aispendguard-proxy

Lightweight cost-tracking proxy for AISpendGuard. Track LLM spend with zero code changes — just swap your base URL.

How it works

1. Point your OpenAI/Anthropic SDK to the proxy URL instead of the provider
2. The proxy forwards requests transparently to the real provider
3. Usage metadata (model, tokens, latency) is extracted from responses
4. Usage events are sent to AISpendGuard for cost tracking

Privacy: The proxy never reads, logs, or stores request/response content. Only metadata is extracted.

Quick start

# Python — OpenAI SDK
from openai import OpenAI

client = OpenAI(
    base_url="https://proxy.aispendguard.com/v1",
    default_headers={
        "X-ASG-Key": "asg_your_api_key",
        "X-ASG-Feature": "chat",
        "X-ASG-Task-Type": "completion",
        "X-ASG-Route": "/api/chat",
    },
)

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

Custom headers

Header	Required	Description
X-ASG-Key	Yes	AISpendGuard workspace API key
X-ASG-Feature	No	Feature tag for cost attribution
X-ASG-Task-Type	No	Task type tag
X-ASG-Route	No	Route tag
X-ASG-Provider	No	Override provider detection
X-ASG-Base-URL	No	Custom provider URL (allowlisted domains only)

Self-hosting

docker build -t asg-proxy .
docker run -p 8787:8787 -e ASG_INGEST_URL=https://aispendguard.com/api/ingest asg-proxy

Development

npm install
npm run dev      # Start with tsx (hot reload)
npm test         # Run tests
npm run build    # Compile TypeScript

License

MIT