langgraph: add handle_unanswered_tool_calls to create_react_agent
#1899
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
vbarda
changed the title
create_react_agent
vbarda
changed the title
create_react_agentcreate_react_agent
vbarda
changed the title
create_react_agentcreate_react_agent
eyurtsev
reviewed
Oct 10, 2024
| If provided, the list of messages sent to the model will be modified according to the specified strategy: | ||
|
|
||
| - "remove_tool_calls": modify AIMessages to remove tool_calls that do not have corresponding ToolMessages | ||
| - "add_tool_messages": add a ToolMessage for each unanswered tool call, following each AIMessage with unanswered tool calls |
There was a problem hiding this comment.
What's the content of those messages?
There was a problem hiding this comment.
good question -- currently it's an empty string, but i wonder if it needs to be parametrized somehow / or if there is a better default
eyurtsev
reviewed
Oct 22, 2024
| @@ -201,6 +302,14 @@ def create_react_agent( | |||
| - str: This is converted to a SystemMessage and added to the beginning of the list of messages in state["messages"]. | |||
| - Callable: This function should take in full graph state and the output is then passed to the language model. | |||
| - Runnable: This runnable should take in full graph state and the output is then passed to the language model. | |||
| handle_unanswered_tool_calls: An optional parameter that defines how to handle AIMessages with tool calls that | |||
| do not have corresponding ToolMessages. This is useful to handle errors from LLM providers that require | |||
There was a problem hiding this comment.
handle errors from LLM providers
What does this mean?
If we're adding this we should probably explain what reasonable situation could lead to such an error.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This change should address situations where LLM API providers error when sending an AI message with tool calls that do not have a corresponding
ToolMessage, e.g.BadRequestError: Error code: 400 - {'error': {'message': "An assistant message with 'tool_calls' must be followed by tool messages responding to each 'tool_call_id'....BadRequestError: Error code: 400 - {'type': 'error', 'error': {'type': 'invalid_request_error', 'message': 'messages.1: Did not find 1tool_resultblock(s) at the beginning of this message....)In the context of the
create_react_agentusage, the simplest way to trigger this is an error inside a tool without error handling inToolNode. This would lead to the agent interruption and any new human message sent to the agent would result in an error. A simple solution to address this is to call.update_stateto modify message history. But since this option might not always be acceptable, we introduce a newhandle_unanswered_tool_callsparameter with the following values:"remove_tool_calls": modify AIMessages to remove tool_calls that do not have corresponding ToolMessages"add_tool_messages": add a ToolMessage for each unanswered tool call, following each AIMessage with unanswered tool callsFalse: do nothing (default)Additional considerations:
call_modelupdate state for the AI messages with removed tool_calls? IMO no, because it would only lead to a confusing output when running agent w/stream_mode=updatesand won't actually affect underlying model behavior significantly (i.e. if the original human message is there, the model might still reintroduce a failing tool call in the future messages)