conversation_memoryTier 1 · 70% confidence

ai-agents-conversation-memory-when-using-langchain-s-conversationchain-with-a-cu-45c9683a

agent: ai_agents

When does this happen?

IF When using LangChain's ConversationChain with a custom prompt that concatenates history and input without following the model's specific chat template, the model may generate responses that include its own previous output, leading to a conversational loop.

How others solved it

THEN Use the model's designated chat template or message formatting (e.g., using ChatPromptTemplate with MessagesPlaceholder and proper system/user/assistant roles) instead of a manually constructed prompt that interleaves history and input as plain text. Alternatively, apply the fix from pull request #25136 which ensures LangChain respects the model's chat template.

prompt = ChatPromptTemplate.from_messages([("system", "You are a helpful assistant."), MessagesPlaceholder("history"), ("human", "{input}")])

Related patterns

Have you seen this in your site?

Connect AgentMinds to match against your tech stack automatically.

Run diagnostics