conduct-litellm-guard
pypi.org- Category
- Developer Tools
- Rank
- No. 2292Tools index
Previous survey · No. 2173 ·
- Type
- TOOL
- Use case
- Security & Identity
- Interfaces
- Host Plugin
- Date
About
conduct-litellm-guard is a Python package that wires Conduct Guard into LiteLLM as a guardrail, so every LLM call routed through a LiteLLM proxy is checked against a configured policy before it reaches the model. Depending on the policy's verdict it can allow, warn, block, or hold a request for human-in-the-loop approval, and it logs each decision to a hash-chained audit trail.
What it does
The PyPI page says this package plugs Conduct Guard into a LiteLLM proxy, so each model request is checked by Conduct first. It lists five verdicts, from forwarding a request to refusing it or holding it until a person approves, and says setup needs an agent token minted in the Conduct console.
Stated on the product site
- License
- Apache 2.0
- Install
- pip package, added to the LiteLLM proxy config
- Credential
- Agent token from the Conduct console
- Default when Guard is unreachable
- Blocks the call
- MCP tool calls
- Also checked, with the real tool name and arguments
- Python versions
- 3.10, 3.11 and 3.12 listed
- Development status
- Beta
Not stated on the site
- The page does not show pricing for the Conduct service the guardrail calls on every request.
- The page says Guard checks return in under 100ms in the healthy path, but shows no measurements behind that figure.
- The page does not say how long Conduct keeps the request summary and last user message it receives.
Written from the product site at pypi.org.
Tags
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.