community/templates: ClickHouse Template

libs/community/langchain_community/vectorstores/clickhouse.py

@@ -72,7 +72,7 @@
     username: Optional[str] = None
     password: Optional[str] = None
 
-    index_type: str = "annoy"
+    index_type: Optional[str] = "annoy"
     # Annoy supports L2Distance and cosineDistance.
     index_param: Optional[Union[List, Dict]] = ["'L2Distance'", 100]
     index_query_params: Dict[str, str] = {}
@@ -177,18 +177,8 @@
             else self.config.index_param
         )
 
-        self.schema = f"""\
-CREATE TABLE IF NOT EXISTS {self.config.database}.{self.config.table}(
-    {self.config.column_map['id']} Nullable(String),
-    {self.config.column_map['document']} Nullable(String),
-    {self.config.column_map['embedding']} Array(Float32),
-    {self.config.column_map['metadata']} JSON,
-    {self.config.column_map['uuid']} UUID DEFAULT generateUUIDv4(),
-    CONSTRAINT cons_vec_len CHECK length({self.config.column_map['embedding']}) = {dim},
-    INDEX vec_idx {self.config.column_map['embedding']} TYPE \
-{self.config.index_type}({index_params}) GRANULARITY 1000
-) ENGINE = MergeTree ORDER BY uuid SETTINGS index_granularity = 8192\
-"""
+        self.schema = self._schema(dim, index_params)
+
         self.dim = dim
         self.BS = "\\"
         self.must_escape = ("\\", "'")
@@ -209,6 +199,32 @@
         self.client.command(f"SET allow_experimental_{self.config.index_type}_index=1")
         self.client.command(self.schema)
 
+    def _schema(self, dim, index_params):
+        if self.config.index_type:
+            return f"""\
+        CREATE TABLE IF NOT EXISTS {self.config.database}.{self.config.table}(
+            {self.config.column_map['id']} Nullable(String),
+            {self.config.column_map['document']} Nullable(String),
+            {self.config.column_map['embedding']} Array(Float32),
+            {self.config.column_map['metadata']} JSON,
+            {self.config.column_map['uuid']} UUID DEFAULT generateUUIDv4(),
+            CONSTRAINT cons_vec_len CHECK length({self.config.column_map['embedding']}) = {dim},
+            INDEX vec_idx {self.config.column_map['embedding']} TYPE \
+        {self.config.index_type}({index_params}) GRANULARITY 1000
+        ) ENGINE = MergeTree ORDER BY uuid SETTINGS index_granularity = 8192\
+        """
+        else:
+            return f"""\
+                CREATE TABLE IF NOT EXISTS {self.config.database}.{self.config.table}(
+                    {self.config.column_map['id']} Nullable(String),
+                    {self.config.column_map['document']} Nullable(String),
+                    {self.config.column_map['embedding']} Array(Float32),
+                    {self.config.column_map['metadata']} JSON,
+                    {self.config.column_map['uuid']} UUID DEFAULT generateUUIDv4(),
+                    CONSTRAINT cons_vec_len CHECK length({self.config.column_map['embedding']}) = {dim}
+                ) ENGINE = MergeTree ORDER BY uuid
+                """
+
     @property
     def embeddings(self) -> Embeddings:
         return self.embedding_function

templates/rag-clickhouse/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2023 LangChain, Inc.
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

templates/rag-clickhouse/README.md

@@ -0,0 +1,89 @@
+
+# rag-clickhouse
+
+This template performs RAG with no reliance on external APIs. 
+
+It utilizes Ollama the LLM, GPT4All for embeddings, and ClickHouse for the vectorstore.
+
+The vectorstore is created in `chain.py` and by default indexes a [blog post about feature stores]([https://lilianweng.github.io/posts/2023-06-23-agent/](https://clickhouse.com/blog/powering-featurestores-with-clickhouse)) for question-answering. 
+
+## Environment Setup
+
+To set up the environment, you need to download Ollama. 
+
+Follow the instructions [here](https://python.langchain.com/docs/integrations/chat/ollama). 
+
+You can choose the desired LLM with Ollama. 
+
+This template uses `mistral`, which can be accessed using `ollama pull mistral`.
+
+There are also [other models available](https://ollama.ai/library).
+
+This package also uses [GPT4All](https://python.langchain.com/docs/integrations/text_embedding/gpt4all) embeddings.
+
+You'll also need to install ClickHouse:
+
+```bash
+curl https://clickhouse.com/ | sh
+./clickhouse server
+```
+
+## Usage
+
+To use this package, you should first have the LangChain CLI installed:
+
+```shell
+pip install -U langchain-cli
+```
+
+To create a new LangChain project (called `my-app`) and install this as the only package, you can do:
+
+```shell
+langchain app new my-app --package rag-clickhouse
+```
+
+If you want to add this to an existing project, you can run:
+
+```shell
+langchain app add rag-clickhouse
+```
+
+And add the following code to your `server.py` file:
+```python
+from rag_clickhouse import chain as rag_clickhouse
+
+add_routes(app, rag_clickhouse, path="/rag-clickhouse")
+```
+
+(Optional) Let's now configure LangSmith. LangSmith will help us trace, monitor and debug LangChain applications. 
+LangSmith is currently in private beta, you can sign up [here](https://smith.langchain.com/). 
+If you don't have access, you can skip this section
+
+```shell
+export LANGCHAIN_TRACING_V2=true
+export LANGCHAIN_API_KEY=<your-api-key>
+export LANGCHAIN_PROJECT=<your-project>  # if not specified, defaults to "default"
+```
+
+If you are inside this directory, then you can spin up a LangServe instance directly by:
+
+```shell
+langchain serve
+```
+
+This will start the FastAPI app with a server is running locally at 
+[http://localhost:8000](http://localhost:8000)
+
+We can see all templates at [http://127.0.0.1:8000/docs](http://127.0.0.1:8000/docs)
+We can access the playground at [http://127.0.0.1:8000/rag-clickhouse/playground](http://127.0.0.1:8000/rag_clickhouse/playground)  
+
+We can access the template from code with:
+
+```python
+from langserve.client import RemoteRunnable
+
+runnable = RemoteRunnable("http://localhost:8000/rag-clickhouse")
+```
+
+The package will create and add documents to the vector database in `chain.py`. 
+By default, it will load a popular blog post on agents. However, you can choose from a large number of document loaders [here](https://python.langchain.com/docs/integrations/document_loaders).

templates/rag-clickhouse/pyproject.toml

@@ -0,0 +1,35 @@
+[tool.poetry]
+name = "rag-clickhouse"
+version = "0.1.0"
+description = "RAG using local LLM, embeddings, vectorstore"
+authors = [
+    "Mark Needham <[email protected]>",
+]
+readme = "README.md"
+
+[tool.poetry.dependencies]
+python = ">=3.8.1,<4.0"
+langchain = "^0.1"
+tiktoken = ">=0.5.1"
+clickhouse-connect = ">=0.7.0"
+gpt4all = ">=1.0.8"
+beautifulsoup4 = ">=4.12.2"
+
+[tool.poetry.group.dev.dependencies]
+langchain-cli = ">=0.0.21"
+
+[tool.langserve]
+export_module = "rag_clickhouse"
+export_attr = "chain"
+
+[tool.templates-hub]
+use-case = "rag"
+author = "LangChain"
+integrations = ["ClickHouse", "Gpt4all", "Ollama"]
+tags = ["vectordbs"]
+
+[build-system]
+requires = [
+    "poetry-core",
+]
+build-backend = "poetry.core.masonry.api"

templates/rag-clickhouse/rag_clickhouse.ipynb

@@ -0,0 +1,100 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "232fd40d-cf6a-402d-bcb8-414184a8e924",
+   "metadata": {},
+   "source": [
+    "## ClickHouse Template\n",
+    "\n",
+    "In `server.py`, set the following:\n",
+    "\n",
+    "```python\n",
+    "from rag_clickhouse import chain as rag_clickhouse_chain\n",
+    "add_routes(app, rag_clickhouse_chain, path=\"/rag-clickhouse\")\n",
+    "```"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "5494f6a2-f809-4210-a49c-b3766ca2d83f",
+   "metadata": {
+    "execution": {
+     "iopub.execute_input": "2024-02-08T16:43:34.381207Z",
+     "iopub.status.busy": "2024-02-08T16:43:34.380716Z",
+     "iopub.status.idle": "2024-02-08T16:43:34.387371Z",
+     "shell.execute_reply": "2024-02-08T16:43:34.386243Z",
+     "shell.execute_reply.started": "2024-02-08T16:43:34.381176Z"
+    }
+   },
+   "source": [
+    "And then run the following to ask a question:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "id": "ce39d358-1934-4404-bd3e-3fd497974aff",
+   "metadata": {
+    "ExecuteTime": {
+     "end_time": "2024-02-07T12:43:16.792520Z",
+     "start_time": "2024-02-07T12:43:16.717875Z"
+    },
+    "execution": {
+     "iopub.execute_input": "2024-02-08T16:42:37.446022Z",
+     "iopub.status.busy": "2024-02-08T16:42:37.445680Z",
+     "iopub.status.idle": "2024-02-08T16:42:39.330910Z",
+     "shell.execute_reply": "2024-02-08T16:42:39.330307Z",
+     "shell.execute_reply.started": "2024-02-08T16:42:37.445996Z"
+    }
+   },
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "' Yes, ClickHouse can be used as a feature store. The documents suggest that ClickHouse can fulfill the role of several components in a feature store and potentially simplify the architecture due to its real-time data warehouse capabilities and performance. However, it is important to note that not all feature stores provide the same components directly, so some degree of architectural flexibility and openness may be required for ClickHouse to be integrated into a specific feature store implementation.'"
+      ]
+     },
+     "execution_count": 5,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "from langserve.client import RemoteRunnable\n",
+    "\n",
+    "rag_app = RemoteRunnable(\"http://127.0.0.1:8000/rag-clickhouse/\")\n",
+    "rag_app.invoke(\"Can you use ClickHouse as a feature store?\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "fa2254e7-a7ae-4b0f-bf6c-2b830b837a27",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.11.6"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

templates/rag-clickhouse/rag_clickhouse/__init__.py

@@ -0,0 +1,3 @@
+from rag_clickhouse.chain import chain
+
+__all__ = ["chain"]

templates/rag-clickhouse/rag_clickhouse/chain.py

@@ -0,0 +1,60 @@
+# Load
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain_community.chat_models import ChatOllama
+from langchain_community.document_loaders import WebBaseLoader
+from langchain_community.embeddings import GPT4AllEmbeddings
+from langchain_community.vectorstores import Clickhouse, ClickhouseSettings
+from langchain_core.output_parsers import StrOutputParser
+from langchain_core.prompts import ChatPromptTemplate
+from langchain_core.pydantic_v1 import BaseModel
+from langchain_core.runnables import RunnableParallel, RunnablePassthrough
+
+loader = WebBaseLoader(
+    "https://clickhouse.com/blog/powering-featurestores-with-clickhouse"
+)
+data = loader.load()
+
+# Split
+
+text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=0)
+all_splits = text_splitter.split_documents(data)
+
+# Add to vectorDB
+settings = ClickhouseSettings(table="clickhouse_vector_search_example", index_type=None)
+vectorstore = Clickhouse.from_documents(
+    documents=all_splits, embedding=GPT4AllEmbeddings(), config=settings
+)
+retriever = vectorstore.as_retriever()
+
+# Prompt
+# Optionally, pull from the Hub
+# from langchain import hub
+# prompt = hub.pull("rlm/rag-prompt")
+# Or, define your own:
+template = """Answer the question based only on the following context:
+{context}
+
+Question: {question}
+"""
+prompt = ChatPromptTemplate.from_template(template)
+
+# LLM
+# Select the LLM that you downloaded
+ollama_llm = "mistral"
+model = ChatOllama(model=ollama_llm)
+
+# RAG chain
+chain = (
+    RunnableParallel({"context": retriever, "question": RunnablePassthrough()})
+    | prompt
+    | model
+    | StrOutputParser()
+)
+
+
+# Add typing for input
+class Question(BaseModel):
+    __root__: str
+
+
+chain = chain.with_types(input_type=Question)