v2.104.0
airweave-ai/airweavev2.104.0Jun 19, 2026by docling-ops[bot]
AI Summary
The update introduces ConfidenceScores into the service response models and adds vLLM support for specific OCR models like chandra/glm-ocr. It also resolves issues with Excel page range handling and aligns argument names for better client compatibility.
Key Highlights
- ConfidenceScores added to service response models.
- vLLM support for chandra/glm-ocr with dots bbox fix.
- Excel documents now respect the page_range parameter.
- Explicit device selection now fails hard when unavailable.
New Features
- ConfidenceScores in service responses
- vLLM support for chandra/glm-ocr
Full Release Notes
### Feature * **service:** Add ConfidenceScores to service response models ([#3651](https://github.com/docling-project/docling/issues/3651)) ([`166003b`](https://github.com/docling-project/docling/commit/166003b64aefd18447fed1513445facb33d732ce)) * Add vLLM support for chandra/glm-ocr + dots bbox fix ([#3404](https://github.com/docling-project/docling/issues/3404)) ([`2c875ca`](https://github.com/docling-project/docling/commit/2c875ca0be3520f73e84acee1c5d7175206eb9a8)) ### Fix * **service:** Align DoclingServiceClient.convert_all arg names ([#3650](https://github.com/docling-project/docling/issues/3650)) ([`f38e222`](https://github.com/docling-project/docling/commit/f38e222964fdfa659b5d10cc399e83750b9ffa48)) * Make explicit device selection fail hard when device unavailable ([#3606](https://github.com/docling-project/docling/issues/3606)) ([`3117bd1`](https://github.com/docling-project/docling/commit/3117bd1c76f94ed75bd3754a97b6fd8bb0609aa0)) * #1649 Update RTX installation instructions for PyTorch ([#3644](https://github.com/docling-project/docling/issues/3644)) ([`f847b6c`](https://github.com/docling-project/docling/commit/f847b6ccc4859aeabe07a125fffac8d70b135826)) * **xlsx:** Respect page_range in MsExcelDocumentBackend ([#3637](https://github.com/docling-project/docling/issues/3637)) ([`dc61cdd`](https://github.com/docling-project/docling/commit/dc61cdd505e4dcd511cac40f016cf39d4609889e))