sciwrite-lint 0.5.0__tar.gz → 0.5.1__tar.gz

{sciwrite_lint-0.5.0/sciwrite_lint.egg-info → sciwrite_lint-0.5.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sciwrite-lint
-Version: 0.5.0
+Version: 0.5.1
 Summary: A linter for scientific manuscripts — reference verification, consistency checking, and structural validation.
 Author: Sergey Samsonau
 Maintainer-email: Authentic Research Partners LLC <info@arpconnect.com>
@@ -158,11 +158,11 @@ Three models — a vision model (Qwen3-VL-2B default, or 8B FP8 via `--vision-ba
 - **14-source full-text cascade** — early exit on first successful download across arXiv, Semantic Scholar, OpenAlex, PubMed Central, Europe PMC, Unpaywall, bioRxiv, NBER, RePEc/IDEAS, HAL, ERIC, NASA ADS, OSF Preprints, CORE
 - **Live monitoring** *(advanced)* — `sciwrite-lint containers monitor` shows service health, VRAM usage, and KV cache utilization in a terminal dashboard
 
-Full pipeline on a 50-reference paper: ~30 minutes initial (dominated by downloads and claim verification), minutes on cached reruns. The pipeline automatically swaps vLLM containers to free GPU for embedding and vision stages (~50× faster than CPU) — same code path on WSL2 and native Linux.
+The pipeline automatically swaps vLLM containers to free GPU for embedding and vision stages (~50× faster than CPU).
 
 ## Install
 
-**Assumed setup:** A workstation with an NVIDIA GPU (16+ GB VRAM). Tested on WSL2 with NVIDIA driver 546.01+. Native Linux is likely to work with GPU memory allocation tuning (see [docs/services.md](https://github.com/authentic-research-partners/sciwrite-lint/blob/main/docs/services.md#embedding-device)). Not tested on macOS.
+**Assumed setup:** A workstation with an NVIDIA GPU (16+ GB VRAM). Tested on WSL2 with NVIDIA driver 546.01+. Native Linux uses the same GPU-sequencing code path with no extra configuration (see [docs/services.md](https://github.com/authentic-research-partners/sciwrite-lint/blob/main/docs/services.md#embedding-device)), though it isn't actively tested. Not tested on macOS.
 
 Requires [uv](https://docs.astral.sh/uv/getting-started/installation/), a container runtime (podman or docker), CUDA drivers, and [NVIDIA Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html).
 
@@ -207,6 +207,8 @@ sciwrite-lint containers start         # start GROBID + vLLM (needs GPU for vLLM
 sciwrite-lint containers monitor       # live dashboard: service health, VRAM, KV cache
 ```
 
+`init` detects `.tex` files and their `.bib` references and generates a `.sciwrite-lint.toml` config. Review to confirm the right files were detected.
+
 ### WSL2: disable CUDA Sysmem Fallback (recommended)
 
 By default Windows silently spills GPU overflow into system RAM via PCIe (30–100× slower than VRAM); you'll see this as a sudden throughput collapse during the heaviest pipeline stages with no error in the logs. Tell the NVIDIA driver to fail loudly instead:
@@ -217,8 +219,6 @@ Driver 546.01+. After this, WSL2 behaves like native Linux on overflow (CUDA OOM
 
 ![Monitor dashboard](https://github.com/authentic-research-partners/sciwrite-lint/raw/main/docs/monitor.png)
 
-`init` detects `.tex` files and their `.bib` references and generates a `.sciwrite-lint.toml` config. Review to confirm the right files were detected.
-
 **Providing references manually — two drop folders.** sciwrite-lint reads two local directories before going to the open-access waterfall, kept separate because the folder itself signals how much to trust the source:
 
 - `local_pdfs/` — **academic sources**. Accepts `.pdf` (primary) and `.md` summaries you've written yourself. Use this for paywalled papers, for OA papers whose publisher requires a browser (Cloudflare/JS walls, Cell, Springer, some PMC pages), and for anything you'd cite as peer-reviewed evidence.
@@ -280,6 +280,22 @@ sciwrite-lint contributions --paper my_paper    # add contribution axes to SciLi
 sciwrite-lint contributions paper.pdf           # standalone file scoring
 ```
 
+**Supported inputs.**
+
+*The manuscript you check:*
+
+| Format | Notes |
+|---|---|
+| `.tex` | LaTeX — hand-written, or pandoc-generated (e.g. from Markdown); pandoc's heading anchors are unwrapped automatically. |
+| `.pdf` | Parsed via GROBID (container required). |
+
+*Cited sources you optionally supply (drop folders — see [Setup](#setup)):*
+
+| Folder | Accepts | Trust level |
+|---|---|---|
+| `local_pdfs/` | `.pdf`, `.md` (your own summaries) | academic |
+| `local_web/` | `.mhtml` / `.mht`, `.md` | web capture |
+
 `check` runs the full pipeline in one command: text checks → figure analysis → LLM consistency → reference verification via APIs → download and parse cited papers → claim verification → consistency checks on cited papers → bibliography verification → aggregate reliability scores → SciLint Score. An initial run on a 50-reference paper takes up to 30 minutes (dominated by downloads and claim verification); subsequent cached runs complete in minutes.
 
 ### Iterative editing — no `--fresh` needed

{sciwrite_lint-0.5.0 → sciwrite_lint-0.5.1}/README.md

@@ -92,11 +92,11 @@ Three models — a vision model (Qwen3-VL-2B default, or 8B FP8 via `--vision-ba
 - **14-source full-text cascade** — early exit on first successful download across arXiv, Semantic Scholar, OpenAlex, PubMed Central, Europe PMC, Unpaywall, bioRxiv, NBER, RePEc/IDEAS, HAL, ERIC, NASA ADS, OSF Preprints, CORE
 - **Live monitoring** *(advanced)* — `sciwrite-lint containers monitor` shows service health, VRAM usage, and KV cache utilization in a terminal dashboard
 
-Full pipeline on a 50-reference paper: ~30 minutes initial (dominated by downloads and claim verification), minutes on cached reruns. The pipeline automatically swaps vLLM containers to free GPU for embedding and vision stages (~50× faster than CPU) — same code path on WSL2 and native Linux.
+The pipeline automatically swaps vLLM containers to free GPU for embedding and vision stages (~50× faster than CPU).
 
 ## Install
 
-**Assumed setup:** A workstation with an NVIDIA GPU (16+ GB VRAM). Tested on WSL2 with NVIDIA driver 546.01+. Native Linux is likely to work with GPU memory allocation tuning (see [docs/services.md](https://github.com/authentic-research-partners/sciwrite-lint/blob/main/docs/services.md#embedding-device)). Not tested on macOS.
+**Assumed setup:** A workstation with an NVIDIA GPU (16+ GB VRAM). Tested on WSL2 with NVIDIA driver 546.01+. Native Linux uses the same GPU-sequencing code path with no extra configuration (see [docs/services.md](https://github.com/authentic-research-partners/sciwrite-lint/blob/main/docs/services.md#embedding-device)), though it isn't actively tested. Not tested on macOS.
 
 Requires [uv](https://docs.astral.sh/uv/getting-started/installation/), a container runtime (podman or docker), CUDA drivers, and [NVIDIA Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html).
 
@@ -141,6 +141,8 @@ sciwrite-lint containers start         # start GROBID + vLLM (needs GPU for vLLM
 sciwrite-lint containers monitor       # live dashboard: service health, VRAM, KV cache
 ```
 
+`init` detects `.tex` files and their `.bib` references and generates a `.sciwrite-lint.toml` config. Review to confirm the right files were detected.
+
 ### WSL2: disable CUDA Sysmem Fallback (recommended)
 
 By default Windows silently spills GPU overflow into system RAM via PCIe (30–100× slower than VRAM); you'll see this as a sudden throughput collapse during the heaviest pipeline stages with no error in the logs. Tell the NVIDIA driver to fail loudly instead:
@@ -151,8 +153,6 @@ Driver 546.01+. After this, WSL2 behaves like native Linux on overflow (CUDA OOM
 
 ![Monitor dashboard](https://github.com/authentic-research-partners/sciwrite-lint/raw/main/docs/monitor.png)
 
-`init` detects `.tex` files and their `.bib` references and generates a `.sciwrite-lint.toml` config. Review to confirm the right files were detected.
-
 **Providing references manually — two drop folders.** sciwrite-lint reads two local directories before going to the open-access waterfall, kept separate because the folder itself signals how much to trust the source:
 
 - `local_pdfs/` — **academic sources**. Accepts `.pdf` (primary) and `.md` summaries you've written yourself. Use this for paywalled papers, for OA papers whose publisher requires a browser (Cloudflare/JS walls, Cell, Springer, some PMC pages), and for anything you'd cite as peer-reviewed evidence.
@@ -214,6 +214,22 @@ sciwrite-lint contributions --paper my_paper    # add contribution axes to SciLi
 sciwrite-lint contributions paper.pdf           # standalone file scoring
 ```
 
+**Supported inputs.**
+
+*The manuscript you check:*
+
+| Format | Notes |
+|---|---|
+| `.tex` | LaTeX — hand-written, or pandoc-generated (e.g. from Markdown); pandoc's heading anchors are unwrapped automatically. |
+| `.pdf` | Parsed via GROBID (container required). |
+
+*Cited sources you optionally supply (drop folders — see [Setup](#setup)):*
+
+| Folder | Accepts | Trust level |
+|---|---|---|
+| `local_pdfs/` | `.pdf`, `.md` (your own summaries) | academic |
+| `local_web/` | `.mhtml` / `.mht`, `.md` | web capture |
+
 `check` runs the full pipeline in one command: text checks → figure analysis → LLM consistency → reference verification via APIs → download and parse cited papers → claim verification → consistency checks on cited papers → bibliography verification → aggregate reliability scores → SciLint Score. An initial run on a 50-reference paper takes up to 30 minutes (dominated by downloads and claim verification); subsequent cached runs complete in minutes.
 
 ### Iterative editing — no `--fresh` needed

{sciwrite_lint-0.5.0 → sciwrite_lint-0.5.1}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "sciwrite-lint"
-version = "0.5.0"
+version = "0.5.1"
 description = "A linter for scientific manuscripts — reference verification, consistency checking, and structural validation."
 requires-python = ">=3.13,<3.14"
 license = "MIT"

{sciwrite_lint-0.5.0 → sciwrite_lint-0.5.1}/sciwrite_lint/__init__.py

@@ -2,7 +2,7 @@
 
 from sciwrite_lint.exceptions import LLMConnectionError, SciWriteLintError
 
-__version__ = "0.5.0"
+__version__ = "0.5.1"
 
 from sciwrite_lint.oa import (
     DownloadResult,

{sciwrite_lint-0.5.0 → sciwrite_lint-0.5.1}/sciwrite_lint/manuscript_store.py

@@ -450,12 +450,17 @@ def _build_context_latex(tex_path: Path, config: LintConfig) -> ManuscriptContex
         split_paragraphs,
         strip_comments,
         strip_non_prose_environments_preserve_lines,
+        unwrap_hypertarget,
     )
 
     # Read + strip_comments exactly once, then pass through to the
     # section/abstract extractors. They each used to re-read the file
     # and re-strip comments — three round trips for the same data.
-    text = strip_comments(tex_path.read_text(encoding="utf-8"))
+    # unwrap_hypertarget normalises pandoc's two-line ``\hypertarget{id}{%
+    # \section{...}}`` heading envelope to plain LaTeX; without it the
+    # section splitter strands the opening brace in the prior section and
+    # pandoc rejects the unbalanced paragraph below.
+    text = unwrap_hypertarget(strip_comments(tex_path.read_text(encoding="utf-8")))
 
     raw_sections = analyze_sections_with_text(tex_path, text=text)
     abstract_raw = get_abstract_text(tex_path, text=text)

{sciwrite_lint-0.5.0 → sciwrite_lint-0.5.1}/sciwrite_lint/tex_parser.py

@@ -55,6 +55,80 @@ def strip_verbatim(text: str) -> str:
     return text
 
 
+def unwrap_hypertarget(text: str) -> str:
+    r"""Unwrap pandoc ``\hypertarget{id}{content}`` envelopes to ``content``.
+
+    Pandoc renders every markdown heading as a two-line anchor wrapper::
+
+        \hypertarget{the-id}{%
+        \section{Title}\label{the-id}}
+
+    The opening ``\hypertarget{the-id}{`` and its matching ``}`` sit on
+    different lines, straddling the heading. Section splitting keys off the
+    ``\section`` line and assigns the line above it to the *preceding*
+    section, which strands the unbalanced opening brace at the end of that
+    section's body. Balanced-brace consumers downstream (the pandoc prose
+    conversion in ``paragraphs_to_markdown``) then choke on the lone brace.
+
+    The anchor is a hyperlink target with no prose meaning, so we drop the
+    wrapper and keep its content, normalising pandoc output to standard
+    LaTeX. Only the wrapper tokens are removed; newlines inside the content
+    are preserved, so source line numbers are unchanged. Nested wrappers are
+    handled (scanning continues inside each content span).
+    """
+    marker = "\\hypertarget{"
+    n = len(text)
+    delete_spans: list[tuple[int, int]] = []
+    i = 0
+    while True:
+        idx = text.find(marker, i)
+        if idx == -1:
+            break
+        # Skip the first argument {id} via balanced-brace matching.
+        j = idx + len(marker)
+        depth = 1
+        while j < n and depth > 0:
+            if text[j] == "{":
+                depth += 1
+            elif text[j] == "}":
+                depth -= 1
+            j += 1
+        # The two-arg form requires the content brace immediately after.
+        if j >= n or text[j] != "{":
+            i = j
+            continue
+        content_start = j + 1
+        depth = 1
+        k = content_start
+        while k < n and depth > 0:
+            if text[k] == "{":
+                depth += 1
+            elif text[k] == "}":
+                depth -= 1
+            k += 1
+        if depth != 0:
+            break  # unbalanced input — leave the remainder untouched
+        # Remove "\hypertarget{id}{" and the matching content-closing "}".
+        delete_spans.append((idx, content_start))
+        delete_spans.append((k - 1, k))
+        # Resume inside the content so nested wrappers are also unwrapped.
+        i = content_start
+
+    if not delete_spans:
+        return text
+
+    # Nested wrappers append the outer closing brace before the inner
+    # spans, so sort before stitching the kept slices back together.
+    delete_spans.sort()
+    out: list[str] = []
+    prev = 0
+    for start, end in delete_spans:
+        out.append(text[prev:start])
+        prev = end
+    out.append(text[prev:])
+    return "".join(out)
+
+
 def extract_body(text: str) -> str:
     r"""Extract content between \begin{document} and \end{document}."""
     start = text.find("\\begin{document}")

{sciwrite_lint-0.5.0 → sciwrite_lint-0.5.1/sciwrite_lint.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sciwrite-lint
-Version: 0.5.0
+Version: 0.5.1
 Summary: A linter for scientific manuscripts — reference verification, consistency checking, and structural validation.
 Author: Sergey Samsonau
 Maintainer-email: Authentic Research Partners LLC <info@arpconnect.com>
@@ -158,11 +158,11 @@ Three models — a vision model (Qwen3-VL-2B default, or 8B FP8 via `--vision-ba
 - **14-source full-text cascade** — early exit on first successful download across arXiv, Semantic Scholar, OpenAlex, PubMed Central, Europe PMC, Unpaywall, bioRxiv, NBER, RePEc/IDEAS, HAL, ERIC, NASA ADS, OSF Preprints, CORE
 - **Live monitoring** *(advanced)* — `sciwrite-lint containers monitor` shows service health, VRAM usage, and KV cache utilization in a terminal dashboard
 
-Full pipeline on a 50-reference paper: ~30 minutes initial (dominated by downloads and claim verification), minutes on cached reruns. The pipeline automatically swaps vLLM containers to free GPU for embedding and vision stages (~50× faster than CPU) — same code path on WSL2 and native Linux.
+The pipeline automatically swaps vLLM containers to free GPU for embedding and vision stages (~50× faster than CPU).
 
 ## Install
 
-**Assumed setup:** A workstation with an NVIDIA GPU (16+ GB VRAM). Tested on WSL2 with NVIDIA driver 546.01+. Native Linux is likely to work with GPU memory allocation tuning (see [docs/services.md](https://github.com/authentic-research-partners/sciwrite-lint/blob/main/docs/services.md#embedding-device)). Not tested on macOS.
+**Assumed setup:** A workstation with an NVIDIA GPU (16+ GB VRAM). Tested on WSL2 with NVIDIA driver 546.01+. Native Linux uses the same GPU-sequencing code path with no extra configuration (see [docs/services.md](https://github.com/authentic-research-partners/sciwrite-lint/blob/main/docs/services.md#embedding-device)), though it isn't actively tested. Not tested on macOS.
 
 Requires [uv](https://docs.astral.sh/uv/getting-started/installation/), a container runtime (podman or docker), CUDA drivers, and [NVIDIA Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html).
 
@@ -207,6 +207,8 @@ sciwrite-lint containers start         # start GROBID + vLLM (needs GPU for vLLM
 sciwrite-lint containers monitor       # live dashboard: service health, VRAM, KV cache
 ```
 
+`init` detects `.tex` files and their `.bib` references and generates a `.sciwrite-lint.toml` config. Review to confirm the right files were detected.
+
 ### WSL2: disable CUDA Sysmem Fallback (recommended)
 
 By default Windows silently spills GPU overflow into system RAM via PCIe (30–100× slower than VRAM); you'll see this as a sudden throughput collapse during the heaviest pipeline stages with no error in the logs. Tell the NVIDIA driver to fail loudly instead:
@@ -217,8 +219,6 @@ Driver 546.01+. After this, WSL2 behaves like native Linux on overflow (CUDA OOM
 
 ![Monitor dashboard](https://github.com/authentic-research-partners/sciwrite-lint/raw/main/docs/monitor.png)
 
-`init` detects `.tex` files and their `.bib` references and generates a `.sciwrite-lint.toml` config. Review to confirm the right files were detected.
-
 **Providing references manually — two drop folders.** sciwrite-lint reads two local directories before going to the open-access waterfall, kept separate because the folder itself signals how much to trust the source:
 
 - `local_pdfs/` — **academic sources**. Accepts `.pdf` (primary) and `.md` summaries you've written yourself. Use this for paywalled papers, for OA papers whose publisher requires a browser (Cloudflare/JS walls, Cell, Springer, some PMC pages), and for anything you'd cite as peer-reviewed evidence.
@@ -280,6 +280,22 @@ sciwrite-lint contributions --paper my_paper    # add contribution axes to SciLi
 sciwrite-lint contributions paper.pdf           # standalone file scoring
 ```
 
+**Supported inputs.**
+
+*The manuscript you check:*
+
+| Format | Notes |
+|---|---|
+| `.tex` | LaTeX — hand-written, or pandoc-generated (e.g. from Markdown); pandoc's heading anchors are unwrapped automatically. |
+| `.pdf` | Parsed via GROBID (container required). |
+
+*Cited sources you optionally supply (drop folders — see [Setup](#setup)):*
+
+| Folder | Accepts | Trust level |
+|---|---|---|
+| `local_pdfs/` | `.pdf`, `.md` (your own summaries) | academic |
+| `local_web/` | `.mhtml` / `.mht`, `.md` | web capture |
+
 `check` runs the full pipeline in one command: text checks → figure analysis → LLM consistency → reference verification via APIs → download and parse cited papers → claim verification → consistency checks on cited papers → bibliography verification → aggregate reliability scores → SciLint Score. An initial run on a 50-reference paper takes up to 30 minutes (dominated by downloads and claim verification); subsequent cached runs complete in minutes.
 
 ### Iterative editing — no `--fresh` needed