Harrison/serialize from llm and tools (#760)

langchain/agents/conversational/base.py

@@ -91,14 +91,22 @@ class ConversationalAgent(Agent):
         llm: BaseLLM,
         tools: List[Tool],
         callback_manager: Optional[BaseCallbackManager] = None,
+        prefix: str = PREFIX,
+        suffix: str = SUFFIX,
         ai_prefix: str = "AI",
         human_prefix: str = "Human",
+        input_variables: Optional[List[str]] = None,
         **kwargs: Any,
     ) -> Agent:
         """Construct an agent from an LLM and tools."""
         cls._validate_tools(tools)
         prompt = cls.create_prompt(
-            tools, ai_prefix=ai_prefix, human_prefix=human_prefix
+            tools,
+            ai_prefix=ai_prefix,
+            human_prefix=human_prefix,
+            prefix=prefix,
+            suffix=suffix,
+            input_variables=input_variables,
         )
         llm_chain = LLMChain(
             llm=llm,

langchain/agents/initialize.py

@@ -54,7 +54,9 @@ def initialize_agent(
             llm, tools, callback_manager=callback_manager
         )
     elif agent_path is not None:
-        agent_obj = load_agent(agent_path, callback_manager=callback_manager)
+        agent_obj = load_agent(
+            agent_path, llm=llm, tools=tools, callback_manager=callback_manager
+        )
     else:
         raise ValueError(
             "Somehow both `agent` and `agent_path` are None, "

langchain/agents/loading.py

@@ -3,7 +3,7 @@ import json
 import os
 import tempfile
 from pathlib import Path
-from typing import Any, Union
+from typing import Any, List, Optional, Union
 
 import requests
 import yaml
@@ -13,7 +13,9 @@ from langchain.agents.conversational.base import ConversationalAgent
 from langchain.agents.mrkl.base import ZeroShotAgent
 from langchain.agents.react.base import ReActDocstoreAgent
 from langchain.agents.self_ask_with_search.base import SelfAskWithSearchAgent
+from langchain.agents.tools import Tool
 from langchain.chains.loading import load_chain, load_chain_from_config
+from langchain.llms.base import BaseLLM
 
 AGENT_TO_CLASS = {
     "zero-shot-react-description": ZeroShotAgent,
@@ -25,10 +27,42 @@ AGENT_TO_CLASS = {
 URL_BASE = "https://raw.githubusercontent.com/hwchase17/langchain-hub/master/agents/"
 
 
-def load_agent_from_config(config: dict, **kwargs: Any) -> Agent:
+def _load_agent_from_tools(
+    config: dict, llm: BaseLLM, tools: List[Tool], **kwargs: Any
+) -> Agent:
+    config_type = config.pop("_type")
+    if config_type not in AGENT_TO_CLASS:
+        raise ValueError(f"Loading {config_type} agent not supported")
+
+    if config_type not in AGENT_TO_CLASS:
+        raise ValueError(f"Loading {config_type} agent not supported")
+    agent_cls = AGENT_TO_CLASS[config_type]
+    combined_config = {**config, **kwargs}
+    return agent_cls.from_llm_and_tools(llm, tools, **combined_config)
+
+
+def load_agent_from_config(
+    config: dict,
+    llm: Optional[BaseLLM] = None,
+    tools: Optional[List[Tool]] = None,
+    **kwargs: Any,
+) -> Agent:
     """Load agent from Config Dict."""
     if "_type" not in config:
         raise ValueError("Must specify an agent Type in config")
+    load_from_tools = config.pop("load_from_llm_and_tools", False)
+    if load_from_tools:
+        if llm is None:
+            raise ValueError(
+                "If `load_from_llm_and_tools` is set to True, "
+                "then LLM must be provided"
+            )
+        if tools is None:
+            raise ValueError(
+                "If `load_from_llm_and_tools` is set to True, "
+                "then tools must be provided"
+            )
+        return _load_agent_from_tools(config, llm, tools, **kwargs)
     config_type = config.pop("_type")
 
     if config_type not in AGENT_TO_CLASS:

langchain/agents/mrkl/base.py

@@ -7,6 +7,8 @@ from typing import Any, Callable, List, NamedTuple, Optional, Tuple
 from langchain.agents.agent import Agent, AgentExecutor
 from langchain.agents.mrkl.prompt import FORMAT_INSTRUCTIONS, PREFIX, SUFFIX
 from langchain.agents.tools import Tool
+from langchain.callbacks.base import BaseCallbackManager
+from langchain.chains import LLMChain
 from langchain.llms.base import BaseLLM
 from langchain.prompts import PromptTemplate
 
@@ -92,6 +94,30 @@ class ZeroShotAgent(Agent):
             input_variables = ["input", "agent_scratchpad"]
         return PromptTemplate(template=template, input_variables=input_variables)
 
+    @classmethod
+    def from_llm_and_tools(
+        cls,
+        llm: BaseLLM,
+        tools: List[Tool],
+        callback_manager: Optional[BaseCallbackManager] = None,
+        prefix: str = PREFIX,
+        suffix: str = SUFFIX,
+        input_variables: Optional[List[str]] = None,
+        **kwargs: Any,
+    ) -> Agent:
+        """Construct an agent from an LLM and tools."""
+        cls._validate_tools(tools)
+        prompt = cls.create_prompt(
+            tools, prefix=prefix, suffix=suffix, input_variables=input_variables
+        )
+        llm_chain = LLMChain(
+            llm=llm,
+            prompt=prompt,
+            callback_manager=callback_manager,
+        )
+        tool_names = [tool.name for tool in tools]
+        return cls(llm_chain=llm_chain, allowed_tools=tool_names, **kwargs)
+
     @classmethod
     def _validate_tools(cls, tools: List[Tool]) -> None:
         for tool in tools: