Tool Retrieval Template (#13104)

Adds a template like
https://python.langchain.com/docs/modules/agents/how_to/custom_agent_with_tool_retrieval

Uses OpenAI functions, LCEL, and FAISS

templates/openai-functions-agent/openai_functions_agent/agent.py

@@ -62,11 +62,11 @@ def _format_chat_history(chat_history: List[Tuple[str, str]]):
 
 class AgentInput(BaseModel):
     input: str
-    chat_history: List[Tuple[str, str]] = Field(..., extra={"widget": {"type": "chat"}})
+    chat_history: List[Tuple[str, str]] = Field(
+        ..., extra={"widget": {"type": "chat", "input": "input", "output": "output"}}
+    )
 
 
 agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True).with_types(
     input_type=AgentInput
 )
-
-agent_executor = agent_executor | (lambda x: x["output"])

templates/openai-functions-tool-retrieval-agent/.gitignore

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

templates/openai-functions-tool-retrieval-agent/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/openai-functions-tool-retrieval-agent/README.md

@@ -0,0 +1,74 @@
+# openai-functions-tool-retrieval-agent
+
+The novel idea introduced in this template is the idea of using retrieval to select the set of tools to use to answer an agent query. This is useful when you have many many tools to select from. You cannot put the description of all the tools in the prompt (because of context length issues) so instead you dynamically select the N tools you do want to consider using at run time.
+
+In this template we will create a somewhat contrived example. We will have one legitimate tool (search) and then 99 fake tools which are just nonsense. We will then add a step in the prompt template that takes the user input and retrieves tool relevant to the query.
+
+This template is based on [this Agent How-To](https://python.langchain.com/docs/modules/agents/how_to/custom_agent_with_tool_retrieval).
+
+## Environment Setup
+
+The following environment variables need to be set:
+
+Set the `OPENAI_API_KEY` environment variable to access the OpenAI models.
+
+Set the `TAVILY_API_KEY` environment variable to access Tavily.
+
+## 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 and install this as the only package, you can do:
+
+```shell
+langchain app new my-app --package openai-functions-tool-retrieval-agent
+```
+
+If you want to add this to an existing project, you can just run:
+
+```shell
+langchain app add openai-functions-tool-retrieval-agent
+```
+
+And add the following code to your `server.py` file:
+```python
+from openai_functions_tool_retrieval_agent import chain as openai_functions_tool_retrieval_agent_chain
+
+add_routes(app, openai_functions_tool_retrieval_agent_chain, path="/openai-functions-tool-retrieval-agent")
+```
+
+(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/openai-functions-tool-retrieval-agent/playground](http://127.0.0.1:8000/openai-functions-tool-retrieval-agent/playground)  
+
+We can access the template from code with:
+
+```python
+from langserve.client import RemoteRunnable
+
+runnable = RemoteRunnable("http://localhost:8000/openai-functions-tool-retrieval-agent")
+```

templates/openai-functions-tool-retrieval-agent/openai_functions_tool_retrieval_agent/__init__.py

@@ -0,0 +1,3 @@
+from openai_functions_tool_retrieval_agent.agent import agent_executor
+
+__all__ = ["agent_executor"]

templates/openai-functions-tool-retrieval-agent/openai_functions_tool_retrieval_agent/agent.py

@@ -0,0 +1,123 @@
+from typing import Dict, List, Tuple
+
+from langchain.agents import (
+    AgentExecutor,
+    Tool,
+)
+from langchain.agents.format_scratchpad import format_to_openai_functions
+from langchain.agents.output_parsers import OpenAIFunctionsAgentOutputParser
+from langchain.chat_models import ChatOpenAI
+from langchain.embeddings import OpenAIEmbeddings
+from langchain.prompts import (
+    ChatPromptTemplate,
+    MessagesPlaceholder,
+)
+from langchain.pydantic_v1 import BaseModel, Field
+from langchain.schema import Document
+from langchain.schema.messages import AIMessage, HumanMessage
+from langchain.schema.runnable import Runnable, RunnableLambda, RunnableParallel
+from langchain.tools.base import BaseTool
+from langchain.tools.render import format_tool_to_openai_function
+from langchain.tools.tavily_search import TavilySearchResults
+from langchain.utilities.tavily_search import TavilySearchAPIWrapper
+from langchain.vectorstores import FAISS
+
+# Create the tools
+search = TavilySearchAPIWrapper()
+description = """"Useful for when you need to answer questions \
+about current events or about recent information."""
+tavily_tool = TavilySearchResults(api_wrapper=search, description=description)
+
+
+def fake_func(inp: str) -> str:
+    return "foo"
+
+
+fake_tools = [
+    Tool(
+        name=f"foo-{i}",
+        func=fake_func,
+        description=("a silly function that gets info " f"about the number {i}"),
+    )
+    for i in range(99)
+]
+ALL_TOOLS: List[BaseTool] = [tavily_tool] + fake_tools
+
+# turn tools into documents for indexing
+docs = [
+    Document(page_content=t.description, metadata={"index": i})
+    for i, t in enumerate(ALL_TOOLS)
+]
+
+vector_store = FAISS.from_documents(docs, OpenAIEmbeddings())
+
+retriever = vector_store.as_retriever()
+
+
+def get_tools(query: str) -> List[Tool]:
+    docs = retriever.get_relevant_documents(query)
+    return [ALL_TOOLS[d.metadata["index"]] for d in docs]
+
+
+assistant_system_message = """You are a helpful assistant. \
+Use tools (only if necessary) to best answer the users questions."""
+assistant_system_message = """You are a helpful assistant. \
+Use tools (only if necessary) to best answer the users questions."""
+prompt = ChatPromptTemplate.from_messages(
+    [
+        ("system", assistant_system_message),
+        MessagesPlaceholder(variable_name="chat_history"),
+        ("user", "{input}"),
+        MessagesPlaceholder(variable_name="agent_scratchpad"),
+    ]
+)
+
+
+def llm_with_tools(input: Dict) -> Runnable:
+    return RunnableLambda(lambda x: x["input"]) | ChatOpenAI(temperature=0).bind(
+        functions=input["functions"]
+    )
+
+
+def _format_chat_history(chat_history: List[Tuple[str, str]]):
+    buffer = []
+    for human, ai in chat_history:
+        buffer.append(HumanMessage(content=human))
+        buffer.append(AIMessage(content=ai))
+    return buffer
+
+
+agent = (
+    RunnableParallel(
+        {
+            "input": lambda x: x["input"],
+            "chat_history": lambda x: _format_chat_history(x["chat_history"]),
+            "agent_scratchpad": lambda x: format_to_openai_functions(
+                x["intermediate_steps"]
+            ),
+            "functions": lambda x: [
+                format_tool_to_openai_function(tool) for tool in get_tools(x["input"])
+            ],
+        }
+    )
+    | {
+        "input": prompt,
+        "functions": lambda x: x["functions"],
+    }
+    | llm_with_tools
+    | OpenAIFunctionsAgentOutputParser()
+)
+
+# LLM chain consisting of the LLM and a prompt
+
+
+class AgentInput(BaseModel):
+    input: str
+    chat_history: List[Tuple[str, str]] = Field(
+        ..., extra={"widget": {"type": "chat", "input": "input", "output": "output"}}
+    )
+
+
+agent_executor = AgentExecutor(agent=agent, tools=ALL_TOOLS).with_types(
+    input_type=AgentInput
+)