The open-source AI infrastructure space just got a new tool for managing proxy traffic. A developer released conduct-litellm-guard, a Python package designed to integrate with LiteLLM, the popular open-source proxy that simplifies API calls across various LLM providers.

The Core Mechanism

The package's pitch is deceptively simple: LiteLLM sends AI requests, and Conduct decides which ones shouldn't be sent. It functions as a guardrail layer, sitting between the client and the underlying model APIs. By intercepting requests at the proxy level, it allows developers to enforce policies or filter out specific types of prompts before they incur token costs or trigger hallucinations.

Integration and Documentation

The tool is available on PyPI, signaling its readiness for immediate use in existing Python-based AI stacks. The official documentation, hosted at docs.litellm.ai/docs/proxy/guardrails/conduct, outlines how to configure this filtering logic within the LiteLLM proxy settings. This integration leverages LiteLLM's existing guardrail framework, allowing teams to drop in Conduct without rewriting their entire proxy configuration.

Market Context

As LLM costs remain a significant operational concern, tools that prevent unnecessary API calls are gaining traction. While the project currently has low visibility on Hacker News, with only a single point and no comments at the time of writing, it addresses a critical pain point for developers building scalable AI applications. The ability to block requests pre-flight is a powerful optimization strategy.

Key Takeaways

  • conduct-litellm-guard is a new PyPI package for LiteLLM proxy integration.
  • It acts as a filter to prevent unwanted AI requests from reaching the model.
  • Documentation is available via the official LiteLLM guardrails docs.
  • The project is currently in early visibility stages on Hacker News.

The Bottom Line

While early-stage, Conduct represents a necessary shift toward cost-aware infrastructure. Blocking requests before they hit the API is the most efficient way to manage LLM spend, and LiteLLM's guardrail framework makes this adoption frictionless for existing users.