, preventing unauthorized
actions or excessive autonomy during execution. Post-execution guardrails validate
the generated response against expected formats, confidence thresholds, or
organizational policies before returning it to the caller. Together, these layers
transform an LLM from a general-purpose text generator into a predictable application
component whose behavior can be monitored, audited, and controlled, making it suitable
for production environments where reliability, compliance, and security are essential.

The Smarter Guardrail app is included in v0.15.0 and later.
It shifts management of LLM guardrail API credentials from an IT
and devops responsibility, to the Account administrators.

.. seealso::

    - :doc:`Smarter Installation Guide <../smarter-platform/installation>`
    - :doc:`OpenAI Getting Started Guide <../smarter-framework/guides/openai-api-getting-started-guide>`

Technical Reference
-------------------

.. toctree::
   :maxdepth: 1

   guardrail/api
   guardrail/const
   guardrail/management
   guardrail/manifest
   guardrail/models
   guardrail/serializers
   guardrail/signals
   guardrail/tasks
   guardrail/views