# Syntropylabs

> LLM evaluation and AI observability platform for engineering teams.

Syntropylabs is a web platform that lets teams rigorously test and evaluate large language models (LLMs) before and after deployment. It supports text, voice, image, and video AI workflows.

## What it does

- **Dataset-based evaluation**: Upload CSV datasets and run automated LLM benchmarks against them.
- **Multi-model comparison**: Test the same dataset across multiple AI providers (OpenAI, Anthropic, Google Gemini, Mistral, Cohere, and others) side by side.
- **Custom evaluators**: Define evaluation rules using custom prompts, LLM-as-judge, or statistical metrics (ROUGE, BLEU).
- **Voice AI evaluation**: Evaluate voice agents with speech-to-text, LLM response, and text-to-speech pipelines — in batch or live session modes.
- **Image and video evaluation**: Test image generation and video generation models.
- **Playground**: Chat directly with any configured model to explore behavior before adding it to a formal evaluation run.
- **Projects and organizations**: Multi-tenant workspace model — teams can organize evaluation work into projects within an organization.

## Who it is for

Engineering teams and AI practitioners who need to measure LLM quality, compare model outputs, and make data-driven decisions about which models to use in production.

## Key concepts

- **Evaluation rule**: A scoring criterion — either a custom LLM judge prompt, a predefined statistical metric, or both.
- **Dataset**: A CSV file with input prompts and (optionally) reference answers. Outputs are generated by running a model against each row.
- **Evaluation job**: A batch run that applies one or more evaluation rules to every row in a dataset and aggregates scores.
- **Voice eval job**: A batch run using a full voice pipeline (STT → LLM → TTS) per row, or an interactive live session.
- **Model catalog**: The user's saved model configurations (provider, model name, API key) reused across evaluations.

## Website

https://syntropylabs.ai/

## Crawlable URLs

- Homepage: https://syntropylabs.ai/
- Blog index: https://syntropylabs.ai/blog
- Sitemap: https://syntropylabs.ai/sitemap.xml
- Robots: https://syntropylabs.ai/robots.txt