extract-cli 0.1.1__tar.gz → 0.1.2__tar.gz

{extract_cli-0.1.1 → extract_cli-0.1.2}/ARCHITECTURE.md

@@ -8,11 +8,14 @@ map.
 ```
 load_source(path)                       extension/content sniff → reader
   ├─ .md/.txt  → utf-8 decode
+  ├─ .html     → stdlib html.parser reader (also auto-detected inside .txt)
   ├─ .docx     → python-docx (if [docx]) else stdlib zipfile/XML reader
   └─ .pdf      → pypdf (if [pdf]) else stdlib zlib + text-operator reader
         │
         ▼  (raw_bytes, text, format, warnings)
 build_extraction(text, raw, fmt, src)   the DETERMINISTIC tier (always on)
+  │  field extractors run on a whitespace-FLATTENED copy (so values that wrap
+  │  across a line are matched whole); clause detection keeps the original text
   ├─ extract_parties          "between X and Y", with role parentheticals
   ├─ extract_dates            effective / expiration, ISO-normalized
   ├─ extract_term             length / auto_renew / notice_period_days

{extract_cli-0.1.1 → extract_cli-0.1.2}/CHANGELOG.md

@@ -6,6 +6,39 @@ to [Semantic Versioning](https://semver.org/). Per the suite convention
 (see [`docs/INTEROP.md`](docs/INTEROP.md)), **backward-incompatible changes to
 the output schema require a major version bump**; new optional fields are minor.
 
+## [0.1.2] - 2026-05-21
+
+More real-world hardening, driven by testing against five additional contracts
+(SEC EDGAR consulting/MSA, lease, and Visteon services agreements; Common Paper
+and Perigon Cloud Service Agreements).
+
+### Added
+- **HTML input** (`.html`/`.htm`, and HTML auto-detected inside `.txt` such as
+  SEC EDGAR full submissions). Stdlib `html.parser`-based reader strips
+  script/style, frames block elements so heading detection still works, and
+  unescapes entities. `document.format` enum gains `html` (backward-compatible
+  widening). This turns the large class of HTML contracts (SEC exhibits, web
+  ToS) from garbage into structured output.
+
+### Fixed
+- **Field extraction now runs on whitespace-flattened text**, so values that
+  wrap across a line break are matched whole — e.g. governing law
+  `the laws of the Province\nof Ontario` now yields `Province of Ontario`, and
+  line-wrapped party names/defined terms are captured.
+- **Party extraction** (continues issue #2): names are trimmed of trailing
+  descriptors (`, a Delaware corporation`, `doing business as …`,
+  `having its offices at …`, `as of …`), and each party must begin with a
+  capital so an `and` *inside* a party's own description no longer splits the
+  parties (`…V6E 3S7 and doing business as …` → real parties recovered).
+
+### Known limitations (documented, not bugs)
+- The stdlib PDF reader cannot decode PDFs that use embedded subset fonts with
+  hex-encoded glyph strings (common in professionally-typeset PDFs); these
+  degrade gracefully to a low-signal warning. Install the `[pdf]` extra (pypdf)
+  for them — verified to recover full text and clause structure.
+- Two-line `ARTICLE N` / title headings (number on one line, title on the next)
+  are not yet detected.
+
 ## [0.1.1] - 2026-05-21
 
 Real-world hardening, driven by testing against a SEC EDGAR employment
@@ -82,5 +115,6 @@ Initial release — the open-loop front door of the contract-ops CLI suite.
   intentionally *not* governed by the output schema (the schema describes the
   full default output).
 
+[0.1.2]: https://github.com/DrBaher/extract-cli/releases/tag/v0.1.2
 [0.1.1]: https://github.com/DrBaher/extract-cli/releases/tag/v0.1.1
 [0.1.0]: https://github.com/DrBaher/extract-cli/releases/tag/v0.1.0

{extract_cli-0.1.1 → extract_cli-0.1.2}/PKG-INFO

@@ -1,7 +1,7 @@
 Metadata-Version: 2.4
 Name: extract-cli
-Version: 0.1.1
-Summary: Open-loop front door of the contract-ops CLI suite: ingest any contract (.md/.txt/.docx/.pdf) and emit structured JSON.
+Version: 0.1.2
+Summary: Open-loop front door of the contract-ops CLI suite: ingest any contract (.md/.txt/.html/.docx/.pdf) and emit structured JSON.
 Project-URL: Homepage, https://cli.drbaher.com/
 Project-URL: Repository, https://github.com/DrBaher/extract-cli
 Project-URL: Suite interop, https://github.com/DrBaher/extract-cli/blob/main/docs/INTEROP.md
@@ -63,8 +63,8 @@ ingest (extract) → review → diff → convert → sign
 
 ## What it does
 
-Give it a contract in **`.md` / `.txt`** (native), **`.docx`**, or **`.pdf`**,
-and it returns structured JSON: the parties, dates, term, governing law, a
+Give it a contract in **`.md` / `.txt` / `.html`** (native), **`.docx`**, or
+**`.pdf`**, and it returns structured JSON: the parties, dates, term, governing law, a
 **clause map** normalized onto the suite's canonical clause vocabulary, a
 defined-term inventory, and a headline value. Every field carries a
 `confidence` and a `source` so downstream tools **verify, don't trust**.
@@ -75,14 +75,15 @@ daemon, no network in the default path.
 ## Install
 
 ```bash
-pip install extract-cli                 # core: .md/.txt + best-effort .docx/.pdf
+pip install extract-cli                 # core: .md/.txt/.html + best-effort .docx/.pdf
 pip install "extract-cli[docx]"         # higher-fidelity .docx (python-docx)
 pip install "extract-cli[pdf]"          # higher-fidelity .pdf (pypdf)
 pip install "extract-cli[docx,pdf]"     # both
 ```
 
-The core has **zero runtime dependencies** and is fully functional on `.md`/`.txt`
-with no extras. `.docx` and `.pdf` work out of the box via stdlib readers; the
+The core has **zero runtime dependencies** and is fully functional on
+`.md`/`.txt`/`.html` with no extras (HTML is also auto-detected when it hides
+inside a `.txt`, e.g. SEC EDGAR filings). `.docx` and `.pdf` work out of the box via stdlib readers; the
 `[docx]`/`[pdf]` extras improve fidelity on complex documents (see
 [ARCHITECTURE.md](ARCHITECTURE.md)).
 

{extract_cli-0.1.1 → extract_cli-0.1.2}/README.md

@@ -25,8 +25,8 @@ ingest (extract) → review → diff → convert → sign
 
 ## What it does
 
-Give it a contract in **`.md` / `.txt`** (native), **`.docx`**, or **`.pdf`**,
-and it returns structured JSON: the parties, dates, term, governing law, a
+Give it a contract in **`.md` / `.txt` / `.html`** (native), **`.docx`**, or
+**`.pdf`**, and it returns structured JSON: the parties, dates, term, governing law, a
 **clause map** normalized onto the suite's canonical clause vocabulary, a
 defined-term inventory, and a headline value. Every field carries a
 `confidence` and a `source` so downstream tools **verify, don't trust**.
@@ -37,14 +37,15 @@ daemon, no network in the default path.
 ## Install
 
 ```bash
-pip install extract-cli                 # core: .md/.txt + best-effort .docx/.pdf
+pip install extract-cli                 # core: .md/.txt/.html + best-effort .docx/.pdf
 pip install "extract-cli[docx]"         # higher-fidelity .docx (python-docx)
 pip install "extract-cli[pdf]"          # higher-fidelity .pdf (pypdf)
 pip install "extract-cli[docx,pdf]"     # both
 ```
 
-The core has **zero runtime dependencies** and is fully functional on `.md`/`.txt`
-with no extras. `.docx` and `.pdf` work out of the box via stdlib readers; the
+The core has **zero runtime dependencies** and is fully functional on
+`.md`/`.txt`/`.html` with no extras (HTML is also auto-detected when it hides
+inside a `.txt`, e.g. SEC EDGAR filings). `.docx` and `.pdf` work out of the box via stdlib readers; the
 `[docx]`/`[pdf]` extras improve fidelity on complex documents (see
 [ARCHITECTURE.md](ARCHITECTURE.md)).
 

{extract_cli-0.1.1 → extract_cli-0.1.2}/docs/spec/extract-output.schema.json

@@ -69,7 +69,8 @@
             "markdown",
             "text",
             "docx",
-            "pdf"
+            "pdf",
+            "html"
           ]
         },
         "sha256": {

{extract_cli-0.1.1 → extract_cli-0.1.2}/extract_cli.py

@@ -4,8 +4,8 @@
 The suite is a contract lifecycle (store -> draft -> review -> diff -> convert
 -> sign) that, until now, only handled documents it authored from its own
 templates. `extract-cli` is "passport control": it ingests ANY document --
-yours or a counterparty's foreign paper -- in .md/.txt (natively), .docx, or
-.pdf, and emits a structured JSON representation that the rest of the suite
+yours or a counterparty's foreign paper -- in .md/.txt/.html (natively), .docx,
+or .pdf, and emits a structured JSON representation that the rest of the suite
 (nda-review-cli, compare-cli, contract-vault) consumes.
 
 Two extraction tiers:
@@ -32,6 +32,7 @@ from __future__ import annotations
 import argparse
 import datetime as _dt
 import hashlib
+import html.parser
 import importlib.util
 import json
 import os
@@ -42,11 +43,11 @@ import urllib.request
 from pathlib import Path
 from typing import Any, Dict, List, Optional, Tuple
 
-__version__ = "0.1.1"
+__version__ = "0.1.2"
 
 # Bumped independently of the package version when the *extraction logic*
 # changes in a way downstream consumers should notice. Embedded in `_meta`.
-EXTRACTOR_VERSION = "0.1.1"
+EXTRACTOR_VERSION = "0.1.2"
 
 # JSON Schema version of the output contract (docs/spec/extract-output.schema.json).
 SCHEMA_VERSION = 1
@@ -492,10 +493,17 @@ _EXPIRE_RE = re.compile(
     re.IGNORECASE,
 )
 
+# Each party must start with a capital letter (optionally "the X"), a quote, or
+# a paren. This is case-sensitive on purpose (no global IGNORECASE -- only the
+# keywords are): it lets the engine skip an "and" that sits INSIDE a party's own
+# description ("...V6E 3S7 and doing business as ...", where the right side
+# starts lowercase) and find the real "and" before the second named entity.
+_PARTY_START = r"(?:(?:[Tt]he|its)\s+)?[A-Z\"“(]"
 _PARTY_BLOCK_RE = re.compile(
-    r"\b(?:by\s+and\s+between|between)\s+(.{2,200}?)\s+\band\b\s+(.{2,200}?)"
-    r"(?=[\.;\n]|\bwhereas\b|\beffective\b|\bdated\b|\bhaving\b|\bwith\s+offices\b|$)",
-    re.IGNORECASE | re.DOTALL,
+    r"(?i:\b(?:by\s+and\s+between|between)\s+)"
+    r"(" + _PARTY_START + r"[^\n]{1,200}?)\s+and\s+"
+    r"(" + _PARTY_START + r"[^\n]{1,200}?)"
+    r"(?=[\.;\n]|(?i:\bwhereas\b|\beffective\b|\bdated\b|\bas\s+of\b|\bwitnesseth\b)|$)",
 )
 _ROLE_PAREN_RE = re.compile(
     r"\(\s*(?:the\s+)?[\"“]?([^\"”()]+?)[\"”]?\s*\)"
@@ -604,8 +612,40 @@ def _date_field(match: Optional["re.Match[str]"]) -> JSON:
     return _date_field_from_str(match.group(1), 0.85)
 
 
+# Trailing descriptors that follow a party's actual name and should be dropped
+# ("Acme Corp., a Delaware corporation", "... doing business as Foo", "... as of
+# March 1", "... having its offices at ..."). Each is matched and everything from
+# it onward is cut.
+_PARTY_CUT_MARKERS: Tuple[str, ...] = (
+    r",\s+an?\s+\w",                                  # ", a Delaware ..." / ", an Ohio ..."
+    r"\s+doing\s+business\s+as\b",
+    r"\s+d/?b/?a\b",
+    r"\s+f/?k/?a\b",
+    r"\s+a[n]?\s+\w+\s+(?:corporation|company|partnership|limited)\b",
+    r"\s+having\b",
+    r"\s+with\s+(?:its\s+)?(?:offices|principal|a\s)\b",
+    r"\s+with\s+offices\b",
+    r"\s+located\b",
+    r"\s+organized\b",
+    r"\s+incorporated\b",
+    r"\s+whose\b",
+    r"\s+(?:as\s+of|dated|effective)\b",
+)
+
+
+def _clean_party_name(s: str) -> str:
+    """Trim a captured party name down to the entity name, dropping trailing
+    descriptors ('a Delaware corporation', 'd/b/a ...', 'as of ...')."""
+    s = re.sub(r"\s+", " ", s).strip().strip(",").strip()
+    for pat in _PARTY_CUT_MARKERS:
+        m = re.search(pat, s, re.IGNORECASE)
+        if m:
+            s = s[: m.start()].strip().strip(",").strip()
+    return s.strip("\"“”").strip()
+
+
 def _split_name_role(s: str) -> Tuple[str, Optional[str]]:
-    s = s.strip().strip(",").strip()
+    s = re.sub(r"\s+", " ", s).strip().strip(",").strip()
     role: Optional[str] = None
     m = _ROLE_PAREN_RE.search(s)
     if m:
@@ -614,9 +654,7 @@ def _split_name_role(s: str) -> Tuple[str, Optional[str]]:
         if len(candidate) <= 40 and candidate.lower() not in ("a", "an", "the"):
             role = candidate
         s = (s[: m.start()] + s[m.end():]).strip().rstrip(",").strip()
-    s = s.strip("\"“”").strip()
-    s = re.sub(r"\s+", " ", s)
-    return s, role
+    return _clean_party_name(s), role
 
 
 def extract_parties(text: str) -> List[JSON]:
@@ -625,9 +663,6 @@ def extract_parties(text: str) -> List[JSON]:
         return []
     out: List[JSON] = []
     for raw in (m.group(1), m.group(2)):
-        # Party names can wrap across lines ("...(the \"Disclosing\nParty\")");
-        # collapse whitespace rather than truncating at the first newline.
-        raw = re.sub(r"\s+", " ", raw).strip()
         name, role = _split_name_role(raw)
         if not name or len(name) < 2 or len(name) > 120:
             continue
@@ -750,21 +785,91 @@ def extract_title(text: str, path: Optional[Path], fmt: str) -> Optional[str]:
 # ---------------------------------------------------------------------------
 
 
+def _looks_like_html(head: str) -> bool:
+    """Heuristic: does this text look like HTML? Catches HTML masquerading as
+    .txt (e.g. SEC EDGAR full submissions wrap HTML exhibits in a .txt)."""
+    low = head.lower()
+    if "<!doctype html" in low or "<html" in low or "<body" in low:
+        return True
+    return len(re.findall(r"</?(?:p|div|table|tr|td|span|br|h[1-6]|font|b|i)\b", low)) >= 6
+
+
 def _detect_format(path: Path, raw: bytes) -> str:
     ext = path.suffix.lower()
-    if ext in (".md", ".markdown"):
-        return "markdown"
-    if ext == ".txt":
-        return "text"
+    if ext in (".htm", ".html", ".xhtml"):
+        return "html"
     if ext == ".docx":
         return "docx"
     if ext == ".pdf":
         return "pdf"
     if raw[:4] == b"%PDF":
         return "pdf"
-    if raw[:2] == b"PK":
+    if raw[:2] == b"PK" and ext not in (".md", ".markdown", ".txt"):
         return "docx"
-    return "text"
+    base = "markdown" if ext in (".md", ".markdown") else "text"
+    # Content sniff: HTML hiding inside a .txt/.md (or extensionless) file.
+    if _looks_like_html(raw[:4096].decode("utf-8", "replace")):
+        return "html"
+    return base
+
+
+class _HTMLTextExtractor(html.parser.HTMLParser):
+    """Stdlib HTML -> text: drops script/style, frames block elements with blank
+    lines (so clause-heading detection still works), and unescapes entities."""
+
+    _SKIP = {"script", "style", "head", "title", "meta", "link", "noscript"}
+    _BLOCK = {
+        "p", "div", "br", "tr", "li", "h1", "h2", "h3", "h4", "h5", "h6",
+        "section", "article", "table", "ul", "ol", "blockquote", "pre", "hr",
+        "thead", "tbody", "header", "footer", "main",
+    }
+
+    def __init__(self) -> None:
+        super().__init__(convert_charrefs=True)
+        self._parts: List[str] = []
+        self._skip = 0
+
+    def handle_starttag(self, tag: str, attrs: Any) -> None:
+        if tag in self._SKIP:
+            self._skip += 1
+        elif tag in self._BLOCK:
+            self._parts.append("\n")
+
+    def handle_endtag(self, tag: str) -> None:
+        if tag in self._SKIP and self._skip > 0:
+            self._skip -= 1
+        elif tag in self._BLOCK:
+            self._parts.append("\n")
+
+    def handle_data(self, data: str) -> None:
+        if self._skip == 0:
+            self._parts.append(data)
+
+    def get_text(self) -> str:
+        # Strip each line; collapse runs of blank lines to a single blank line
+        # (gives ALL-CAPS / numbered headings their blank-line frame).
+        lines = [re.sub(r"[ \t]+", " ", ln).strip() for ln in "".join(self._parts).split("\n")]
+        out: List[str] = []
+        blank = False
+        for ln in lines:
+            if ln:
+                out.append(ln)
+                blank = False
+            elif not blank:
+                out.append("")
+                blank = True
+        return "\n".join(out).strip()
+
+
+def _read_html(raw_text: str) -> str:
+    parser = _HTMLTextExtractor()
+    try:
+        parser.feed(raw_text)
+        parser.close()
+    except Exception:
+        # Never crash on malformed markup; fall back to a crude tag strip.
+        return re.sub(r"<[^>]+>", " ", raw_text)
+    return parser.get_text()
 
 
 def _read_docx(path: Path, raw: bytes, prefer_optional: bool = True) -> Tuple[str, List[str]]:
@@ -986,6 +1091,8 @@ def load_source(path: Path, prefer_optional: bool = True) -> Tuple[bytes, str, s
     warnings: List[str] = []
     if fmt in ("markdown", "text"):
         text = raw.decode("utf-8", "replace")
+    elif fmt == "html":
+        text = _read_html(raw.decode("utf-8", "replace"))
     elif fmt == "docx":
         text, w = _read_docx(path, raw, prefer_optional)
         warnings += w
@@ -1011,6 +1118,13 @@ def build_extraction(text: str, raw: bytes, fmt: str,
                      source_path: Optional[str]) -> JSON:
     """Run the deterministic tier and assemble the output contract object."""
     sha = hashlib.sha256(raw).hexdigest()
+    # Field extractors (parties, dates, governing law, term, value, defined
+    # terms) run on a whitespace-flattened copy so values that wrap across a
+    # line break in the source -- "...laws of the Province\nof Ontario", a party
+    # name split mid-line -- are matched whole. Clause detection and the title
+    # keep the original text, which depends on line structure.
+    flat = re.sub(r"[ \t\r\f\v]*\n[ \t\r\f\v]*", " ", text)
+    flat = re.sub(r"[ \t]+", " ", flat)
     return {
         "document": {
             "title": extract_title(text, Path(source_path) if source_path else None, fmt),
@@ -1018,13 +1132,13 @@ def build_extraction(text: str, raw: bytes, fmt: str,
             "sha256": sha,
             "source_path": source_path,
         },
-        "parties": extract_parties(text),
-        "dates": extract_dates(text),
-        "term": extract_term(text),
-        "governing_law": extract_governing_law(text),
+        "parties": extract_parties(flat),
+        "dates": extract_dates(flat),
+        "term": extract_term(flat),
+        "governing_law": extract_governing_law(flat),
         "clauses": extract_clauses(text),
-        "defined_terms": extract_defined_terms(text),
-        "value": extract_value(text),
+        "defined_terms": extract_defined_terms(flat),
+        "value": extract_value(flat),
         "_meta": {
             "extractor_version": EXTRACTOR_VERSION,
             "tiers_used": ["deterministic"],
@@ -1336,7 +1450,7 @@ def output_schema() -> JSON:
                 "required": ["title", "format", "sha256", "source_path"],
                 "properties": {
                     "title": {"type": ["string", "null"]},
-                    "format": {"enum": ["markdown", "text", "docx", "pdf"]},
+                    "format": {"enum": ["markdown", "text", "docx", "pdf", "html"]},
                     "sha256": {"type": "string", "pattern": "^[0-9a-f]{64}$"},
                     "source_path": {"type": ["string", "null"]},
                 },
@@ -1687,7 +1801,7 @@ def _add_common_output_flags(p: argparse.ArgumentParser) -> None:
 def build_parser() -> argparse.ArgumentParser:
     parser = argparse.ArgumentParser(
         prog="extract",
-        description="Ingest any contract (.md/.txt/.docx/.pdf) and emit structured "
+        description="Ingest any contract (.md/.txt/.html/.docx/.pdf) and emit structured "
                     "JSON for the contract-ops CLI suite. See docs/INTEROP.md.",
     )
     parser.add_argument("-V", "--version", action="version",
@@ -1721,7 +1835,7 @@ def build_parser() -> argparse.ArgumentParser:
 
 
 def _build_extract_args(p: argparse.ArgumentParser) -> None:
-    p.add_argument("path", help="Path to the document (.md/.txt/.docx/.pdf).")
+    p.add_argument("path", help="Path to the document (.md/.txt/.html/.docx/.pdf).")
     p.add_argument("--llm", action="store_true",
                    help="Opt-in LLM enrichment of fuzzy fields (renewal, obligations). "
                         "Off by default; the deterministic core is fully useful without it.")

{extract_cli-0.1.1 → extract_cli-0.1.2}/pyproject.toml

@@ -4,8 +4,8 @@ build-backend = "hatchling.build"
 
 [project]
 name = "extract-cli"
-version = "0.1.1"
-description = "Open-loop front door of the contract-ops CLI suite: ingest any contract (.md/.txt/.docx/.pdf) and emit structured JSON."
+version = "0.1.2"
+description = "Open-loop front door of the contract-ops CLI suite: ingest any contract (.md/.txt/.html/.docx/.pdf) and emit structured JSON."
 readme = "README.md"
 requires-python = ">=3.9"
 license = { text = "MIT" }

{extract_cli-0.1.1 → extract_cli-0.1.2}/tests/_make_goldens.py

@@ -20,7 +20,8 @@ from tests._fixtures_build import ensure_binary_fixtures  # noqa: E402
 FIXTURES = Path(__file__).resolve().parent / "fixtures"
 
 DOCS = ["nda_h2.md", "services_bold.txt", "lease_allcaps.txt",
-        "employment_docx.docx", "license_pdf.pdf", "scanned.pdf"]
+        "employment_docx.docx", "license_pdf.pdf", "services_html.html",
+        "scanned.pdf"]
 
 
 def golden_for(name: str) -> dict:

{extract_cli-0.1.1 → extract_cli-0.1.2}/tests/conftest.py

@@ -26,6 +26,7 @@ CORPUS: Tuple[Tuple[str, str, str], ...] = (
     ("lease_allcaps.txt", "all-caps", "text"),
     ("employment_docx.docx", "bold-numbered", "docx"),
     ("license_pdf.pdf", "all-caps", "pdf"),
+    ("services_html.html", "numbered", "html"),
 )
 
 

{extract_cli-0.1.1 → extract_cli-0.1.2}/tests/fixtures/employment_docx.docx.expected.json

@@ -138,7 +138,7 @@
     "source": "deterministic"
   },
   "_meta": {
-    "extractor_version": "0.1.1",
+    "extractor_version": "0.1.2",
     "tiers_used": [
       "deterministic"
     ],

{extract_cli-0.1.1 → extract_cli-0.1.2}/tests/fixtures/lease_allcaps.txt.expected.json

@@ -133,7 +133,7 @@
     "source": "deterministic"
   },
   "_meta": {
-    "extractor_version": "0.1.1",
+    "extractor_version": "0.1.2",
     "tiers_used": [
       "deterministic"
     ],

{extract_cli-0.1.1 → extract_cli-0.1.2}/tests/fixtures/license_pdf.pdf.expected.json

@@ -133,7 +133,7 @@
     "source": "deterministic"
   },
   "_meta": {
-    "extractor_version": "0.1.1",
+    "extractor_version": "0.1.2",
     "tiers_used": [
       "deterministic"
     ],

{extract_cli-0.1.1 → extract_cli-0.1.2}/tests/fixtures/nda_h2.md.expected.json

@@ -121,6 +121,11 @@
       "confidence": 0.6,
       "source": "deterministic"
     },
+    {
+      "term": "Disclosing Party",
+      "confidence": 0.6,
+      "source": "deterministic"
+    },
     {
       "term": "Receiving Party",
       "confidence": 0.6,
@@ -138,7 +143,7 @@
     "source": "none"
   },
   "_meta": {
-    "extractor_version": "0.1.1",
+    "extractor_version": "0.1.2",
     "tiers_used": [
       "deterministic"
     ],

{extract_cli-0.1.1 → extract_cli-0.1.2}/tests/fixtures/scanned.pdf.expected.json

@@ -48,7 +48,7 @@
     "source": "none"
   },
   "_meta": {
-    "extractor_version": "0.1.1",
+    "extractor_version": "0.1.2",
     "tiers_used": [
       "deterministic"
     ],

{extract_cli-0.1.1 → extract_cli-0.1.2}/tests/fixtures/services_bold.txt.expected.json

@@ -133,7 +133,7 @@
     "source": "deterministic"
   },
   "_meta": {
-    "extractor_version": "0.1.1",
+    "extractor_version": "0.1.2",
     "tiers_used": [
       "deterministic"
     ],

extract_cli-0.1.2/tests/fixtures/services_html.html

@@ -0,0 +1,35 @@
+<!DOCTYPE html>
+<html>
+<head>
+  <title>Exhibit 10.1</title>
+  <style>body { font-family: serif; } .hidden { display:none; }</style>
+  <script>var x = "(this should never appear in output)";</script>
+</head>
+<body>
+  <p align="center"><b>MASTER SERVICES AGREEMENT</b></p>
+
+  <p>This Master Services Agreement (the &ldquo;Agreement&rdquo;) is entered
+  into as of March 15, 2023 (the &quot;Effective Date&quot;), by and between
+  Initrode&nbsp;Systems,&nbsp;Inc., a Delaware corporation (&ldquo;Provider&rdquo;),
+  and Hooli&nbsp;LLC (&ldquo;Customer&rdquo;).</p>
+
+  <p>1. Services</p>
+  <p>Provider shall perform the services described in each Statement of Work.</p>
+
+  <p>2. Fees and Payment</p>
+  <p>Customer shall pay Provider the fees set forth in the applicable Statement
+  of Work, not to exceed $500,000 in the aggregate.</p>
+
+  <p>3. Term and Termination</p>
+  <p>The initial term of this Agreement is two (2) years. Either party may
+  terminate upon sixty (60) days&rsquo; written notice. This Agreement shall
+  automatically renew for successive one-year terms.</p>
+
+  <p>4. Confidentiality</p>
+  <p>Each party shall protect the other&rsquo;s &ldquo;Confidential
+  Information&rdquo; using reasonable care.</p>
+
+  <p>5. Governing Law</p>
+  <p>This Agreement shall be governed by the laws of the State of California.</p>
+</body>
+</html>

extract_cli-0.1.2/tests/fixtures/services_html.html.expected.json

@@ -0,0 +1,157 @@
+{
+  "document": {
+    "title": "MASTER SERVICES AGREEMENT",
+    "format": "html",
+    "sha256": "088b40f13135e6b5d8f8548b162d657f10725d348388c7c3a416d11d7fc65300",
+    "source_path": "services_html.html"
+  },
+  "parties": [
+    {
+      "name": "Initrode Systems, Inc.",
+      "confidence": 0.9,
+      "source": "deterministic",
+      "role": "Provider"
+    },
+    {
+      "name": "Hooli LLC",
+      "confidence": 0.9,
+      "source": "deterministic",
+      "role": "Customer"
+    }
+  ],
+  "dates": {
+    "effective": {
+      "value": "2023-03-15",
+      "confidence": 0.9,
+      "source": "deterministic"
+    },
+    "expiration": {
+      "value": null,
+      "confidence": 0.0,
+      "source": "none"
+    }
+  },
+  "term": {
+    "length": {
+      "value": "2 years",
+      "confidence": 0.7,
+      "source": "deterministic"
+    },
+    "auto_renew": {
+      "value": true,
+      "confidence": 0.65,
+      "source": "deterministic"
+    },
+    "notice_period_days": {
+      "value": 60,
+      "confidence": 0.7,
+      "source": "deterministic"
+    }
+  },
+  "governing_law": {
+    "value": "State of California",
+    "confidence": 0.85,
+    "source": "deterministic"
+  },
+  "clauses": [
+    {
+      "canonical_title": "Services",
+      "detected_title": "1. Services",
+      "tier": "numbered",
+      "span": {
+        "start": 242,
+        "end": 329
+      },
+      "confidence": 0.6,
+      "source": "deterministic",
+      "mapped": false
+    },
+    {
+      "canonical_title": "Payment",
+      "detected_title": "2. Fees and Payment",
+      "tier": "numbered",
+      "span": {
+        "start": 329,
+        "end": 476
+      },
+      "confidence": 0.8,
+      "source": "deterministic",
+      "mapped": true
+    },
+    {
+      "canonical_title": "Termination",
+      "detected_title": "3. Term and Termination",
+      "tier": "numbered",
+      "span": {
+        "start": 476,
+        "end": 692
+      },
+      "confidence": 0.8,
+      "source": "deterministic",
+      "mapped": true
+    },
+    {
+      "canonical_title": "Confidentiality",
+      "detected_title": "4. Confidentiality",
+      "tier": "numbered",
+      "span": {
+        "start": 692,
+        "end": 800
+      },
+      "confidence": 0.8,
+      "source": "deterministic",
+      "mapped": true
+    },
+    {
+      "canonical_title": "Governing Law",
+      "detected_title": "5. Governing Law",
+      "tier": "numbered",
+      "span": {
+        "start": 800,
+        "end": 890
+      },
+      "confidence": 0.8,
+      "source": "deterministic",
+      "mapped": true
+    }
+  ],
+  "defined_terms": [
+    {
+      "term": "Agreement",
+      "confidence": 0.6,
+      "source": "deterministic"
+    },
+    {
+      "term": "Effective Date",
+      "confidence": 0.6,
+      "source": "deterministic"
+    },
+    {
+      "term": "Provider",
+      "confidence": 0.6,
+      "source": "deterministic"
+    },
+    {
+      "term": "Customer",
+      "confidence": 0.6,
+      "source": "deterministic"
+    },
+    {
+      "term": "Confidential Information",
+      "confidence": 0.6,
+      "source": "deterministic"
+    }
+  ],
+  "value": {
+    "value": "$500,000",
+    "confidence": 0.6,
+    "source": "deterministic"
+  },
+  "_meta": {
+    "extractor_version": "0.1.2",
+    "tiers_used": [
+      "deterministic"
+    ],
+    "llm_used": false
+  }
+}

{extract_cli-0.1.1 → extract_cli-0.1.2}/tests/test_deterministic.py

@@ -12,8 +12,8 @@ def test_parties_between_simple() -> None:
     assert all(0.0 <= p["confidence"] <= 1.0 for p in parties)
 
 
-def test_parties_with_roles_and_linebreak() -> None:
-    text = ('by and between Acme Corp. (the "Disclosing\nParty") and '
+def test_parties_with_roles() -> None:
+    text = ('by and between Acme Corp. (the "Disclosing Party") and '
             'Beta LLC (the "Receiving Party"), dated March 1, 2024.')
     parties = ex.extract_parties(text)
     assert parties[0]["name"] == "Acme Corp."
@@ -22,6 +22,30 @@ def test_parties_with_roles_and_linebreak() -> None:
     assert parties[1]["role"] == "Receiving Party"
 
 
+def test_parties_linebreak_handled_by_build() -> None:
+    # build_extraction flattens whitespace, so a party/role that wraps across a
+    # line is matched whole.
+    text = ('This Agreement is made by and between Acme Corp. (the "Disclosing\n'
+            'Party") and Beta LLC (the "Receiving Party").')
+    r = ex.build_extraction(text, text.encode("utf-8"), "text", "x.txt")
+    assert [p["name"] for p in r["parties"]] == ["Acme Corp.", "Beta LLC"]
+    assert r["parties"][0]["role"] == "Disclosing Party"
+
+
+def test_parties_skip_and_inside_description() -> None:
+    # An "and" inside a party's own description must not split the parties.
+    text = ("between Blade Ventures Inc., a Nevada corporation having offices at "
+            "1 Main St and doing business as Foo (\"Client\"), and KPMG LP")
+    parties = ex.extract_parties(text)
+    assert [p["name"] for p in parties] == ["Blade Ventures Inc.", "KPMG LP"]
+
+
+def test_party_name_descriptors_trimmed() -> None:
+    assert ex._clean_party_name("Visteon Corporation, a Delaware corporation") == "Visteon Corporation"
+    assert ex._clean_party_name("Foo Inc. doing business as Bar") == "Foo Inc."
+    assert ex._clean_party_name("Baz LLC having its principal office at X") == "Baz LLC"
+
+
 def test_parties_none() -> None:
     assert ex.extract_parties("There are no parties named here.") == []
 
@@ -80,6 +104,15 @@ def test_governing_law_stops_before_trailing_clause() -> None:
     assert out["value"] == "State of Delaware"
 
 
+def test_governing_law_linebreak_handled_by_build() -> None:
+    # A jurisdiction that wraps a line ("...the Province\nof Ontario") is
+    # matched whole because build_extraction flattens whitespace first.
+    text = ("This Agreement shall be governed by the laws of the Province\n"
+            "of Ontario and the federal laws of Canada.")
+    r = ex.build_extraction(text, text.encode("utf-8"), "text", "x.txt")
+    assert r["governing_law"]["value"] == "Province of Ontario"
+
+
 def test_governing_law_missing() -> None:
     assert ex.extract_governing_law("nothing about law")["source"] == "none"
 

{extract_cli-0.1.1 → extract_cli-0.1.2}/tests/test_misc.py

@@ -142,6 +142,32 @@ def test_pdf_unescape() -> None:
     assert ex._pdf_unescape(r"\101\102") == "AB"  # octal escapes
 
 
+def test_html_extraction() -> None:
+    raw, text, fmt, _w = ex.load_source(FIXTURES / "services_html.html")
+    assert fmt == "html"
+    # script/style content is dropped; entities are unescaped.
+    assert "this should never appear" not in text
+    result = ex.build_extraction(text, raw, fmt, "services_html.html")
+    assert result["document"]["format"] == "html"
+    assert [p["name"] for p in result["parties"]] == ["Initrode Systems, Inc.", "Hooli LLC"]
+    assert result["governing_law"]["value"] == "State of California"
+    assert result["dates"]["effective"]["value"] == "2023-03-15"
+    canon = {c["canonical_title"] for c in result["clauses"]}
+    assert {"Payment", "Termination", "Confidentiality", "Governing Law"} <= canon
+
+
+def test_html_detected_by_content_sniff(tmp_path: Any) -> None:
+    # HTML masquerading as .txt (e.g. a SEC EDGAR full submission) is sniffed.
+    p = tmp_path / "exhibit.txt"
+    p.write_text("<html><body><p>between A Co and B Co</p></body></html>")
+    _raw, _text, fmt, _w = ex.load_source(p)
+    assert fmt == "html"
+
+
+def test_html_malformed_does_not_crash() -> None:
+    assert ex._read_html("<p>unclosed <b>bold <div>text") is not None
+
+
 def test_pdf_text_only_inside_bt_et() -> None:
     # Strings outside BT/ET (font/signature/metadata stream bytes that happen to
     # contain parentheses) must be ignored; only text objects yield text.