PromptAnalyzerAI-powered

About PromptAnalyzer

PromptAnalyzer is an open-source, local-first observability and prompt versioning toolkit for LLM applications. Add a single @track decorator to an LLM-powered Python function and PromptAnalyzer automatically captures prompt versions, inputs, responses, models, providers, token usage, costs and latency. Every system prompt is SHA-256 versioned, making it easy to see exactly which prompt produced a response and compare changes using Git-style diffs. A built-in local dashboard provides run history, usage analytics, full-text search and detailed inference inspection. PromptAnalyzer works with OpenAI, Anthropic Claude, Gemini, Ollama, vLLM, LiteLLM, OpenRouter, Groq, Mistral, Azure OpenAI and custom LLM libraries. It is designed to be lightweight and private: no cloud account, Docker stack or external observability service is required. Data can stay locally in SQLite, while PostgreSQL is also supported. The core idea is especially strong because installation is simply pip install "promptanalyzer[dashboard]", after which a developer decorates their function and can launch the dashboard locally.

Why choose PromptAnalyzer

• One decorator to start tracking your LLM application
• Local-first — your observability data can stay on your machine
• No Docker, cloud account or frontend setup required
• Works across major LLM providers and custom libraries
• Git-style version history makes prompt changes traceable
• Open-source and MIT licensed

Open source

At a glance