LangChain-Perplexity v1.3.2 Fixes Tool Call Serialization

LangChain has released langchain-perplexity v1.3.2, a minor patch that fixes a serialization bug in the integration with Perplexity AI. The change, tracked in PR #37911, addresses how ToolMessage and AIMessage.tool_calls are serialized. Previously, when Perplexity models generated tool calls within an agent loop, the serialization process could fail or produce inconsistent results, breaking the ability to pass tool outputs back to the model for subsequent turns. This fix ensures that both the tool call metadata and the actual message payloads are correctly marshaled, making the integration robust for multi-step agentic workflows.

The update is available via pip (`pip install langchain-perplexity==1.3.2`) and is signed with a verified GitHub signature (GPG key ID B5690EEEBB952194). While this is a small point release, it highlights the ongoing refinement of LangChain's model-specific integrations as the ecosystem matures. For teams building agent-based applications with Perplexity's large language models, this fix reduces the risk of silent failures during tool execution—a common pain point in production agent systems. The patch also improves compatibility with LangChain's evolving message schema, which increasingly relies on standardized tool call structures across different model providers.