dottxt-ai/outlines

AI Summary: The task requires implementing a custom logits processor for the Xgrammar backend in the Outlines library to support LlamaCpp and MLXLM models. This involves adapting the existing logits processing logic to handle numpy and mlx tensors, mirroring the approach used for the LLGuidance backend.

AI Summary: The task is to standardize exception handling within the Outlines library. Currently, different Large Language Model (LLM) providers throw unique exceptions for issues like timeouts or rate limits. The goal is to create a consistent set of custom exceptions within Outlines that wrap these provider-specific exceptions, providing a uniform interface for developers regardless of the underlying LLM.

AI Summary: The task is to add support for Ollama's asynchronous client to the Outlines library. This involves modifying the existing Ollama integration to utilize the asynchronous capabilities of the Ollama Python client, allowing for non-blocking interactions with the Ollama model.