apple/ml-clara

Health v2 · maintenance only

Flags: no_releases no_license

Adoption not part of the score

1153 stars · 99 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

CLaRa is Apple's open-source end-to-end Retrieval-Augmented Generation model that compresses documents into continuous latent representations (32x-64x compression) unifying retrieval and generation in a single optimized model. The repository provides training code, evaluation data, and 7B model checkpoints (Base, Instruct, E2E) on Hugging Face.

Use cases

• build a RAG system with compressed document representations
• reduce long-context costs in retrieval-augmented generation
• fine-tune a model that jointly optimizes retrieval and generation
• compress knowledge documents into latent embeddings for LLMs
• evaluate state-of-the-art RAG compression approaches
• run an end-to-end RAG model without a separate retriever

When to choose

• you want cutting-edge research RAG with document compression
• long context lengths are a bottleneck in your RAG pipeline
• you want to reproduce or build on the CLaRa paper
• you need jointly trained retrieval and generation instead of separate components

When to avoid

• you need a production-ready plug-and-play RAG framework with broad ecosystem support
• you require permissive licensing (Apple license, NOASSERTION)
• you need small models or CPU-only inference
• you want a mature tool with long-term community support

Facets

library · maturity active

rag llm-inference machine-learning search-engine large-language-models machine-learning python document-compression latent-reasoning huggingface research-model apple retrieval-augmented-generation natural-language-processing gpu linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="apple/ml-clara")