Starstruk Logo MarkStarstruk

v0.4.0

D4Vinci/Scraplingv0.4.0Nov 9, 2025by github-actions[bot]

AI Summary

Adds support for the PaddleOCR-VL backend, introduces lazy loading for faster startup, and enhances sampling controls across the CLI and server.

Key Highlights

  • Adds PaddleOCR-VL backend (SigLIP + Ernie 0.9B) alongside DeepSeek-OCR.
  • Implements lazy loading to defer weight memory mapping and reduce startup time.
  • Expands sampling controls (--do-sample, --temperature, --top-p, etc.) to CLI and Server.
  • Improves prompt handling with model-aware feedback and bilingual warnings.
  • Refactors common utilities into deepseek-ocr-core crate.

New Features

  • PaddleOCR-VL backend integration.
  • Lazy loading mechanism.
  • Comprehensive sampling controls (--do-sample, --temperature, --top-p, --top-k, --repetition-penalty, --no-repeat-ngram-size).
  • Model-aware prompt builder with bilingual Markdown feedback.
  • ModelScope provider fixes for arbitrary repo IDs and file paths.

Full Release Notes

## New Features

  - `PaddleOCR‑VL` backend (SigLIP + Ernie 0.9B with FlashAttention) is now selectable alongside `DeepSeek‑OCR`. Documentation covers model switching and architecture/memory differences.
  - Model-aware prompts with bilingual Markdown feedback when requests omit <image> placeholders.
  - Lazy loading: the server defers weight mmap until the first request, reducing startup time.
  - Server/CLI sampling controls: `--do-sample`, `--temperature`, `--top-p`, `--top-k`, `--repetition-penalty`, and `--no-repeat-ngram-size` are now recognized across both entry points.

  ## Improvements

  - Common utilities (token sampling, embedding gathers, transformer KV cache) moved into deepseek-ocr-core, trimming duplication between DeepSeek and Paddle crates.
  - Documentation clarifies multi-model selection, DeepSeek-only dynamic crop mode, bilingual terminology, and per-flag behavior.
  - CLI/Server prompt builders choose the correct format per model, improving output quality without manual tweaks.

  ## Bug Fixes

  - ModelScope provider now respects arbitrary repo IDs and exact file paths, fixing Paddle asset downloads that previously fetched the wrong config.json.
  - Requests without images no longer throw transport errors; both sync and streaming responses return a structured bilingual warning instead.
  - Prompt/image mismatches surface as normal assistant replies instead of opaque “prompt formatting failed” errors, keeping clients compatible with standard OpenAI flows.