Driftlock v0.3.0

Update docs

Driftlock v0.1.0 — Initial Release

Driftlock v0.1.0 — Initial Release

Driftlock is an open-source control layer for monitoring and enforcing policies on LLM API usage.

This first release introduces the core foundation of the project, including request tracking, cost monitoring, and a flexible policy framework for controlling how LLM calls are made in production systems.

Features

• LLM Request Observability

  • Track prompt tokens, completion tokens, latency, and cost per request

• Spend Monitoring

  • Real-time cost accounting based on model pricing

• Policy Engine

  • Define rules that evaluate each request before execution

• Local SQLite Storage

  • Lightweight database for request history and cost tracking

• CLI Interface

  • Inspect usage and cost statistics from the command line

• Developer-Friendly Integration

  • Simple Python API designed to wrap existing LLM calls

Example Use Cases

• Prevent runaway LLM costs in production systems
• Automatically downgrade models when budgets are exceeded
• Track usage across users, services, or pipelines
• Add governance and observability to AI applications

Quick Example

from driftlock import Driftlock

dl = Driftlock()

response = dl.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Hello world"}]
)

Driftlock records token usage, latency, and cost automatically.

Project Status

This is the initial public release. The core tracking and policy system are stable, but the ecosystem is still evolving. Future releases will introduce dashboards, integrations, and more advanced policy capabilities.

Links

GitHub: https://github.com/maddox-214/driftlock
Documentation: Coming soon