Templates (langchain-ai#12294)

Co-authored-by: Harrison Chase <[email protected]>
Co-authored-by: Lance Martin <[email protected]>
Co-authored-by: Jacob Lee <[email protected]>

libs/cli/langchain_cli/constants.py

@@ -1,3 +1,3 @@
 DEFAULT_GIT_REPO = "https://github.com/langchain-ai/langchain.git"
 DEFAULT_GIT_SUBDIRECTORY = "templates"
-DEFAULT_GIT_REF = "langserve-templates"
+DEFAULT_GIT_REF = "master"

libs/langchain/langchain/agents/agent_toolkits/__init__.py

@@ -21,9 +21,6 @@
 from langchain.agents.agent_toolkits.conversational_retrieval.openai_functions import (
     create_conversational_retrieval_agent,
 )
-from langchain.agents.agent_toolkits.conversational_retrieval.tool import (
-    create_retriever_tool,
-)
 from langchain.agents.agent_toolkits.csv.base import create_csv_agent
 from langchain.agents.agent_toolkits.file_management.toolkit import (
     FileManagementToolkit,
@@ -59,6 +56,7 @@
 )
 from langchain.agents.agent_toolkits.xorbits.base import create_xorbits_agent
 from langchain.agents.agent_toolkits.zapier.toolkit import ZapierToolkit
+from langchain.tools.retriever import create_retriever_tool
 
 __all__ = [
     "AINetworkToolkit",
@@ -87,12 +85,12 @@
     "create_pbi_agent",
     "create_pbi_chat_agent",
     "create_python_agent",
-    "create_retriever_tool",
     "create_spark_dataframe_agent",
     "create_spark_sql_agent",
     "create_sql_agent",
     "create_vectorstore_agent",
     "create_vectorstore_router_agent",
     "create_xorbits_agent",
     "create_conversational_retrieval_agent",
+    "create_retriever_tool",
 ]

libs/langchain/langchain/agents/agent_toolkits/conversational_retrieval/tool.py

@@ -1,25 +1,3 @@
-from langchain.schema import BaseRetriever
-from langchain.tools import Tool
+from langchain.tools.retriever import create_retriever_tool
 
-
-def create_retriever_tool(
-    retriever: BaseRetriever, name: str, description: str
-) -> Tool:
-    """Create a tool to do retrieval of documents.
-
-    Args:
-        retriever: The retriever to use for the retrieval
-        name: The name for the tool. This will be passed to the language model,
-            so should be unique and somewhat descriptive.
-        description: The description for the tool. This will be passed to the language
-            model, so should be descriptive.
-
-    Returns:
-        Tool class to pass to an agent
-    """
-    return Tool(
-        name=name,
-        description=description,
-        func=retriever.get_relevant_documents,
-        coroutine=retriever.aget_relevant_documents,
-    )
+__all__ = ["create_retriever_tool"]

libs/langchain/langchain/retrievers/you.py

@@ -1,4 +1,4 @@
-from typing import Any, Dict, List
+from typing import Any, Dict, List, Optional
 
 from langchain.callbacks.manager import CallbackManagerForRetrieverRun
 from langchain.pydantic_v1 import root_validator
@@ -17,6 +17,7 @@ class YouRetriever(BaseRetriever):
     """
 
     ydc_api_key: str
+    k: Optional[int] = None
     endpoint_type: str = "web"
 
     @root_validator(pre=True)
@@ -45,6 +46,8 @@ def _get_relevant_documents(
             for hit in results["hits"]:
                 for snippet in hit["snippets"]:
                     docs.append(Document(page_content=snippet))
+                    if self.k is not None and len(docs) >= self.k:
+                        return docs
             return docs
         elif self.endpoint_type == "snippet":
             results = requests.get(

libs/langchain/langchain/tools/retriever.py

@@ -0,0 +1,25 @@
+from langchain.schema import BaseRetriever
+from langchain.tools import Tool
+
+
+def create_retriever_tool(
+    retriever: BaseRetriever, name: str, description: str
+) -> Tool:
+    """Create a tool to do retrieval of documents.
+
+    Args:
+        retriever: The retriever to use for the retrieval
+        name: The name for the tool. This will be passed to the language model,
+            so should be unique and somewhat descriptive.
+        description: The description for the tool. This will be passed to the language
+            model, so should be descriptive.
+
+    Returns:
+        Tool class to pass to an agent
+    """
+    return Tool(
+        name=name,
+        description=description,
+        func=retriever.get_relevant_documents,
+        coroutine=retriever.aget_relevant_documents,
+    )

pyproject.toml

@@ -34,7 +34,7 @@ codespell = "^2.2.0"
 
 
 [tool.codespell]
-skip = '.git,*.pdf,*.svg,*.pdf,*.yaml,*.ipynb,poetry.lock,*.min.js,*.css,package-lock.json,example_data,_dist,examples'
+skip = '.git,*.pdf,*.svg,*.pdf,*.yaml,*.ipynb,poetry.lock,*.min.js,*.css,package-lock.json,example_data,_dist,examples,templates'
 # Ignore latin etc
 ignore-regex = '.*(Stati Uniti|Tense=Pres).*'
 # whats is a typo but used frequently in queries so kept as is

templates/.gitignore

@@ -0,0 +1 @@
+__pycache__

templates/INDEX.md

@@ -0,0 +1,9 @@
+# Templates
+
+A list of all template repos
+
+⭐Retrieval Augmented Generation Chatbot: Build a chatbot over your data. Uses OpenAI and Pinecone.
+⭐Extraction with OpenAI Functions: Do extraction of structured data from unstructured data. Uses OpenAI function calling.
+⭐Local Retrieval Augmented Generation: Build a chatbot over your data. Uses only local tooling: Ollama, GPT4all, Chroma.
+⭐OpenAI Functions Agent: Build a chatbot that can take actions. Uses OpenAI function calling and Tavily.
+⭐XML Agent: Build a chatbot that can take actions. Uses Anthropic and You.com.

templates/README.md

@@ -0,0 +1,71 @@
+# LangServe Hub
+
+Packages that can be easily hosted by LangServe using the `langserve` cli.
+
+## Using LangServe Hub
+
+You can install the `langservehub` CLI and use it as follows:
+```bash
+# install langservehub CLI
+pip install --upgrade langservehub
+
+langservehub new my-app
+cd my-app
+
+poetry install
+
+# if you have problems with poe, use `poetry run poe ...` instead
+
+# add the simple-pirate package
+poe add --repo=pingpong-templates/hub simple-pirate
+
+# adding other GitHub repo packages, defaults to repo root
+poe add --repo=hwchase17/chain-of-verification
+
+# with a custom api mount point (defaults to `/{package_name}`)
+poe add --repo=pingpong-templates/hub simple-translator --api_path=/my/custom/path/translator
+
+poe list
+
+poe start
+^C
+
+# remove packages by their api path:
+poe remove my/custom/path/translator
+```
+
+## Creating New Packages
+
+You can also create new packages with the `langservehub package new` command
+
+```bash
+# starting from this directory in langserve-hub
+langservehub package new simple-newpackage
+```
+
+Now you can edit the chain in `simple-newpackage/simple_newpackage/chain.py` and put up a PR!
+
+Your package will be usable as `poe add --repo=pingpong-templates/hub simple-newpackage` when it's merged in.
+
+## Data Format
+
+What makes these packages work?
+
+- Poetry
+- pyproject.toml files
+
+### Installable Packages
+
+Everything is a Poetry package currently. This allows poetry to manage our dependencies for us :).
+
+In addition to normal keys in the `pyproject.toml` file, you'll notice an additional `tool.langserve` key ([link](https://github.com/langchain-ai/langserve-hub/blob/main/simple/pirate/pyproject.toml#L13-L15)).
+
+This allows us to identify which module and attribute to import as the chain/runnable for the langserve `add_routes` call.
+
+### Apps (with installed langserve packages)
+
+Let's say you add the pirate package with `poe add --repo=pingpong-templates/hub simple-pirate`.
+
+First this downloads the simple-pirate package to pirate
+
+Then this adds a `poetry` path dependency, which gets picked up from `add_package_routes`.

templates/anthropic-functions/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/anthropic-functions/README.md

@@ -0,0 +1,39 @@
+# Function calling with Anthropic
+
+This template enables [Anthropic function calling](https://python.langchain.com/docs/integrations/chat/anthropic_functions).
+
+Function calling can be used for various tasks, such as extraction or tagging. 
+
+Specify the function you want to use in `chain.py`
+
+By default, it will tag the input text using the following fields:
+
+* sentiment
+* aggressiveness
+* language
+
+##  LLM
+
+This template will use `Claude2` by default. 
+
+Be sure that `ANTHROPIC_API_KEY` is set in your enviorment.
+
+## Adding the template
+
+Create your LangServe app:
+```
+langchain serve new my-app
+cd my-app
+```
+
+Add template:
+```
+langchain serve add anthropic-functions
+```
+
+Start server:
+```
+langchain start
+```
+
+See Jupyter notebook `anthropic_functions` for various way to connect to the template.

templates/anthropic-functions/anthropic_functions.ipynb

@@ -0,0 +1,83 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "4ae4b789",
+   "metadata": {},
+   "source": [
+    "## Document Loading\n",
+    "\n",
+    "Load a blog post on agents."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "id": "5d6bd62e",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from langchain.document_loaders import WebBaseLoader\n",
+    "loader = WebBaseLoader(\"https://lilianweng.github.io/posts/2023-06-23-agent/\")\n",
+    "text = loader.load()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "8e21575d",
+   "metadata": {},
+   "source": [
+    "## Run Template\n",
+    "\n",
+    "As shown in the README, add template and start server:\n",
+    "```\n",
+    "langchain serve add anthropic-functions\n",
+    "langchain start\n",
+    "```\n",
+    "\n",
+    "We can now look at the endpoints:\n",
+    "\n",
+    "http://127.0.0.1:8000/docs#\n",
+    "\n",
+    "And specifically at our loaded template:\n",
+    "\n",
+    "http://127.0.0.1:8000/docs#/default/invoke_anthropic-functions_invoke_post\n",
+    " \n",
+    "We can also use remote runnable to call it:"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "id": "92edba86",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from langserve.client import RemoteRunnable\n",
+    "anthropic_function_model = RemoteRunnable('http://localhost:8000/anthropic-functions')\n",
+    "anthropic_function_model.invoke(text[0].page_content[0:1500])"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "langserve",
+   "language": "python",
+   "name": "langserve"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.9.16"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

templates/anthropic-functions/anthropic_functions/__init__.py

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

templates/anthropic-functions/anthropic_functions/chain.py

@@ -0,0 +1,15 @@
+from langchain.chains import create_tagging_chain
+from langchain_experimental.llms.anthropic_functions import AnthropicFunctions
+
+model = AnthropicFunctions(model='claude-2')
+
+schema = {
+    "properties": {
+        "sentiment": {"type": "string"},
+        "aggressiveness": {"type": "integer"},
+        "language": {"type": "string"},
+    }
+}
+
+# This is LLMChain, which implements invoke 
+chain = create_tagging_chain(schema, model)