Revalvo

Description

Revalvo is a local-first workbench designed for evaluating large language models (LLMs) by running prompts across multiple models simultaneously. It provides a comprehensive environment where users can write, run, version, and evaluate prompts before they reach production. Unlike traditional chat UIs, Revalvo offers full control over prompts, models, and variables, ensuring transparency and traceability with scores, diffs, and version history.

The platform is not a typical SaaS; it requires no account creation, and all data remains in the user's browser, ensuring privacy and security. Users can connect their providers using keys, OAuth, or local runtimes, and create separate workspaces for each provider. The playground feature allows users to write once and run everywhere, with models responding to the same prompt in parallel, providing metrics such as latency, tokens, and cost per response.

Revalvo's version control system treats prompts like code, with immutable snapshots, Git-style version history, and the ability to roll back to any previous version. Users can export versions as API snippets or agent prompts and duplicate prompts across workspaces for fresh evaluations.

The platform supports large-scale testing with datasets, offering 40 batch-runnable scorers for structure, AI judgment, performance, and safety. Reports generated from batch runs include pass rates, latency, and cost per model, providing a detailed receipt of readiness for shipping.

Revalvo is optimized for efficiency, with thoughtful defaults and minimal setup time. It balances the speed of chat playgrounds with the rigor of full evaluation platforms, making it suitable for both first-time users and experienced prompt engineers.

Revalvo's Core Features

• Local-first LLM evaluation

• Real-time scoring and version control

• No account required

• Full transparency with scores and diffs

• Connect via keys, OAuth, or local runtimes

• Parallel model response evaluation

• Git-style version history

• Batch testing with 40 scorers

How to use Revalvo?

1. 1

   Configure: Connect your provider using keys or OAuth

2. 2

   Use: Run prompts across multiple models in parallel

3. 3

   Optimise: Save and version your prompts with Git-style history

4. 4

   Test: Run batch tests with datasets and scorers

Revalvo's Use Cases

• Prompt Evaluation
• Version Control
• Batch Testing
• Model Comparison
• Data Privacy

FAQ from Revalvo

• What is Revalvo?

  Revalvo is a local-first LLM evaluation workbench that allows users to run prompts on multiple models simultaneously with real-time scoring and version control.

• Do I need an account to use Revalvo?

  No, Revalvo does not require an account. All data is stored locally in your browser.

• How does Revalvo handle version control?

  Revalvo uses Git-style version control, allowing users to save immutable snapshots and roll back to previous versions.

• Can I connect my own model providers?

  Yes, you can connect providers using keys, OAuth, or local runtimes.

• What metrics does Revalvo provide?

  Revalvo provides metrics such as latency, tokens, and cost per response for each model.

• Is Revalvo suitable for large-scale testing?

  Yes, Revalvo supports large-scale testing with datasets and offers 40 batch-runnable scorers.

• How does Revalvo ensure data privacy?

  All data remains in the user's browser, and keys are never uploaded or logged server-side.

• What is the playground feature in Revalvo?

  The playground allows users to write prompts once and run them across multiple models in parallel, with real-time scoring.