LLM Workbench - Local-first playground to test compare & version LLM prompts

Hey Product Hunt community! 👋 We are thrilled to officially launch LLM Workbench v1.0.0 — our first major stable release built for developers, prompt engineers, and AI researchers who need a fast, flexible, and completely private environment to test and evaluate LLMs. Here is what’s packed into the v1.0.0 milestone: 🧪 Assertions & PASS/FAIL Evaluation: Define validation rules (regex, JSON schema, substring matching) to check output quality instantly across models. 📊 Bulk Dataset Testing: Execute prompt runs against custom CSV/JSON datasets with variable mapping and quick export options. 🛠️ Multi-Turn Tool Calling Mocks: Test agentic function-calling workflows by mocking tool responses on a 2nd-turn execution. 💻 1-Click Local Model Discovery: Instantly discover and switch between local models via Ollama and LM Studio — fully supported in air-gapped/offline setups. 🔍 Side-by-Side Visual Diffs: Compare response differences across up to 4 models simultaneously using side-by-side or unified diff views. 🔐 Zero-Trust Security Vault: Encrypted via AES-256-GCM with session-only keys. Your API keys are kept strictly in memory and never persisted on disk or sent to external servers. 📄 Git-Native .prompt Support: Import and export your prompts as Markdown files with YAML frontmatter for version control.