Harrison/update docstore (#47)

change docstore interface

examples/react.ipynb

@@ -23,22 +23,17 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Search Scott Derrickson\n",
-      "Could not find [Scott Derrickson]. Similar: ['Scott Derrickson', 'The Black Phone', 'Sinister (film)', 'Doctor Strange (2016 film)', 'The Day the Earth Stood Still (2008 film)', 'The Exorcism of Emily Rose', 'Deliver Us from Evil (2014 film)', 'C. Robert Cargill', 'Scott Teems', 'Sean Harris']\n",
-      "Search Scott Derrickson (film director)\n",
-      "Scott Derrickson (born July 16, 1966) is an American filmmaker. He is best known for directing the films The Exorcism of Emily Rose (2005), The Day the Earth Stood Still (2008), Sinister (2012), Deliver Us from Evil (2014), Doctor Strange (2016), and The Black Phone (2021).\n",
-      "Search Ed Wood\n",
-      "Edward Davis Wood Jr.  (October 10, 1924 – December 10, 1978) was an American filmmaker, \n",
-      "actor, and pulp novel author.\n",
-      "In the 1950s, Wood directed several low-budget science fiction, crime and horror films that later became cult classics, notably Glen or Glenda (1953), Jail Bait (1954), Bride of the Monster (1955), Plan 9 from Outer Space (1957) and Night of the Ghouls (1959). In the 1960s and 1970s, he moved towards sexploitation and pornographic films such as  The Sinister Urge (1960), Orgy of the Dead (1965) and Necromania (1971), and wrote over 80 pulp crime and sex novels.\n",
-      "Notable for their campy aesthetics, technical errors, unsophisticated special effects, use of poorly-matched stock footage, eccentric casts, idiosyncratic stories and non sequitur dialogue, Wood's films remained largely obscure until he was posthumously awarded a Golden Turkey Award for Worst Director of All Time in 1980, renewing public interest in his life and work.Following the publication of Rudolph Grey's 1992 oral biography Nightmare of Ecstasy: The Life and Art of Edward D. Wood Jr., a biopic of his life, Ed Wood (1994), was directed by Tim Burton. Starring Johnny Depp as Wood and Martin Landau as Bela Lugosi, the film received critical acclaim and various awards, including two Academy Awards.\n",
-      "Finish yes\n"
+      "Search David Chanoff\n",
+      "David Chanoff is a noted author of non-fiction work. His work has typically involved collaborations with the principal protagonist of the work concerned. His collaborators have included; Augustus A. White, Joycelyn Elders, Đoàn Văn Toại, William J. Crowe, Ariel Sharon, Kenneth Good and Felix Zandman. He has also written about a wide range of subjects including literary history, education and foreign for The Washington Post, The New Republic and The New York Times Magazine. He has published more than twelve books.\n",
+      "Search William J. Crowe\n",
+      "William James Crowe Jr. (January 2, 1925 – October 18, 2007) was a United States Navy admiral and diplomat who served as the 11th chairman of the Joint Chiefs of Staff under Presidents Ronald Reagan and George H. W. Bush, and as the ambassador to the United Kingdom and Chair of the Intelligence Oversight Board under President Bill Clinton.\n",
+      "Finish Bill Clinton\n"
      ]
     },
     {
      "data": {
       "text/plain": [
-       "'yes'"
+       "'Bill Clinton'"
       ]
      },
      "execution_count": 2,
@@ -47,7 +42,7 @@
     }
    ],
    "source": [
-    "question = \"Were Scott Derrickson and Ed Wood of the same nationality?\"\n",
+    "question = \"Author David Chanoff has collaborated with a U.S. Navy admiral who served as the ambassador to the United Kingdom under which President?\"\n",
     "react.run(question)"
    ]
   },

langchain/chains/react/base.py

@@ -8,6 +8,7 @@ from langchain.chains.base import Chain
 from langchain.chains.llm import LLMChain
 from langchain.chains.react.prompt import PROMPT
 from langchain.docstore.base import Docstore
+from langchain.docstore.document import Document
 from langchain.llms.base import LLM
 
 
@@ -85,7 +86,13 @@ class ReActChain(Chain, BaseModel):
             prompt += ret_text
             print(action, directive)
             if action == "Search":
-                observation, document = self.docstore.search(directive)
+                result = self.docstore.search(directive)
+                if isinstance(result, Document):
+                    document = result
+                    observation = document.summary
+                else:
+                    document = None
+                    observation = result
                 print(observation)
             elif action == "Lookup":
                 if document is None:

langchain/docstore/base.py

@@ -1,6 +1,6 @@
 """Interface to access to place that stores documents."""
 from abc import ABC, abstractmethod
-from typing import Optional, Tuple
+from typing import Union
 
 from langchain.docstore.document import Document
 
@@ -9,7 +9,7 @@ class Docstore(ABC):
     """Interface to access to place that stores documents."""
 
     @abstractmethod
-    def search(self, search: str) -> Tuple[str, Optional[Document]]:
+    def search(self, search: str) -> Union[str, Document]:
         """Search for document.
 
         If page exists, return the page summary, and a Document object.

langchain/docstore/wikipedia.py

@@ -1,7 +1,7 @@
 """Wrapper around wikipedia API."""
 
 
-from typing import Optional, Tuple
+from typing import Union
 
 from langchain.docstore.base import Docstore
 from langchain.docstore.document import Document
@@ -20,7 +20,7 @@ class Wikipedia(Docstore):
                 "Please it install it with `pip install wikipedia`."
             )
 
-    def search(self, search: str) -> Tuple[str, Optional[Document]]:
+    def search(self, search: str) -> Union[str, Document]:
         """Try to search for wiki page.
 
         If page exists, return the page summary, and a PageWithLookups object.
@@ -30,16 +30,13 @@ class Wikipedia(Docstore):
 
         try:
             page_content = wikipedia.page(search).content
-            wiki_page = Document(page_content=page_content)
-            observation = wiki_page.summary
+            result: Union[str, Document] = Document(page_content=page_content)
         except wikipedia.PageError:
-            wiki_page = None
-            observation = (
+            result = (
                 f"Could not find [{search}]. " f"Similar: {wikipedia.search(search)}"
             )
         except wikipedia.DisambiguationError:
-            wiki_page = None
-            observation = (
+            result = (
                 f"Could not find [{search}]. " f"Similar: {wikipedia.search(search)}"
             )
-        return observation, wiki_page
+        return result

tests/integration_tests/chains/test_react.py

@@ -1,13 +1,18 @@
 """Integration test for self ask with search."""
 
 from langchain.chains.react.base import ReActChain
+from langchain.docstore.wikipedia import Wikipedia
 from langchain.llms.openai import OpenAI
 
 
 def test_react() -> None:
     """Test functionality on a prompt."""
     llm = OpenAI(temperature=0)
-    react = ReActChain(llm=llm)
-    question = "Were Scott Derrickson and Ed Wood of the same nationality?"
+    react = ReActChain(llm=llm, docstore=Wikipedia())
+    question = (
+        "Author David Chanoff has collaborated with a U.S. Navy admiral "
+        "who served as the ambassador to the United Kingdom under "
+        "which President?"
+    )
     output = react.run(question)
-    assert output == "yes"
+    assert output == "Bill Clinton"

tests/unit_tests/chains/test_react.py

@@ -1,6 +1,6 @@
 """Unit tests for ReAct."""
 
-from typing import List, Optional, Tuple
+from typing import List, Optional, Union
 
 import pytest
 
@@ -39,10 +39,10 @@ class FakeListLLM(LLM):
 class FakeDocstore(Docstore):
     """Fake docstore for testing purposes."""
 
-    def search(self, search: str) -> Tuple[str, Optional[Document]]:
+    def search(self, search: str) -> Union[str, Document]:
         """Return the fake document."""
         document = Document(page_content=_PAGE_CONTENT)
-        return document.summary, document
+        return document
 
 
 def test_predict_until_observation_normal() -> None: