# Gaussia

## Docs

- [Introduction](https://docs.gaussia.ai/index.md): Scientific metrics for intelligent behaviors. If you can't trace it to a paper, it's not a metric — it's an opinion.
- [Contributing](https://docs.gaussia.ai/development.md): How to contribute to Gaussia — from proposing metrics to building SDKs.
- [Introduction](https://docs.gaussia.ai/sdks/python/index.md): Gaussia is a comprehensive performance-measurement library for evaluating AI models and assistants
- [Installation](https://docs.gaussia.ai/sdks/python/installation.md): Install Gaussia and its dependencies
- [Quickstart](https://docs.gaussia.ai/sdks/python/quickstart.md): Get started with Gaussia in minutes
- [Architecture](https://docs.gaussia.ai/sdks/python/concepts/architecture.md): Understanding Gaussia's core architecture and design patterns
- [Retriever](https://docs.gaussia.ai/sdks/python/concepts/retriever.md): Implement custom data retrievers to load conversation data from any source
- [Datasets and batches](https://docs.gaussia.ai/sdks/python/concepts/datasets.md): Understand the Dataset and Batch data models that structure conversation data for evaluation
- [Statistical modes](https://docs.gaussia.ai/sdks/python/concepts/statistical-modes.md): Choose between frequentist point estimates and Bayesian credible intervals for metric aggregation
- [LLM judge](https://docs.gaussia.ai/sdks/python/concepts/llm-judge.md): Use any LangChain-compatible model as an evaluation judge for metric scoring
- [Metrics Overview](https://docs.gaussia.ai/sdks/python/metrics/overview.md): Overview of all available metrics in Gaussia
- [Context](https://docs.gaussia.ai/sdks/python/metrics/context.md): Evaluate how well AI responses align with provided context, with session-level aggregation and pluggable statistical modes
- [Conversational](https://docs.gaussia.ai/sdks/python/metrics/conversational.md): Evaluate dialogue quality using Grice's Maxims with session-level aggregation and pluggable statistical modes
- [Toxicity](https://docs.gaussia.ai/sdks/python/metrics/toxicity.md): Measure toxic language with clustering and demographic group profiling
- [Bias](https://docs.gaussia.ai/sdks/python/metrics/bias.md): Detect bias across protected attributes using guardian-based analysis
- [Humanity](https://docs.gaussia.ai/sdks/python/metrics/humanity.md): Measure emotional profiling and entropy of AI assistant responses using NRC emotion lexicons
- [BestOf](https://docs.gaussia.ai/sdks/python/metrics/best-of.md): Tournament-style comparison of multiple AI assistants using king-of-the-hill evaluation
- [Agentic](https://docs.gaussia.ai/sdks/python/metrics/agentic.md): Evaluate AI agent responses with pass@K metrics, tool correctness, and pluggable statistical modes
- [Vision](https://docs.gaussia.ai/sdks/python/metrics/vision.md): Evaluate vision-language model descriptions for similarity and hallucination detection
- [Regulatory](https://docs.gaussia.ai/sdks/python/metrics/regulatory.md): Evaluate AI response compliance against a regulatory document corpus
- [Use Gaussia as an EvalHub Provider](https://docs.gaussia.ai/sdks/python/integrations/evalhub.md): Run Gaussia as a BYOF provider in EvalHub, map conversational payloads into gaussia.Dataset, and orchestrate benchmark execution with MLflow-backed evidence.
- [Generators](https://docs.gaussia.ai/sdks/python/advanced/generators.md): Generate synthetic evaluation datasets from context documents using LLM-powered generation
- [Prompt optimizer](https://docs.gaussia.ai/sdks/python/advanced/prompt-optimizer.md): Optimize LLM prompts using GEPA and MIPROv2 algorithms
- [Explainability](https://docs.gaussia.ai/sdks/python/advanced/explainability.md): Analyze token-level attributions to understand which input tokens drive model outputs
- [Gaussia TypeScript SDK](https://docs.gaussia.ai/sdks/typescript/index.md): Evaluate LLM applications in TypeScript with the @gaussia/sdk package.
- [Installation](https://docs.gaussia.ai/sdks/typescript/installation.md): Install @gaussia/sdk and its peer dependencies for Node or the browser.
- [Quickstart](https://docs.gaussia.ai/sdks/typescript/quickstart.md): Score a dataset end to end: a retriever supplies data, an evaluator uses a model to judge each answer.
- [Architecture](https://docs.gaussia.ai/sdks/typescript/concepts/architecture.md): Module boundaries, subpath exports, and the isomorphic design of @gaussia/sdk.
- [Evaluators](https://docs.gaussia.ai/sdks/typescript/concepts/evaluators.md): Subclass the Gaussia base class to evaluate datasets batch by batch.
- [Language models & adapters](https://docs.gaussia.ai/sdks/typescript/concepts/language-models.md): The vendor-neutral LanguageModel contract and the adapters that satisfy it.
- [Schemas](https://docs.gaussia.ai/sdks/typescript/concepts/schemas.md): The Zod-derived data model exported from @gaussia/sdk/schemas.
- [Generators](https://docs.gaussia.ai/sdks/typescript/generators/overview.md): Turn context documents into validated evaluation datasets with @gaussia/sdk/generators.
- [GEPA prompt optimizer](https://docs.gaussia.ai/sdks/typescript/prompt-optimizer/gepa.md): Improve a system prompt against a dataset with GEPA — custom judges, custom executors, tuning, and full history.
- [Logprob evaluator](https://docs.gaussia.ai/sdks/typescript/prompt-optimizer/logprob-evaluator.md): Score prompt-optimizer candidates from Yes/No token log probabilities for better-calibrated judging.
- [Boltzmann Python SDK](https://docs.gaussia.ai/sdks/boltzmann/index.md): An SDK for the Boltzmann Protocol: portable, verifiable, model-agnostic knowledge.
- [Installation](https://docs.gaussia.ai/sdks/boltzmann/installation.md): Install pyboltzmann, and understand why the install name and the import name differ.
- [Quickstart](https://docs.gaussia.ai/sdks/boltzmann/quickstart.md): The whole lifecycle against a real OCI layout: ingest, query, prove, publish, remove.
- [Architecture](https://docs.gaussia.ai/sdks/boltzmann/concepts/architecture.md): Blocks, modules, compositions and snapshots: the four layers a brain is made of.
- [Memory types](https://docs.gaussia.ai/sdks/boltzmann/concepts/memory-types.md): The five typed blocks, what each one is for, and which rules each module obeys.
- [Identity](https://docs.gaussia.ai/sdks/boltzmann/concepts/identity.md): Canonical serialization, the three levels of hashes, and the values a payload refuses.
- [Merkle DAGs](https://docs.gaussia.ai/sdks/boltzmann/concepts/merkle.md): The layout the SDK commits to, why it sorts leaves, and how membership is proven.
- [Interfaces](https://docs.gaussia.ai/sdks/boltzmann/concepts/interfaces.md): The protocol surface a client satisfies, and the four things you plug in.
- [Ingestion](https://docs.gaussia.ai/sdks/boltzmann/guides/ingestion.md): Preserve the source, delegate the interpretation, govern what is stored.
- [Query](https://docs.gaussia.ai/sdks/boltzmann/guides/query.md): Ask for data with its provenance, never prose, and verify every match against the snapshot.
- [Retention](https://docs.gaussia.ai/sdks/boltzmann/guides/retention.md): Four distinct removal mechanisms, the cascade a drop requires, and what pruning reclaims.
- [Distribution](https://docs.gaussia.ai/sdks/boltzmann/guides/distribution.md): Pack, push and pull: a brain moves between a remote artifact and a local layout.
- [Conformance](https://docs.gaussia.ai/sdks/boltzmann/guides/conformance.md): Prove an implementation conforms — in Python by inheriting the suite, in any language from the golden vectors.

## OpenAPI Specs

- [openapi](https://docs.gaussia.ai/api-reference/openapi.json)

## Optional

- [GitHub](https://github.com/gaussia-labs)
- [Papers](https://github.com/gaussia-labs/papers)