Add new OCR model profiles and move local runtimes to UV-managed installs - #18
Merged
Conversation
BREAKING CHANGE: The `vllm` OCR provider and `VLLMOptions` are removed. Use a served vLLM runtime through `provider="openai-compatible"` instead.
Remove the repo-managed vLLM install and serve path in favor of external OpenAI-compatible servers. BREAKING CHANGE: The `vllm` extra, `churro-ocr install vllm`, and `churro-ocr serve-vllm` were removed. Use an external vLLM OpenAI-compatible server with the `openai-compatible` backend instead.
Update the published HF dependency range to `transformers>=5,<6` and refresh the lockfile accordingly. Add and align test coverage for the Transformers v5 chat-template contract and package metadata expectations.
Register the built-in rednote-hilab/dots.mocr profile for both Hugging Face and openai-compatible backends, export the template preset, and cover the integration with provider, CLI, benchmark, and documentation updates.
Record the CHURRO test-set vLLM benchmark entry for rednote-hilab/dots.mocr in benchmark_results.json.
s-jse
changed the title
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
This PR expands Churro’s built-in OCR model coverage while simplifying runtime setup and standardizing the self-hosted serving path.
churro-ocr install <target>and slim package extras so heavy local runtimes are installed into the active environment on demandvllmprovider and managed vLLM runtime support; documentopenai-compatibleas the supported path for self-hosted servers such as vLLM, and makeapi_keyoptional for that backendolmOCR(including FP8),datalab-to/chandra-ocr-2,LiquidAI/LFM2.5-VL-1.6B,PaddleOCR-VL-1.5,rednote-hilab/dots.mocr, anddeepseek-ai/DeepSeek-OCR-2transformers5.xBreaking Changes
provider="vllm"/--backend vllmhas been removed. Served local models should now be accessed throughprovider="openai-compatible"/--backend openai-compatible.mistral-ocr-2505ormistral-ocr-2512);mistral-ocr-latestis no longer acceptedDocs And Benchmarks
benchmark_results.jsonand the leaderboard UI with model icons plus new benchmark entries/results includingdots.mocr,DeepSeek OCR 2, andGemini 3.1 Pro Preview