ElectriPy Studio
Production-grade Python toolkit for AI product engineering.
Composable, typed, production-minded Python building blocks for LLM gateways, policy guardrails, retrieval evaluation, telemetry, resilience, and deterministic AI workflows. Built for engineers who ship AI products to production.
pip install electripy-studioWhat It Is
A toolkit, not a framework.
ElectriPy Studio is a collection of focused, composable Python components. Each component does one thing well. You pick the pieces you need — LLM gateway, policy enforcement, RAG evaluation, telemetry — and wire them together your way.
No magic, no monolithic framework lock-in. Provider-agnostic LLM access, typed APIs throughout, and deterministic testing support built in. Designed to survive contact with production.
Early alpha. The interfaces are stabilizing. Contributions and feedback welcome.
Component Categories
7 focused categories. 30+ components.
Core
5 components
Foundational building blocks and utilities
Concurrency & Resilience
5 components
Resilience primitives for production workloads
AI / LLM Infrastructure
7 components
Provider-agnostic LLM infrastructure
Safety / Policy / Validation
5 components
Policy seams, guardrails, and validation utilities
RAG / Evaluation
4 components
Retrieval-augmented generation evaluation tooling
Telemetry / Observability
4 components
Observability and cost tracking for AI workloads
CLI Tooling
4 components
Command-line developer experience
Featured Components
Highlights from the toolkit.
LLM Gateway
Unified interface for LLM providers with policy hooks and telemetry.
Policy Gateway
Pre/post request policy hooks with block, warn, and transform actions.
Circuit Breaker
Trip-and-recover circuit breaker for external service calls.
RAG Evaluation Runner
Run evaluation suites against RAG pipelines with configurable scorers.
AI Telemetry
Structured telemetry events for LLM requests, latency, and outcomes.
Config Loader
Type-safe configuration loading with env, file, and override support.
Structured Output Engine
Force structured JSON output from any LLM with schema validation.
Rate Limiter
Token-bucket and sliding-window rate limiters for API calls.
Sensitive Data Scanner
Detect and redact PII, credentials, and sensitive patterns in prompts.
Replay Tape
Record and replay LLM interactions for deterministic testing.
Cost Ledger
Token-level cost tracking across providers and models.
Fallback Chain
Automatic provider fallback with configurable priority and conditions.
Answer Quality Evaluator
LLM-as-judge evaluation for answer correctness and coherence.
Recipes
End-to-end working examples.
CLI Tool
Build typed CLI tools with ElectriPy's CLI framework. Subcommands, typed arguments, help generation, and config loading wired together.
LLM Gateway
Provider-agnostic LLM gateway with sync/async support. Wraps OpenAI, Anthropic, Ollama, and HTTP-JSON providers behind a unified interface.
Policy-Governed LLM Flow
LLM requests with pre/post policy hooks. Block, warn, or transform requests and responses at runtime using a declarative policy chain.
Documentation
Everything you need to get started.
Getting Started
User Guide
- CLI Guide
Build and ship typed CLI tools using the ElectriPy CLI framework.
- AI Product Engineering
Patterns and primitives for production AI product development.
- Policy Gateway
Enforce content policies and guardrails on LLM traffic.
- Agent Collaboration
Coordinate multi-agent workflows with shared context and handoffs.
- RAG Eval Runner
Evaluate retrieval-augmented generation quality systematically.
Reference
Status & Maturity
Early alpha. Honest about it.
ElectriPy Studio is under active development. Components have varying maturity levels. We label everything clearly so you can make informed decisions.
API is settled. Suitable for production use with minor version risk.
API may have minor changes. Tested, but breaking changes possible.
API is stabilizing. Expect changes. Use with awareness of churn.
Early proof-of-concept. May be removed or fundamentally changed.
Key Facts
- →Python 3.11+ required. Type-checked throughout.
- →Async-first design with sync wrappers where sensible.
- →No heavy frameworks as dependencies — stdlib and minimal third-party.
- →MIT License. Use it, fork it, build on it.
- →Actively developed. Interfaces stabilizing as of 2025.
Ready to explore the toolkit?
Install ElectriPy Studio, browse the component docs, and check out the recipes to see how the pieces fit together.