Feature Request: Support `Dict[str, str]` for labels in `inference()` (label descriptions)

[Manikandan-t]

Hi team, I am using a custom model built on top of GLiNER (knowledgator/gliner-relex-large-v1.0) and wanted to ask if there are plans to support dictionary-based labels in the base library.

The Context
The model card for the custom knowledgator model instructs users to pass a dictionary (mapping labels to their descriptions) to the inference() method:

entity_labels = {
    "person": "A human individual, including fictional characters",
    "organization": "A company, institution, agency, or other group of people",
}

entities, relations = model.inference(
    texts=[text],
    labels=entity_labels,  # Passing a Dict instead of a List
    # ...
)

The Issue
Running this code currently crashes the base library with a TypeError: unhashable type: 'dict' (or KeyError: 0).

This happens because BaseEncoderGLiNER.prepare_batch() (around line 2069) assumes labels is a list, and evaluating labels[0] on a dictionary triggers a key lookup error. Furthermore, the base type signature for inference() explicitly only accepts Union[str, List[str], List[List[str]]].

Environment:

• gliner version: 0.2.27
• Python: 3.13
• Branch: main

Question
Are there any plans for the base GLiNER library to officially support passing label descriptions as a Dict[str, str]? Or is this a custom feature that downstream model creators should be implementing themselves via custom wrappers?

[Ingvarstep]

@Manikandan-t , thnaks for mentioning this issue, it will be fixed in the next release.

[Manikandan-t]

@Ingvarstep Thanks. When is the next release planned for, just wanted to know? Also any development branch available having this changes that I can use for now?