Harrison/agents rewrite #15028
Harrison/agents rewrite #15028
Conversation
dosubot
bot
added
the
size:L
|
The latest updates on your projects. Learn more about Vercel for Git ↗︎
|
dosubot
bot
added
Ɑ: agent
|
|
||
|
|
||
| def create_openai_functions_agent( | ||
| llm: BaseLanguageModel, tools: Sequence[BaseTool], prompt: ChatPromptTemplate |
There was a problem hiding this comment.
does prompt need to be typed as ChatPromptTemplate
There was a problem hiding this comment.
guessing it's an intentional choice not to make it optional?
There was a problem hiding this comment.
[Not design related] Since they're deprecating the functions API shouldn't we be pushing a "tools" agent?
| messages: List[Union[BaseMessagePromptTemplate, BaseMessage]] | ||
| messages = [SystemMessage(content=system_message)] | ||
| if include_chat_history: | ||
| messages.append(MessagesPlaceholder(variable_name="chat_history")) | ||
| messages.extend( | ||
| [ | ||
| HumanMessagePromptTemplate.from_template("{input}"), | ||
| MessagesPlaceholder(variable_name="agent_scratchpad"), | ||
| ] | ||
| ) | ||
| return ChatPromptTemplate(messages=messages) |
There was a problem hiding this comment.
nit: find ChatPromptTemplate.from_messages cleaner
messages = [("system", system_message),]
if include_chat_history:
messages.append(MessagesPlaceholder())
messages.extend([
("human", "{input}"),
MessagesPlaceholder()
])
return ChatPromptTemplate.from_messages(messages)
| messages: List[Union[BaseMessagePromptTemplate, BaseMessage]] | ||
| messages = [SystemMessage(content=system_message)] | ||
| if include_chat_history: | ||
| messages.append(MessagesPlaceholder(variable_name="chat_history")) |
There was a problem hiding this comment.
I would really love to find a good way to standardize or reserve this name - always struck me as a bit arbitrary and inconsistent across modules
There was a problem hiding this comment.
Possibly as memory becomes more defined
| agent = ( | ||
| RunnablePassthrough.assign( | ||
| agent_scratchpad=lambda x: format_to_openai_function_messages( | ||
| x["intermediate_steps"] |
There was a problem hiding this comment.
for trying to run agent manually (while experimenting) would be nice if "intermediate_steps" wasn't required (ie do x.get('intermediate_steps', []))
| functions=[format_tool_to_openai_function(t) for t in tools] | ||
| ) | ||
| agent = ( | ||
| RunnablePassthrough.assign( |
There was a problem hiding this comment.
should we add a nice run_name
| - `input`: this is the user input | ||
| - `agent_scratchpad`: this is the history of steps the agent has taken so far. | ||
| This gets populated by the AgentExecutor. | ||
| - `chat_history`: This is optionally required as input, if |
There was a problem hiding this comment.
"optionally required" is confusing
| return ChatPromptTemplate(messages=messages) | ||
|
|
||
|
|
||
| def create_openai_functions_agent( |
There was a problem hiding this comment.
Worth renaming to create_functions_agent and dropping the openai and having typing be some kind of FunctionCallingChatModel since you could do technically do this with any functions-capable model?
dosubot
bot
added
size:XXL
|
|
||
| prompt = prompt.partial( | ||
| tools=render_text_description(list(tools)), | ||
| tool_names=", ".join([t.name for t in tools]), |
There was a problem hiding this comment.
Don't think you need this here? It's not in the hub prompt
|
|
||
|
|
||
| def create_xml_agent( | ||
| llm: BaseLanguageModel, tools: Sequence[BaseTool], prompt: ChatPromptTemplate |
There was a problem hiding this comment.
No reason this one can't be BasePromptTemplate?
cookbook/human_approval.ipynb
Outdated
There was a problem hiding this comment.
Do we need a redirect for these?
CC @baskaryan
| This agent uses the [ReAct](https://arxiv.org/pdf/2210.03629) framework to determine which tool to use | ||
| based solely on the tool's description. Any number of tools can be provided. | ||
| This agent requires that a description is provided for each tool. | ||
| Whether this agent is intended for Chat Models (takes in messages, outputs message) or LLMs (takes in string, outputs string). The main thing this affects is the prompting strategy used. You can use an agent with a different type of model than it is intended for, but it likely won't product as good of results. |
There was a problem hiding this comment.
product -> produce results of the same quality
<!-- Thank you for contributing to LangChain! Please title your PR "<package>: <description>", where <package> is whichever of langchain, community, core, experimental, etc. is being modified. Replace this entire comment with: - **Description:** a description of the change, - **Issue:** the issue # it fixes if applicable, - **Dependencies:** any dependencies required for this change, - **Twitter handle:** we announce bigger features on Twitter. If your PR gets announced, and you'd like a mention, we'll gladly shout you out!
Made a stupid typo in the last PR which got already merged😅 revamp getting started (#15070) Harrison/agents rewrite (#15028) havent marked the class as deprecated yet, will likely want to do all in one go (with other classes) --------- Co-authored-by: Nuno Campos <[email protected]> ModelIO revamp (#15230) Retrieval Docs Revamp (#15238) [core, langchain] modelio code improvements (#15277) [langchain] agents code changes (#15278) <!-- Thank you for contributing to LangChain! Please title your PR "<package>: <description>", where <package> is whichever of langchain, community, core, experimental, etc. is being modified. Replace this entire comment with: - **Description:** a description of the change, - **Issue:** the issue # it fixes if applicable, - **Dependencies:** any dependencies required for this change, - **Twitter handle:** we announce bigger features on Twitter. If your PR gets announced, and you'd like a mention, we'll gladly shout you out! move stuff
havent marked the class as deprecated yet, will likely want to do all in one go (with other classes)