Merge pull request #738 from cheshire-cat-ai/develop

1.5.1 fastembed compatibility and model cache

.gitignore

@@ -24,6 +24,10 @@ test_db.db
 # old admin folder
 admin
 
+# data folder (settings and models)
+core/cat/data/*
+!core/cat/data/.gitkeep
+
 # plugins are not tracked by git
 core/cat/plugins/*
 # plugins readme can stay

core/cat/factory/embedder.py

@@ -10,7 +10,7 @@
 )
 from langchain_openai import OpenAIEmbeddings, AzureOpenAIEmbeddings
 from langchain_google_genai import GoogleGenerativeAIEmbeddings
-from fastembed.embedding import Embedding
+from fastembed.embedding import TextEmbedding
 from cat.factory.custom_embedder import DumbEmbedder, CustomOpenAIEmbeddings
 from cat.mad_hatter.mad_hatter import MadHatter
 
@@ -125,7 +125,7 @@ class EmbedderCohereConfig(EmbedderSettings):
     "FastEmbedModels",
     {
         item["model"].replace("/", "_").replace("-", "_"): item["model"]
-        for item in Embedding.list_supported_models()
+        for item in TextEmbedding.list_supported_models()
     },
 )
 
@@ -134,6 +134,7 @@ class EmbedderQdrantFastEmbedConfig(EmbedderSettings):
     model_name: FastEmbedModels = Field(title="Model name", default="BAAI/bge-base-en")
     max_length: int = 512  # Unknown behavior for values > 512.
     doc_embed_type: str = "passage"  # as suggest on fastembed documentation, "passage" is the best option for documents.
+    cache_dir: str = "cat/data/models/fast_embed"
 
     _pyclass: Type = FastEmbedEmbeddings
 

core/pyproject.toml

@@ -1,7 +1,7 @@
 [project]
 name = "Cheshire-Cat"
 description = "Production ready AI assistant framework"
-version = "1.5.0"
+version = "1.5.1"
 requires-python = ">=3.10"
 license = { file="LICENSE" }
 authors = [
@@ -16,18 +16,18 @@ dependencies = [
     "scikit-learn==1.2.1",
     "qdrant_client==1.7.2",
     "langchain==0.1.10",
-    "langchain-community==0.0.25",
+    "langchain-community==0.0.27",
     "langchain-openai",
     "langchain-google-genai",
     "openai==1.10.0",
     "cohere==4.39",
-    "huggingface-hub==0.19.4",
+    "huggingface-hub==0.20.3",
     "beautifulsoup4==4.12.0",
     "pdfminer.six==20221105",
     "unstructured==0.5.7",
     "tiktoken==0.5.2",
     "tomli",
-    "loguru==0.7.0",
+    "loguru==0.7.2",
     "anthropic==0.2.9",
     "gunicorn==20.1.0",
     "uvicorn[standard]==0.20.0",
@@ -40,7 +40,7 @@ dependencies = [
     "pylint-actions",
     "pytest",
     "httpx",
-    "fastembed==0.1.3",
+    "fastembed==0.2.2",
     "rapidfuzz==3.6.1",
 ]
 

readme/CHANGELOG.md

@@ -1,5 +1,18 @@
 # Changelog
 
+## 1.5.0 ( 2024-03-07 )
+
+New in version 1.5.0: **Sir Punctilious Cat, the IV**! 
+
+- `@form` and `CatForm` to build goal oriented conversation in plugins - @Pingdred, @MaxDam, @pieroit and pizza challenge contributors
+- `@tool` `examples` to easily trigger tools - @zAlweNy26, @Pingdred
+- Agent refactoring (both `forms` ant `tools` are unified under procedural memory) - @Pingdred, @pieroit
+- Async tools - @Pingdred 
+- Update Azure LLM - @cristianorevil 
+- Update HuggingFaceEndpoints LLM - @alessioserra
+- Dependencies update
+- More unit tests, more stability
+
 ## 1.4.8 ( 2024-02-10 )
 
 New in version 1.4.8