{"id":51326822,"url":"https://github.com/darylalim/nuextract-studio","last_synced_at":"2026-07-01T19:02:28.738Z","repository":{"id":337872570,"uuid":"1124437549","full_name":"darylalim/nuextract-studio","owner":"darylalim","description":"Streamlit application for structured extraction, document understanding, and template generation with NuMind NuExtract on Apple Silicon with MLX.","archived":false,"fork":false,"pushed_at":"2026-07-01T02:53:16.000Z","size":1723,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-07-01T04:26:30.480Z","etag":null,"topics":["document-to-markdown","document-understanding","mlx-vlm","nuextract","structured-extraction"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/darylalim.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2025-12-29T02:52:48.000Z","updated_at":"2026-07-01T02:53:20.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/darylalim/nuextract-studio","commit_stats":null,"previous_names":["darylalim/text-extraction-pipeline","darylalim/nuextract-pipeline","darylalim/nuextract-studio"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/darylalim/nuextract-studio","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/darylalim%2Fnuextract-studio","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/darylalim%2Fnuextract-studio/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/darylalim%2Fnuextract-studio/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/darylalim%2Fnuextract-studio/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/darylalim","download_url":"https://codeload.github.com/darylalim/nuextract-studio/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/darylalim%2Fnuextract-studio/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":35019037,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-07-01T02:00:05.325Z","response_time":130,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["document-to-markdown","document-understanding","mlx-vlm","nuextract","structured-extraction"],"created_at":"2026-07-01T19:02:28.046Z","updated_at":"2026-07-01T19:02:28.733Z","avatar_url":"https://github.com/darylalim.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# NuExtract Studio\n\nStreamlit application for structured extraction, document understanding, and template generation with NuMind NuExtract on Apple Silicon with MLX. Mirrors the [official NuExtract3 Hugging Face Space](https://huggingface.co/spaces/numind/NuExtract3) but runs entirely locally via [mlx-vlm](https://github.com/Blaizzy/mlx-vlm), no GPU or external API required.\n\n## Features\n\n- **Three modes** — structured JSON extraction, document-to-markdown conversion, and natural-language → template generation\n- **Multimodal input** — upload an image (screenshot, scan, photo) and/or paste text\n- **Typed template system** — `verbatim-string`, `string`, `integer`, `number`, `date`, `boolean`, enums, multi-enums, and more (see [TYPES.md on Hugging Face](https://huggingface.co/numind/NuExtract3/blob/main/TYPES.md))\n- **Streaming output** — results stream in token-by-token\n- **Optional reasoning mode** — model emits `\u003cthink\u003e...\u003c/think\u003e` traces shown in a dedicated pane\n- **Download buttons** — save the JSON, markdown, or generated template to disk\n- **Local Apple Silicon inference** — no API key, no network calls during extraction\n\n## Requirements\n\n- macOS with Apple Silicon (M1/M2/M3/M4)\n- Python 3.12+\n- ~6 GB free disk — the model is ~5 GB (downloaded on first run), plus headroom\n- 16 GB unified memory recommended (more is better for long-context inference)\n\n## Installation\n\n```bash\nuv sync\n```\n\n## Usage\n\n```bash\nuv run streamlit run streamlit_app.py\n```\n\nFirst run downloads the ~5 GB model. Subsequent runs use the local cache.\n\n## Modes\n\n| Button | Inputs | Output |\n|---|---|---|\n| **Extract JSON** | Image and/or text + JSON template | Structured JSON matching the template |\n| **Convert to Markdown** | Image (required) | Clean markdown with HTML tables and embedded structure |\n| **Generate template** | Image or text describing the document | A JSON template you can paste back into the editor |\n\n## Testing\n\n```bash\nuv run pytest                              # Run all tests\nuv run python scripts/probe_mlx_vlm.py     # End-to-end model probe (downloads + extracts)\n```\n\n## Development\n\n```bash\nuv run ruff check .      # Lint\nuv run ruff format .     # Format\nuv run ty check          # Type check\n```\n\nCI (GitHub Actions, Apple Silicon runners) runs lint, format check, type check, and tests on every push and PR to `main`.\n\n## Project Structure\n\n```\nstreamlit_app.py                    # UI: two-pane layout, buttons + streamed output in an st.fragment\nnuextract.py                        # mlx-vlm wrapper: load, render prompt, stream extraction\n.streamlit/\n  config.toml                       # Theme: GitHub-inspired light/dark palette\nscripts/\n  probe_mlx_vlm.py                  # Verifies model + template kwargs flow-through end-to-end\ntests/\n  conftest.py                       # sys.path setup\n  test_nuextract.py                 # Wrapper tests (40)\n  test_streamlit_app.py             # App helper tests (25)\n  test_streamlit_app_apptest.py     # End-to-end UI tests via Streamlit AppTest (25)\n```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fdarylalim%2Fnuextract-studio","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fdarylalim%2Fnuextract-studio","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fdarylalim%2Fnuextract-studio/lists"}