David Young
David Young
No movement on this in ~8 months... Any update?
This should be closed. The model isn't available to all users, but works fine for those who do have access.
@lixincheng4026 The docs for @vnktsh's answer are here: https://python.langchain.com/docs/modules/agents/how_to/custom_agent_with_tool_retrieval This issue can probably be closed.
I've found a *partial* solution for this [here](https://python.langchain.com/en/latest/modules/prompts/chat_prompt_template.html?highlight=jedi#different-types-of-messageprompttemplate): If you use `ChatMessagePromptTemplate.from_template("{input}", role="Bob")`, then the "current" messages will be prefixed properly. Unfortunately, this will also cause the history messages to...
@scoupleux It looks like the defaults actually lead to the history messages saying `Human: Human: ...` anyways... something feels off here, but I haven't gotten to the bottom of it...
@Joshua-Shepherd Removing that (at least in the Python library, which this repository is for) causes there to be no role labels at all in the generated history messages. The problem...
Also, I'm thinking https://github.com/hwchase17/langchain/blob/master/langchain/memory/chat_memory.py#L35 might have something to do with it as well... No matter the role you give your message, this will always create a `HumanMessage`.
I've come up with a solution that works for arbitrary roles by using `ChatMessage` instead of `HumanMessage`: ```python from typing import Any from langchain.memory import ConversationBufferMemory from langchain.schema import ChatMessage...
@Avinash-Raj I've just updated to 0.0.208, and I'm not getting the exception. I am using this as memory for a `ChatOpenAI`, and not with agents, if that makes a difference.
@hwchase17 Is this resolved? I am pretty sure that by now almost every chain has implemented streaming, and there are specific issues open for most instances that are still not...