Spaces:
Runtime error
Runtime error
File size: 4,009 Bytes
ed4d993 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 |
import logging
from typing import List, Optional
import requests
from langchain_core.embeddings import Embeddings
from langchain_core.pydantic_v1 import BaseModel
logger = logging.getLogger(__name__)
class LlamafileEmbeddings(BaseModel, Embeddings):
"""Llamafile lets you distribute and run large language models with a
single file.
To get started, see: https://github.com/Mozilla-Ocho/llamafile
To use this class, you will need to first:
1. Download a llamafile.
2. Make the downloaded file executable: `chmod +x path/to/model.llamafile`
3. Start the llamafile in server mode with embeddings enabled:
`./path/to/model.llamafile --server --nobrowser --embedding`
Example:
.. code-block:: python
from langchain_community.embeddings import LlamafileEmbeddings
embedder = LlamafileEmbeddings()
doc_embeddings = embedder.embed_documents(
[
"Alpha is the first letter of the Greek alphabet",
"Beta is the second letter of the Greek alphabet",
]
)
query_embedding = embedder.embed_query(
"What is the second letter of the Greek alphabet"
)
"""
base_url: str = "http://localhost:8080"
"""Base url where the llamafile server is listening."""
request_timeout: Optional[int] = None
"""Timeout for server requests"""
def _embed(self, text: str) -> List[float]:
try:
response = requests.post(
url=f"{self.base_url}/embedding",
headers={
"Content-Type": "application/json",
},
json={
"content": text,
},
timeout=self.request_timeout,
)
except requests.exceptions.ConnectionError:
raise requests.exceptions.ConnectionError(
f"Could not connect to Llamafile server. Please make sure "
f"that a server is running at {self.base_url}."
)
# Raise exception if we got a bad (non-200) response status code
response.raise_for_status()
contents = response.json()
if "embedding" not in contents:
raise KeyError(
"Unexpected output from /embedding endpoint, output dict "
"missing 'embedding' key."
)
embedding = contents["embedding"]
# Sanity check the embedding vector:
# Prior to llamafile v0.6.2, if the server was not started with the
# `--embedding` option, the embedding endpoint would always return a
# 0-vector. See issue:
# https://github.com/Mozilla-Ocho/llamafile/issues/243
# So here we raise an exception if the vector sums to exactly 0.
if sum(embedding) == 0.0:
raise ValueError(
"Embedding sums to 0, did you start the llamafile server with "
"the `--embedding` option enabled?"
)
return embedding
def embed_documents(self, texts: List[str]) -> List[List[float]]:
"""Embed documents using a llamafile server running at `self.base_url`.
llamafile server should be started in a separate process before invoking
this method.
Args:
texts: The list of texts to embed.
Returns:
List of embeddings, one for each text.
"""
doc_embeddings = []
for text in texts:
doc_embeddings.append(self._embed(text))
return doc_embeddings
def embed_query(self, text: str) -> List[float]:
"""Embed a query using a llamafile server running at `self.base_url`.
llamafile server should be started in a separate process before invoking
this method.
Args:
text: The text to embed.
Returns:
Embeddings for the text.
"""
return self._embed(text)
|