try make guarded/retriable output parser an instance of parser

docs/modules/prompts/examples/guarded_output_parsers.ipynb

@@ -3,7 +3,7 @@
   {
    "cell_type": "code",
    "execution_count": 1,
-   "id": "10d915e4",
+   "id": "4e75cce5",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -13,7 +13,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "c09a2430",
+   "id": "a930f49b",
    "metadata": {},
    "source": [
     "# Guarded Output Parsers\n",
@@ -22,15 +22,15 @@
     "\n",
     "Unfortunately, small models to date don't have capacity to generate well-formed, schema-adherent json while large models still sometimes fail.\n",
     "\n",
-    "In this notebook, we showcase a `GuardedOutputParser` which can be dropped in for an `OutputParser` in an `LLMChain`. It will catch errors at parsing time and try resolve them, initially by re-invoking an LLM.\n",
+    "In this notebook, we showcase a \"Guarded\" `OutputParser` which can be dropped in for an `OutputParser` in an `LLMChain`. It will catch errors at parsing time and try resolve them, initially by re-invoking an LLM\n",
     "\n",
     "Below are some examples:"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 2,
-   "id": "3db24806",
+   "execution_count": 1,
+   "id": "8e6bd7ef",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -38,7 +38,7 @@
     "from typing import List\n",
     "\n",
     "from langchain.chains import LLMChain\n",
-    "from langchain.guardrails.parsing import GuardedOutputParser\n",
+    "from langchain.guardrails.parsing import RetriableOutputParser\n",
     "from langchain.llms import OpenAI\n",
     "from langchain.output_parsers import PydanticOutputParser, OutputParserException\n",
     "from langchain.prompts import PromptTemplate"
@@ -46,7 +46,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "8d40796c",
+   "id": "8bc39653",
    "metadata": {},
    "source": [
     "## 1st example"
@@ -54,8 +54,8 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 3,
-   "id": "eec61f2d",
+   "execution_count": 2,
+   "id": "01337598",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -69,8 +69,8 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 4,
-   "id": "42b8f744",
+   "execution_count": 3,
+   "id": "618dca6a",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -89,8 +89,8 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 6,
-   "id": "67a1d356",
+   "execution_count": 4,
+   "id": "6ccff983",
    "metadata": {},
    "outputs": [
     {
@@ -112,7 +112,7 @@
       "\u001b[0m\n",
       "Dang!\n",
       "Failed to parse FloatArray from completion \n",
-      "Fiboacci is a sequence of numbers defined by the Fibonacci sequence: 0, 1, 1, 2, 3, 5, 8, 13, 21.. Got: Expecting value: line 1 column 1 (char 0)\n"
+      "Fiboacci is a sequence of numbers that are the sum of the previous two numbers in the sequence.. Got: Expecting value: line 1 column 1 (char 0)\n"
      ]
     }
    ],
@@ -132,8 +132,8 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 7,
-   "id": "c568ebbb",
+   "execution_count": 5,
+   "id": "3c68ad59",
    "metadata": {},
    "outputs": [
     {
@@ -160,17 +160,17 @@
     {
      "data": {
       "text/plain": [
-       "FloatArray(values=[0.0, 1.0, 1.0, 2.0, 3.0, 5.0, 8.0, 13.0, 21.0, 34.0, 55.0, 89.0, 144.0])"
+       "FloatArray(values=[1.0, 2.0, 3.0, 5.0, 8.0, 13.0, 21.0, 34.0, 55.0, 89.0])"
       ]
      },
-     "execution_count": 7,
+     "execution_count": 5,
      "metadata": {},
      "output_type": "execute_result"
     }
    ],
    "source": [
     "# We can replace the parser with a guarded parser that tries to fix errors with a bigger model.\n",
-    "guarded_parser = GuardedOutputParser(\n",
+    "guarded_parser = RetriableOutputParser(\n",
     "    parser=parser, retry_llm=OpenAI(model_name=\"text-davinci-003\"))\n",
     "prompt.output_parser = guarded_parser\n",
     "\n",
@@ -179,27 +179,11 @@
   },
   {
    "cell_type": "markdown",
-   "id": "aa0e7b3c",
+   "id": "f5fc247e",
    "metadata": {},
    "source": [
     "This example is demonstrative though. If your goal is to generate data structures, probably you'll want to start a large enough model."
    ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "3edbc886",
-   "metadata": {},
-   "outputs": [],
-   "source": []
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "d83cfb61",
-   "metadata": {},
-   "outputs": [],
-   "source": []
   }
  ],
  "metadata": {

langchain/chains/llm.py

@@ -15,8 +15,6 @@ from langchain.schema import (
     LLMResult,
     PromptValue,
 )
-from langchain.output_parsers import BaseOutputParser
-from langchain.guardrails.parsing import GuardedOutputParser
 
 
 class LLMChain(Chain, BaseModel):
@@ -140,13 +138,7 @@ class LLMChain(Chain, BaseModel):
         """Get the final output from a list of generations for a prompt."""
         completion = generations[0].text
         if self.prompt.output_parser:
-            parser = self.prompt.output_parser
-            if isinstance(parser, BaseOutputParser):
-                completion = parser.parse(completion)
-            elif isinstance(parser, GuardedOutputParser):
-                # TODO: not ideal to hide retry calls from user. can we expose it in colored log?
-                completion = parser.parse(prompt_value, completion)
-
+            completion = self.prompt.output_parser.parse_with_prompt(completion, prompt_value)
         return completion
 
     def create_outputs(

langchain/guardrails/parsing.py

@@ -1,4 +1,3 @@
-from pydantic import BaseModel
 from typing import Any
 
 from langchain.guardrails.retry import naive_retry
@@ -6,13 +5,13 @@ from langchain.output_parsers import BaseOutputParser, OutputParserException
 from langchain.schema import BaseLanguageModel, PromptValue
 
 
-class GuardedOutputParser(BaseModel):
+class RetriableOutputParser(BaseOutputParser):
     """Wraps a parser and tries to fix parsing errors."""
 
     parser: BaseOutputParser
     retry_llm: BaseLanguageModel
 
-    def parse(self, prompt_value: PromptValue, completion: str) -> Any:
+    def parse_with_prompt(self, completion: str, prompt_value: PromptValue) -> Any:
         try:
             parsed_completion = self.parser.parse(completion)
         except OutputParserException as e:
@@ -22,4 +21,10 @@ class GuardedOutputParser(BaseModel):
             )
             parsed_completion = self.parser.parse(new_completion)
         
-        return parsed_completion
+        return parsed_completion
+
+    def parse(self, completion: str):
+        raise NotImplementedError
+
+    def get_format_instructions(self) -> str:
+        return self.parser.get_format_instructions()

langchain/output_parsers/base.py

@@ -1,10 +1,11 @@
 from __future__ import annotations
 
 from abc import ABC, abstractmethod
-from typing import Any, Dict
+from typing import Any, Dict, Optional
 
 from pydantic import BaseModel
 
+from langchain.schema import PromptValue
 
 class BaseOutputParser(BaseModel, ABC):
     """Class to parse the output of an LLM call."""
@@ -13,6 +14,9 @@ class BaseOutputParser(BaseModel, ABC):
     def parse(self, text: str) -> Any:
         """Parse the output of an LLM call."""
 
+    def parse_with_prompt(self, completion: str, prompt: Optional[PromptValue] = None) -> Any:
+        return self.parse(completion)
+
     def get_format_instructions(self) -> str:
         raise NotImplementedError
 

langchain/schema.py

@@ -6,8 +6,6 @@ from typing import Any, Dict, List, NamedTuple, Optional
 
 from pydantic import BaseModel, Extra, Field, root_validator
 
-from langchain.output_parsers import BaseOutputParser
-
 
 def get_buffer_string(
     messages: List[BaseMessage], human_prefix: str = "Human", ai_prefix: str = "AI"