Starstruk Logo MarkStarstruk

v0.4.0

apache/answerv0.4.0Nov 9, 2025by github-actions[bot]

AI Summary

Adds the PaddleOCR-VL backend alongside DeepSeek-OCR, introduces lazy loading of weights to improve startup times, and unifies sampling controls between the CLI and server.

Key Highlights

  • New PaddleOCR-VL backend (SigLIP + Ernie 0.9B) selectable alongside DeepSeek-OCR.
  • Lazy loading of weights reduces server startup time by deferring mmap until the first request.
  • Unified sampling controls including --do-sample, --temperature, --top-p, and --top-k across CLI and server.
  • Model-aware prompts provide bilingual Markdown feedback when image placeholders are missing.

New Features

  • PaddleOCR-VL backend support.
  • Lazy loading implementation.
  • Unified sampling controls.
  • Model-aware prompt feedback system.

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.