SAVI SDK

Overview

SAVI SDK is a lightweight Python observability SDK for LLM applications. It wraps OpenAI, Anthropic, Bedrock, Cohere, Mistral, and Vertex AI to capture cost, latency, and PII. Local mode runs without an account and makes no network call.

Best for

Best for
Python developers who need quick, offline LLM cost and latency observability

Use cases

• Track per-request LLM cost across multiple providers
• Monitor latency of OpenAI, Anthropic, and other model calls
• Detect PII in prompts and responses during development

Notes

SAVI SDK is a lightweight Python observability SDK for LLM applications. It wraps OpenAI, Anthropic, Bedrock, Cohere, Mistral, and Vertex AI to capture cost, latency, and PII. Local mode runs without an account and makes no network call.

0 stars on GitHub. Last updated 2026-09-28. Licensed MIT.

Use cases

• Track per-request LLM cost across multiple providers
• Monitor latency of OpenAI, Anthropic, and other model calls
• Detect PII in prompts and responses during development

Pros

• Supports six major LLM providers in one SDK
• Local mode works offline with no account setup
• Lightweight, focused on cost, latency, and PII

Cons

• Zero GitHub stars suggests limited community validation
• Python only, no other language bindings
• No full tracing or span-level debugging, only cost, latency, and PII

Indexed from awesome-llmops and enriched against its public facts.