Add filter option (#45)

Default list is long now. This would let you do something like

```
registry.filter(Type="RetrievalTask")
```

langchain_benchmarks/rag/tasks/langchain_docs/indexing/retriever_registry.py

@@ -7,6 +7,9 @@
 from langchain.schema.retriever import BaseRetriever
 from langchain.vectorstores.chroma import Chroma
 
+from langchain_benchmarks.rag.utils._downloading import (
+    fetch_remote_file,
+)
 from langchain_benchmarks.rag.utils.indexing import (
     get_hyde_retriever,
     get_parent_document_retriever,
@@ -30,6 +33,10 @@ def load_docs_from_parquet(filename: Optional[str] = None) -> Iterable[Document]
             "Please install pandas to use the langchain docs benchmarking task.\n"
             "pip install pandas"
         )
+    if filename is None:
+        filename = DOCS_FILE
+    if not os.path.exists(filename):
+        fetch_remote_file(REMOTE_DOCS_FILE, filename)
     df = pd.read_parquet(filename)
     docs_transformed = [Document(**row) for row in df.to_dict(orient="records")]
     for doc in docs_transformed:
@@ -43,11 +50,13 @@ def load_docs_from_parquet(filename: Optional[str] = None) -> Iterable[Document]
 
 def _chroma_retriever_factory(
     embedding: Embeddings,
+    *,
+    docs: Optional[Iterable[Document]] = None,
     search_kwargs: Optional[dict] = None,
     transform_docs: Optional[Callable] = None,
     transformation_name: Optional[str] = None,
 ) -> BaseRetriever:
-    docs = load_docs_from_parquet(DOCS_FILE)
+    docs = docs or load_docs_from_parquet()
     embedding_name = embedding.__class__.__name__
     vectorstore = Chroma(
         collection_name=f"langchain-benchmarks-classic-{embedding_name}",
@@ -67,9 +76,11 @@ def _chroma_retriever_factory(
 
 def _chroma_parent_document_retriever_factory(
     embedding: Embeddings,
+    *,
+    docs: Optional[Iterable[Document]] = None,
     search_kwargs: Optional[dict] = None,
 ) -> BaseRetriever:
-    docs = load_docs_from_parquet(DOCS_FILE)
+    docs = docs or load_docs_from_parquet()
     embedding_name = embedding.__class__.__name__
     vectorstore = Chroma(
         collection_name=f"langchain-benchmarks-parent-doc-{embedding_name}",
@@ -87,9 +98,11 @@ def _chroma_parent_document_retriever_factory(
 
 def _chroma_hyde_retriever_factory(
     embedding: Embeddings,
+    *,
+    docs: Optional[Iterable[Document]] = None,
     search_kwargs: Optional[dict] = None,
 ) -> BaseRetriever:
-    docs = load_docs_from_parquet(DOCS_FILE)
+    docs = docs or load_docs_from_parquet()
     embedding_name = embedding.__class__.__name__
     vectorstore = Chroma(
         collection_name=f"langchain-benchmarks-hyde-{embedding_name}",

langchain_benchmarks/rag/tasks/langchain_docs/task.py

@@ -1,4 +1,6 @@
-from functools import partial
+from typing import Iterable
+
+from langchain.schema.document import Document
 
 from langchain_benchmarks.rag.tasks.langchain_docs import architectures, indexing
 from langchain_benchmarks.rag.tasks.langchain_docs.indexing.retriever_registry import (
@@ -11,12 +13,18 @@
     "452ccafc-18e1-4314-885b-edd735f17b9d"  # ID of public LangChain Docs dataset
 )
 
+
+def load_cached_docs() -> Iterable[Document]:
+    """Load the docs from the cached file."""
+    return load_docs_from_parquet(DOCS_FILE)
+
+
 LANGCHAIN_DOCS_TASK = RetrievalTask(
     name="LangChain Docs Q&A",
     dataset_id=DATASET_ID,
     retriever_factories=indexing.RETRIEVER_FACTORIES,
     architecture_factories=architectures.ARCH_FACTORIES,
-    get_docs=partial(load_docs_from_parquet, DOCS_FILE),
+    get_docs=load_cached_docs,
     description=(
         """\
 Questions and answers based on a snapshot of the LangChain python docs.

langchain_benchmarks/rag/tasks/semi_structured_earnings/indexing/retriever_registry.py

@@ -126,6 +126,7 @@ def _chroma_parent_document_retriever_factory(
     *,
     docs: Optional[Iterable[Document]] = None,
     search_kwargs: Optional[dict] = None,
+    transformation_name: Optional[str] = None,
 ) -> BaseRetriever:
     docs = docs or load_docs()
     embedding_name = embedding.__class__.__name__
@@ -140,6 +141,7 @@ def _chroma_parent_document_retriever_factory(
         vectorstore,
         collection_name="semi-structured-earnings",
         search_kwargs=search_kwargs or _DEFAULT_SEARCH_KWARGS,
+        transformation_name=transformation_name,
     )
 
 
@@ -148,6 +150,7 @@ def _chroma_hyde_retriever_factory(
     *,
     docs: Optional[Iterable[Document]] = None,
     search_kwargs: Optional[dict] = None,
+    transformation_name: Optional[str] = None,
 ) -> BaseRetriever:
     docs = docs or load_docs()
     embedding_name = embedding.__class__.__name__
@@ -162,6 +165,7 @@ def _chroma_hyde_retriever_factory(
         vectorstore,
         collection_name="semi-structured-earnings",
         search_kwargs=search_kwargs or _DEFAULT_SEARCH_KWARGS,
+        transformation_name=transformation_name,
     )
 
 

langchain_benchmarks/schema.py

@@ -1,5 +1,8 @@
 """Schema for the Langchain Benchmarks."""
+from __future__ import annotations
+
 import dataclasses
+import inspect
 from typing import Any, Callable, Dict, Iterable, List, Optional, Type, Union
 
 from langchain.prompts import ChatPromptTemplate
@@ -104,6 +107,7 @@ def _table(self) -> List[List[str]]:
         return table + [
             ["Retriever Factories", ", ".join(self.retriever_factories.keys())],
             ["Architecture Factories", ", ".join(self.architecture_factories.keys())],
+            ["get_docs", self.get_docs],
         ]
 
 
@@ -149,6 +153,29 @@ def _repr_html_(self) -> str:
         ]
         return tabulate(table, headers=headers, tablefmt="html")
 
+    def filter(
+        self,
+        Type: Optional[str],
+        dataset_id: Optional[str] = None,
+        name: Optional[str] = None,
+        description: Optional[str] = None,
+    ) -> Registry:
+        """Filter the tasks in the registry."""
+        tasks = self.tasks
+        if Type is not None:
+            tasks = [task for task in tasks if task.__class__.__name__ == Type]
+        if dataset_id is not None:
+            tasks = [task for task in tasks if task.dataset_id == dataset_id]
+        if name is not None:
+            tasks = [task for task in tasks if task.name == name]
+        if description is not None:
+            tasks = [
+                task
+                for task in tasks
+                if description.lower() in task.description.lower()
+            ]
+        return Registry(tasks=tasks)
+
     def __getitem__(self, key: Union[int, str]) -> BaseTask:
         """Get an environment from the registry."""
         if isinstance(key, slice):