q

langchain/chains/research/api.py

@@ -85,10 +85,12 @@ class Research(Chain):
         urls = search_results["urls"]
         blobs = self.downloader.download(urls)
         parser = MarkdownifyHTMLParser()
-        docs = itertools.chain.from_iterable(parser.lazy_parse(blob) for blob in blobs)
-        inputs = [{"doc": doc, "question": question} for doc in docs]
+        docs = itertools.chain.from_iterable(
+            parser.lazy_parse(blob) for blob in blobs if blob is not None
+        )
+        _inputs = [{"doc": doc, "question": question} for doc in docs]
         results = self.reader(
-            inputs, callbacks=run_manager.get_child() if run_manager else None
+            _inputs, callbacks=run_manager.get_child() if run_manager else None
         )
         return {
             "docs": [result["answer"] for result in results["inputs"]],
@@ -111,9 +113,9 @@ class Research(Chain):
         docs = itertools.chain.from_iterable(
             parser.lazy_parse(blob) for blob in blobs if blob is not None
         )
-        inputs = [{"doc": doc, "question": question} for doc in docs]
+        _inputs = [{"doc": doc, "question": question} for doc in docs]
         results = await self.reader.acall(
-            inputs,
+            _inputs,
             callbacks=run_manager.get_child() if run_manager else None,
         )
         return {
@@ -157,7 +159,9 @@ class Research(Chain):
         if isinstance(text_splitter, str):
             if text_splitter == "recursive":
                 _text_splitter_kwargs = text_splitter_kwargs or {}
-                _text_splitter = RecursiveCharacterTextSplitter(**_text_splitter_kwargs)
+                _text_splitter: TextSplitter = RecursiveCharacterTextSplitter(
+                    **_text_splitter_kwargs
+                )
             else:
                 raise ValueError(f"Invalid text splitter: {text_splitter}")
         elif isinstance(text_splitter, TextSplitter):
@@ -167,7 +171,7 @@ class Research(Chain):
 
         if isinstance(download_handler, str):
             if download_handler == "auto":
-                _download_handler = AutoDownloadHandler()
+                _download_handler: DownloadHandler = AutoDownloadHandler()
             else:
                 raise ValueError(f"Invalid download handler: {download_handler}")
         elif isinstance(download_handler, DownloadHandler):

langchain/chains/research/crawling.py

@@ -79,6 +79,8 @@ class ChainCrawler(BlobCrawler):
 
     def crawl(self, blob: Blob, question: str) -> List[str]:
         """Explore the blob and suggest additional content to explore."""
+        if not blob.source:
+            raise NotImplementedError()
         records, columns = _extract_records(blob)
 
         result = self.chain(

langchain/chains/research/search.py

@@ -2,6 +2,7 @@
 from __future__ import annotations
 
 import asyncio
+import typing
 from typing import Any, Dict, List, Mapping, Optional, Sequence
 
 from bs4 import BeautifulSoup
@@ -257,8 +258,12 @@ class GenericSearcher(Chain):
         run_manager: Optional[CallbackManagerForChainRun] = None,
     ) -> Dict[str, Any]:
         question = inputs["question"]
-        queries = self.query_generator.predict_and_parse(
-            callbacks=run_manager.get_child(), question=question
+        queries = typing.cast(
+            List[str],
+            self.query_generator.predict_and_parse(
+                callbacks=run_manager.get_child() if run_manager else None,
+                question=question,
+            ),
         )
         results = _run_searches(queries, top_k=self.top_k_per_search)
         deuped_results = _deduplicate_objects(results, "link")
@@ -271,7 +276,7 @@ class GenericSearcher(Chain):
                 "question": question,
                 "choices": records,
             },
-            callbacks=run_manager.get_child(),
+            callbacks=run_manager.get_child() if run_manager else None,
         )
         return {"urls": [result["link"] for result in response_["selected"]]}
 
@@ -281,8 +286,12 @@ class GenericSearcher(Chain):
         run_manager: Optional[AsyncCallbackManagerForChainRun] = None,
     ) -> Dict[str, Any]:
         question = inputs["question"]
-        queries = await self.query_generator.apredict_and_parse(
-            callbacks=run_manager.get_child(), question=question
+        queries = typing.cast(
+            List[str],
+            await self.query_generator.apredict_and_parse(
+                callbacks=run_manager.get_child() if run_manager else None,
+                question=question,
+            ),
         )
         results = _run_searches(queries, top_k=self.top_k_per_search)
         deuped_results = _deduplicate_objects(results, "link")
@@ -295,7 +304,7 @@ class GenericSearcher(Chain):
                 "question": question,
                 "choices": records,
             },
-            callbacks=run_manager.get_child(),
+            callbacks=run_manager.get_child() if run_manager else None,
         )
         return {"urls": [result["link"] for result in response_["selected"]]}
 

tests/unit_tests/chains/research/test_downloader.py

@@ -4,7 +4,7 @@ from langchain.chains.research.download import (
 )
 
 
-def test_is_javascript_required():
+def test_is_javascript_required() -> None:
     """Check whether a given page should be re-downloaded with javascript executed."""
     assert not _is_javascript_required(
         """