mirror of
https://github.com/hwchase17/langchain.git
synced 2025-05-11 18:16:12 +00:00
bf5193bb99
Upgrade to using a literal for specifying the extra which is the
recommended approach in pydantic 2.
This works correctly also in pydantic v1.
```python
from pydantic.v1 import BaseModel
class Foo(BaseModel, extra="forbid"):
x: int
Foo(x=5, y=1)
```
And
```python
from pydantic.v1 import BaseModel
class Foo(BaseModel):
x: int
class Config:
extra = "forbid"
Foo(x=5, y=1)
```
## Enum -> literal using grit pattern:
```
engine marzano(0.1)
language python
or {
`extra=Extra.allow` => `extra="allow"`,
`extra=Extra.forbid` => `extra="forbid"`,
`extra=Extra.ignore` => `extra="ignore"`
}
```
Resorted attributes in config and removed doc-string in case we will
need to deal with going back and forth between pydantic v1 and v2 during
the 0.3 release. (This will reduce merge conflicts.)
## Sort attributes in Config:
```
engine marzano(0.1)
language python
function sort($values) js {
return $values.text.split(',').sort().join("\n");
}
class_definition($name, $body) as $C where {
$name <: `Config`,
$body <: block($statements),
$values = [],
$statements <: some bubble($values) assignment() as $A where {
$values += $A
},
$body => sort($values),
}
```
106 lines
3.2 KiB
Python
106 lines
3.2 KiB
Python
"""KNN Retriever.
|
|
Largely based on
|
|
https://github.com/karpathy/randomfun/blob/master/knn_vs_svm.ipynb"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import concurrent.futures
|
|
from typing import Any, Iterable, List, Optional
|
|
|
|
import numpy as np
|
|
from langchain_core.callbacks import CallbackManagerForRetrieverRun
|
|
from langchain_core.documents import Document
|
|
from langchain_core.embeddings import Embeddings
|
|
from langchain_core.retrievers import BaseRetriever
|
|
|
|
|
|
def create_index(contexts: List[str], embeddings: Embeddings) -> np.ndarray:
|
|
"""
|
|
Create an index of embeddings for a list of contexts.
|
|
|
|
Args:
|
|
contexts: List of contexts to embed.
|
|
embeddings: Embeddings model to use.
|
|
|
|
Returns:
|
|
Index of embeddings.
|
|
"""
|
|
with concurrent.futures.ThreadPoolExecutor() as executor:
|
|
return np.array(list(executor.map(embeddings.embed_query, contexts)))
|
|
|
|
|
|
class KNNRetriever(BaseRetriever):
|
|
"""`KNN` retriever."""
|
|
|
|
embeddings: Embeddings
|
|
"""Embeddings model to use."""
|
|
index: Any
|
|
"""Index of embeddings."""
|
|
texts: List[str]
|
|
"""List of texts to index."""
|
|
metadatas: Optional[List[dict]] = None
|
|
"""List of metadatas corresponding with each text."""
|
|
k: int = 4
|
|
"""Number of results to return."""
|
|
relevancy_threshold: Optional[float] = None
|
|
"""Threshold for relevancy."""
|
|
|
|
class Config:
|
|
arbitrary_types_allowed = True
|
|
|
|
@classmethod
|
|
def from_texts(
|
|
cls,
|
|
texts: List[str],
|
|
embeddings: Embeddings,
|
|
metadatas: Optional[List[dict]] = None,
|
|
**kwargs: Any,
|
|
) -> KNNRetriever:
|
|
index = create_index(texts, embeddings)
|
|
return cls(
|
|
embeddings=embeddings,
|
|
index=index,
|
|
texts=texts,
|
|
metadatas=metadatas,
|
|
**kwargs,
|
|
)
|
|
|
|
@classmethod
|
|
def from_documents(
|
|
cls,
|
|
documents: Iterable[Document],
|
|
embeddings: Embeddings,
|
|
**kwargs: Any,
|
|
) -> KNNRetriever:
|
|
texts, metadatas = zip(*((d.page_content, d.metadata) for d in documents))
|
|
return cls.from_texts(
|
|
texts=texts, embeddings=embeddings, metadatas=metadatas, **kwargs
|
|
)
|
|
|
|
def _get_relevant_documents(
|
|
self, query: str, *, run_manager: CallbackManagerForRetrieverRun
|
|
) -> List[Document]:
|
|
query_embeds = np.array(self.embeddings.embed_query(query))
|
|
# calc L2 norm
|
|
index_embeds = self.index / np.sqrt((self.index**2).sum(1, keepdims=True))
|
|
query_embeds = query_embeds / np.sqrt((query_embeds**2).sum())
|
|
|
|
similarities = index_embeds.dot(query_embeds)
|
|
sorted_ix = np.argsort(-similarities)
|
|
|
|
denominator = np.max(similarities) - np.min(similarities) + 1e-6
|
|
normalized_similarities = (similarities - np.min(similarities)) / denominator
|
|
|
|
top_k_results = [
|
|
Document(
|
|
page_content=self.texts[row],
|
|
metadata=self.metadatas[row] if self.metadatas else {},
|
|
)
|
|
for row in sorted_ix[0 : self.k]
|
|
if (
|
|
self.relevancy_threshold is None
|
|
or normalized_similarities[row] >= self.relevancy_threshold
|
|
)
|
|
]
|
|
return top_k_results
|