{ "cells": [ { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Installing hub:\u001b[35m/\u001b[0m\u001b[35m/guardrails/\u001b[0m\u001b[95mtwo_words...\u001b[0m\n", "✅Successfully installed guardrails/two_words version \u001b[1;36m0.0\u001b[0m.\u001b[1;36m0\u001b[0m!\n", "\n", "\n" ] } ], "source": [ "! guardrails hub install hub://guardrails/two_words --quiet" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Input Validation\n", "\n", "Guardrails supports validating inputs (messages) with string validators." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "In XML, specify the validators on the `messages` tag, as such:" ] }, { "cell_type": "code", "execution_count": 2, "metadata": { "is_executing": true }, "outputs": [], "source": [ "from guardrails import Guard\n", "\n", "rail_spec = \"\"\"\n", "\n", "\n", "\n", "This is not two words\n", "\n", "\n", "\n", "\n", "\n", "\n", "\"\"\"\n", "\n", "guard = Guard.for_rail_string(rail_spec)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "When `fix` is specified as the on-fail handler, the prompt will automatically be amended before calling the LLM.\n", "\n", "In any other case (for example, `exception`), a `ValidationError` will be returned in the outcome." ] }, { "cell_type": "code", "execution_count": 3, "metadata": { "is_executing": true }, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "/Users/calebcourier/Projects/guardrails/docs/.venv/lib/python3.10/site-packages/guardrails/validator_service/__init__.py:75: UserWarning: Could not obtain an event loop. Falling back to synchronous validation.\n", " warnings.warn(\n" ] } ], "source": [ "from guardrails.errors import ValidationError\n", "\n", "# Add your OPENAI_API_KEY as an environment variable if it's not already set\n", "# import os\n", "# os.environ[\"OPENAI_API_KEY\"] = \"YOUR_API_KEY\"\n", "\n", "try:\n", " guard(model=\"gpt-4o\")\n", "except ValidationError as e:\n", " print(e)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "When using pydantic to initialize a `Guard`, input validators can be specified by composition, as such:" ] }, { "cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Validation failed for field with errors: Value must be exactly two words\n" ] } ], "source": [ "from guardrails.hub import TwoWords\n", "from pydantic import BaseModel\n", "\n", "\n", "class Pet(BaseModel):\n", " name: str\n", " age: int\n", "\n", "\n", "guard = Guard.for_pydantic(Pet)\n", "guard.use(TwoWords(on_fail=\"exception\"), on=\"messages\")\n", "\n", "try:\n", " guard(\n", " model=\"gpt-4o\",\n", " messages=[{\"role\": \"user\", \"content\": \"This is not two words\"}],\n", " )\n", "except ValidationError as e:\n", " print(e)" ] } ], "metadata": { "kernelspec": { "display_name": ".venv (3.10.16)", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.10.16" } }, "nbformat": 4, "nbformat_minor": 1 }