Initialize GuardrailsOutputParser documentation

langchain-ai · Dec 13, 2023 · b34088c · b34088c
1 parent 1989142
commit b34088c
Showing 1 changed file with 226 additions and 0 deletions.
diff --git a/docs/docs/modules/model_io/output_parsers/guardrails.ipynb b/docs/docs/modules/model_io/output_parsers/guardrails.ipynb
@@ -0,0 +1,226 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "a1ae632a",
+   "metadata": {},
+   "source": [
+    "# Guardrails parser\n",
+    "\n",
+    "This output parser invokes [Guardrails](https://docs.guardrailsai.com/) to parse the output of a language model, given a schema defined either in XML or as a pydantic model. Guardrails validates strings, or fields in a JSON, using [its predefined validators](https://docs.guardrailsai.com/api_reference/validators/), or a custom validator you define."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "outputs": [],
+   "source": [
+    "from guardrails import Guard\n",
+    "from guardrails.validators import LowerCase, TwoWords, UpperCase\n",
+    "from langchain.chat_models import ChatOpenAI\n",
+    "from langchain.output_parsers import GuardrailsOutputParser\n",
+    "from langchain.prompts import PromptTemplate\n",
+    "from pydantic import BaseModel, Field"
+   ],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "end_time": "2023-12-13T17:40:28.263352Z",
+     "start_time": "2023-12-13T17:40:28.258570Z"
+    }
+   }
+  },
+  {
+   "cell_type": "markdown",
+   "source": [
+    "## With strings"
+   ],
+   "metadata": {
+    "collapsed": false
+   }
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "outputs": [
+    {
+     "data": {
+      "text/plain": "'colorful comforts'"
+     },
+     "execution_count": 11,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "guard = Guard.from_string([LowerCase(on_fail=\"fix\")])\n",
+    "\n",
+    "guardrails_parser = GuardrailsOutputParser.from_guard(guard)\n",
+    "\n",
+    "prompt = PromptTemplate.from_template(\n",
+    "    \"What is a good name for a company that makes {product}?\"\n",
+    ")\n",
+    "\n",
+    "runnable = prompt | ChatOpenAI() | guardrails_parser\n",
+    "\n",
+    "runnable.invoke({\"product\": \"colorful socks\"})"
+   ],
+   "metadata": {
+    "collapsed": false,
+    "ExecuteTime": {
+     "end_time": "2023-12-13T17:40:29.345599Z",
+     "start_time": "2023-12-13T17:40:28.664661Z"
+    }
+   }
+  },
+  {
+   "cell_type": "markdown",
+   "source": [
+    "## With pydantic"
+   ],
+   "metadata": {
+    "collapsed": false
+   }
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 20,
+   "id": "b3f16168",
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2023-12-13T17:42:45.201398Z",
+     "start_time": "2023-12-13T17:42:44.058209Z"
+    }
+   },
+   "outputs": [
+    {
+     "data": {
+      "text/plain": "{'pet_type': 'DOG', 'name': 'Buddy'}"
+     },
+     "execution_count": 20,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "# Here's an example of a Guard constructed from a pydantic model\n",
+    "\n",
+    "\n",
+    "class Pet(BaseModel):\n",
+    "    pet_type: str = Field(validators=[UpperCase(on_fail=\"fix\")])\n",
+    "    name: str = Field(validators=[TwoWords(on_fail=\"fix\")])\n",
+    "\n",
+    "\n",
+    "prompt = \"\"\"\n",
+    "    What kind of pet should I get and what should I name it?\n",
+    "\n",
+    "    ${gr.complete_json_suffix_v2}\n",
+    "\"\"\"\n",
+    "\n",
+    "guard = Guard.from_pydantic(Pet, prompt=prompt)\n",
+    "parser = GuardrailsOutputParser.from_guard(guard)\n",
+    "\n",
+    "prompt_template = PromptTemplate(\n",
+    "    template=\"Answer the user query.\\n{format_instructions}\\n{prompt}\\n\",\n",
+    "    input_variables=[\"prompt\"],\n",
+    "    partial_variables={\"format_instructions\": parser.get_format_instructions()},\n",
+    ")\n",
+    "\n",
+    "runnable = prompt_template | ChatOpenAI() | parser\n",
+    "\n",
+    "runnable.invoke(\n",
+    "    {\n",
+    "        \"product\": \"colorful socks\",\n",
+    "        \"prompt\": prompt,\n",
+    "    }\n",
+    ")"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "fd8c3347-4d0b-459b-ab7b-cf5443297026",
+   "metadata": {
+    "collapsed": false
+   },
+   "source": [
+    "### With XML"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 19,
+   "id": "03049f88",
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2023-12-13T17:42:35.469586Z",
+     "start_time": "2023-12-13T17:42:34.578496Z"
+    }
+   },
+   "outputs": [
+    {
+     "data": {
+      "text/plain": "{'pet_type': 'DOG', 'name': 'Buddy'}"
+     },
+     "execution_count": 19,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "# Here's another example, of a Guard constructed from an XML RAIL schema\n",
+    "\n",
+    "rail_spec = \"\"\"\n",
+    "<rail version=\"0.1\">\n",
+    "<output>\n",
+    "    <string name=\"pet_type\" validators=\"uppercase\" />\n",
+    "    <string name=\"name\" validators=\"two-words\" />\n",
+    "</output>\n",
+    "<prompt>\n",
+    "    What kind of pet should I get and what should I name it?\n",
+    "    ${gr.complete_json_suffix_v2}\n",
+    "</prompt>\n",
+    "</rail>\n",
+    "\"\"\"\n",
+    "\n",
+    "guard = Guard.from_rail_string(rail_spec)\n",
+    "parser = GuardrailsOutputParser.from_guard(guard)\n",
+    "\n",
+    "prompt_template = PromptTemplate(\n",
+    "    template=\"Answer the user query.\\n{format_instructions}\\n{prompt}\\n\",\n",
+    "    input_variables=[\"prompt\"],\n",
+    "    partial_variables={\"format_instructions\": parser.get_format_instructions()},\n",
+    ")\n",
+    "\n",
+    "runnable = prompt_template | ChatOpenAI() | parser\n",
+    "\n",
+    "runnable.invoke(\n",
+    "    {\n",
+    "        \"product\": \"colorful socks\",\n",
+    "        \"prompt\": prompt,\n",
+    "    }\n",
+    ")"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "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": "3.11.3"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}