Update the version to 0.1.25 (#367)

Signed-off-by: SimFG <[email protected]>

.github/workflows/Nightly_CI_main.yaml

@@ -21,9 +21,17 @@ on:
 
 jobs:
   nightly-CI-gpt-cache-cli-main:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-20.04
     strategy:
       fail-fast: false
+    services:
+      postgres:
+        image: ankane/pgvector
+        ports:
+          - 5432:5432
+        env:
+          POSTGRES_PASSWORD: postgres
+        options: --health-cmd pg_isready --health-interval 10s --health-timeout 5s --health-retries 5
     steps:
       - uses: actions/checkout@main
 
@@ -49,6 +57,12 @@ jobs:
         run: |
           pip install -r requirements.txt
 
+      - name: Download the `en_core_web_sm` model
+        shell: bash
+        working-directory: tests
+        run: |
+          python3 -m spacy download en_core_web_sm
+
       - name: Nightly CI Tests
         timeout-minutes: 10
         shell: bash

.github/workflows/pylint.yaml

@@ -28,7 +28,7 @@ on:
 
 jobs:
   pylint:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-20.04
     steps:
     - uses: actions/[email protected]
     - name: Setup Python

.github/workflows/unit_test_main.yaml

@@ -25,7 +25,7 @@ on:
 
 jobs:
   unit-test-main:
-    runs-on: ubuntu-latest
+    runs-on: ubuntu-20.04
     strategy:
       fail-fast: false
     services:

docs/release_note.md

@@ -5,6 +5,26 @@ To read the following content, you need to understand the basic use of GPTCache,
 - [Readme doc](https://github.com/zilliztech/GPTCache)
 - [Usage doc](https://github.com/zilliztech/GPTCache/blob/main/docs/usage.md)
 
+## v0.1.25 (2023.5.18)
+
+1. Support the DocArray vector database
+
+```python
+from gptcache.manager import manager_factory
+
+data_manager = manager_factory("sqlite,docarray")
+```
+
+2. Add rwkv model for embedding
+
+```python
+from gptcache.embedding import Rwkv
+
+test_sentence = 'Hello, world.'
+encoder = Rwkv(model='sgugger/rwkv-430M-pile')
+embed = encoder.to_embeddings(test_sentence)
+```
+
 ## v0.1.24 (2023.5.15)
 
 1. Support the langchain embedding

gptcache/embedding/__init__.py

@@ -1,4 +1,16 @@
-__all__ = ["OpenAI", "Huggingface", "SBERT", "Cohere", "Onnx", "FastText", "Data2VecAudio", "Timm", "ViT", "LangChain"]
+__all__ = [
+    "OpenAI",
+    "Huggingface",
+    "SBERT",
+    "Cohere",
+    "Onnx",
+    "FastText",
+    "Data2VecAudio",
+    "Timm",
+    "ViT",
+    "LangChain",
+    "Rwkv",
+]
 
 
 from gptcache.utils.lazy_import import LazyImport
@@ -52,10 +64,9 @@ def ViT(model="google/vit-base-patch16-384"):
     return vit.ViT(model)
 
 
-def LangChain(embeddings, dimension = 0):
+def LangChain(embeddings, dimension=0):
     return langchain.LangChain(embeddings, dimension)
 
 
 def Rwkv(model="sgugger/rwkv-430M-pile"):
     return rwkv.Rwkv(model)
-

gptcache/embedding/rwkv.py

@@ -1,12 +1,13 @@
 import numpy as np
 
-from gptcache.utils import import_huggingface
 from gptcache.embedding.base import BaseEmbedding
+from gptcache.utils import import_huggingface
 
 import_huggingface()
 
 from transformers import AutoTokenizer, RwkvModel  # pylint: disable=C0413
 
+
 class Rwkv(BaseEmbedding):
     """Generate sentence embedding for given text using RWKV models.
 
@@ -23,6 +24,7 @@ class Rwkv(BaseEmbedding):
             encoder = Rwkv(model='sgugger/rwkv-430M-pile')
             embed = encoder.to_embeddings(test_sentence)
     """
+
     def __init__(self, model: str = "sgugger/rwkv-430M-pile"):
         self.model = RwkvModel.from_pretrained(model)
         self.model.eval()
@@ -46,7 +48,7 @@ def to_embeddings(self, data, **_):
         """
         inputs = self.tokenizer(data, return_tensors="pt")
         outputs = self.model(inputs["input_ids"])
-        emb = outputs.last_hidden_state[0,0,:].detach().numpy()
+        emb = outputs.last_hidden_state[0, 0, :].detach().numpy()
         return np.array(emb).astype("float32")
 
     @property

gptcache/manager/factory.py

@@ -1,8 +1,9 @@
 import os
-from typing import Union, Callable
 from pathlib import Path
-from gptcache.manager.data_manager import SSDataManager, MapDataManager
+from typing import Union, Callable
+
 from gptcache.manager import CacheBase, VectorBase, ObjectBase
+from gptcache.manager.data_manager import SSDataManager, MapDataManager
 
 
 def manager_factory(manager="map",
@@ -74,10 +75,9 @@ def manager_factory(manager="map",
 
     if vector_params is None:
         vector_params = {}
-    if vector == "faiss":
-        vector_params["index_path"] = os.path.join(data_dir, "faiss.index")
-    elif vector == "hnswlib":
-        vector_params["index_path"] = os.path.join(data_dir, "hnswlib.index")
+    local_vector_type = ["faiss", "hnswlib", "docarray"]
+    if vector in local_vector_type:
+        vector_params["index_path"] = os.path.join(data_dir, f"{vector}.index")
     elif vector == "milvus" and vector_params.get("local_mode", False) is True:
         vector_params["local_data"] = os.path.join(data_dir, "milvus_data")
     v = VectorBase(name=vector, **vector_params)

gptcache/processor/context/summarization_context.py

@@ -21,7 +21,7 @@ class SummarizationContextProcess(ContextProcess):
     :param target_length: The length of the summarized text.
     :type target_length: int
     """
-    def __init__(self, summarizer = transformers.pipeline("summarization", model="facebook/bart-large-cnn"),
+    def __init__(self, summarizer=transformers.pipeline("summarization", model="facebook/bart-large-cnn"),
                   tokenizer=None, target_length=512):
         self.summarizer = summarizer
         self.target_length = target_length
@@ -30,7 +30,7 @@ def __init__(self, summarizer = transformers.pipeline("summarization", model="fa
         self.tokenizer = tokenizer
         self.content = ""
 
-    def summarize_to_sentence(self, summarizer, sentences, target_size = 1000):
+    def summarize_to_sentence(self, summarizer, sentences, target_size=1000):
         lengths = []
         for sentence in sentences:
             lengths.append(len(sentence))

setup.py

@@ -18,7 +18,7 @@ def parse_requirements(file_name: str) -> List[str]:
 setuptools.setup(
     name="gptcache",
     packages=find_packages(),
-    version="0.1.24",
+    version="0.1.25",
     author="SimFG",
     author_email="[email protected]",
     description="GPTCache, a powerful caching library that can be used to speed up and lower the cost of chat "

tests/unit_tests/embedding/test_rwkv.py

@@ -1,6 +1,8 @@
 import pytest
-from gptcache.embedding import Rwkv
+
 from gptcache.adapter.api import _get_model
+from gptcache.embedding import Rwkv
+
 
 @pytest.mark.tags("L2")
 def test_rwkv():

tests/unit_tests/manager/test_chromadb.py

@@ -1,10 +1,13 @@
 import unittest
+
 import numpy as np
+import pytest
 
 from gptcache.manager import VectorBase
 from gptcache.manager.vector_data.base import VectorData
 
 
+@pytest.mark.tags("L2")
 class TestChromadb(unittest.TestCase):
     def test_normal(self):
         db = VectorBase("chromadb", client_settings={}, top_k=3)

tests/unit_tests/manager/test_local_index.py

@@ -1,14 +1,16 @@
 import unittest
-import numpy as np
 from functools import partial
-from tempfile import TemporaryDirectory
 from pathlib import Path
+from tempfile import TemporaryDirectory
+
+import numpy as np
+import pytest
 
+from gptcache.manager.vector_data import VectorBase
+from gptcache.manager.vector_data.base import VectorData
 from gptcache.manager.vector_data.docarray_index import DocArrayIndex
 from gptcache.manager.vector_data.faiss import Faiss
 from gptcache.manager.vector_data.hnswlib_store import Hnswlib
-from gptcache.manager.vector_data import VectorBase
-from gptcache.manager.vector_data.base import VectorData
 
 DIM = 512
 MAX_ELEMENTS = 10000
@@ -30,6 +32,7 @@ def test_faiss(self):
                 name='faiss', top_k=3, dimension=DIM, index_path=index_path
             )
 
+    @pytest.mark.tags("L2")
     def test_hnswlib(self):
         cls = partial(Hnswlib, max_elements=MAX_ELEMENTS, dimension=DIM)
         self._internal_test_normal(cls)

tests/unit_tests/manager/test_pgvector.py

@@ -1,11 +1,14 @@
 import os
 import unittest
+
 import numpy as np
+import pytest
 
 from gptcache.manager.vector_data import VectorBase
 from gptcache.manager.vector_data.base import VectorData
 
 
+@pytest.mark.tags("L2")
 class TestPgvector(unittest.TestCase):
     def test_normal(self):
         size = 1000