v0.4.0

Summary

• Fix streaming usage under-reporting (auto stream_options.include_usage for OpenAI-compatible streams; merge Anthropic SSE usage)
• SQLite WAL mode with batched writes (~200ms / 50 records)
• Per-customer spend allowances with 429/402 enforcement
• Optional X-Call-Id / X-Step, X-Unit-Id, and retry correlation (X-Correlation-Id / X-Attempt)
• Dashboard: step drill-down, cost-per-unit, waste spend metric

Package: llm-agent-metering==0.4.0

v0.3.1

llm-agent-metering 0.3.1

PyPI distribution renamed to llm-agent-metering (previous name was blocked as too similar to existing packages).

Install

pip install llm-agent-metering

Import path unchanged: import agent_metering

PyPI Trusted Publisher (required once)

On https://pypi.org → Publishing → Pending publisher:

• Project: llm-agent-metering
• Owner: prantakhandaker
• Repository: agent_metering
• Workflow: publish.yml
• Environment: leave empty

v0.3.0

agent-metering 0.3.0

First PyPI release via GitHub Actions Trusted Publishing.

• Any-language HTTP metering proxy
• Python install-only autoload (.pth)
• Automatic user attribution (frameworks + X-User-Id / body user)

Install

pip install agent-metering

Note: Before this workflow succeeds, configure a Pending publisher on https://pypi.org for project agent-metering (owner prantakhandaker, repo agent_metering, workflow publish.yml).