nbd-jev3 0.1.0__tar.gz

nbd_jev3-0.1.0/PKG-INFO

@@ -0,0 +1,7 @@
+Metadata-Version: 2.4
+Name: nbd-jev3
+Version: 0.1.0
+Summary: Jev classifiers for text_classifier and Parseer pages
+Author-email: Shaked Degani <shaked@webz.io>
+Requires-Python: >=3.12
+Requires-Dist: requests

nbd_jev3-0.1.0/README.md

@@ -0,0 +1,16 @@
+# nbd-jev3
+
+Standalone Jev client. Requires Python 3.12 and `JEV_API_KEY`. The model is `typesafe/jev-1.13`.
+
+```python
+from nbd_jev import classify_text_classifier, classify_parseer
+
+classify_text_classifier(text)  # text_classifier labels, or "not sure"
+classify_parseer(text)          # Parseer page labels, or "not sure"
+```
+
+A choice below 0.5 confidence is `not sure`. A failed request is also `not sure`.
+
+```bash
+python -m nbd_jev.check
+```

nbd_jev3-0.1.0/nbd_jev/__init__.py

@@ -0,0 +1,8 @@
+"""Jev classifiers. Requires JEV_API_KEY. The model is pinned in this package."""
+
+from nbd_jev.classify import classify_parseer, classify_text_classifier
+
+__all__ = [
+    "classify_text_classifier",
+    "classify_parseer",
+]

nbd_jev3-0.1.0/nbd_jev/check.py

@@ -0,0 +1,21 @@
+"""Offline checks. Run: python -m nbd_jev.check"""
+
+from nbd_jev.classify import (
+    NOT_SURE,
+    PARSEER_CRITERIA,
+    TEXT_CLASSIFIER_CRITERIA,
+    _resolve,
+)
+
+assert "wiki page" not in TEXT_CLASSIFIER_CRITERIA
+assert "news article" in TEXT_CLASSIFIER_CRITERIA
+assert set(PARSEER_CRITERIA) == {
+    "news article", "not sure", "other", "product page",
+    "advertisement", "job listing", "forum post", "wiki page",
+}
+assert _resolve("news article", 0.5, TEXT_CLASSIFIER_CRITERIA) == "news article"
+assert _resolve("news article", 0.49, TEXT_CLASSIFIER_CRITERIA) == NOT_SURE
+assert _resolve("made up", 0.9, TEXT_CLASSIFIER_CRITERIA) == NOT_SURE
+assert _resolve(None, None, TEXT_CLASSIFIER_CRITERIA) == NOT_SURE
+
+print("ok")

nbd_jev3-0.1.0/nbd_jev/classify.py

@@ -0,0 +1,101 @@
+"""Jev decision calls: text_classifier labels and Parseer labels."""
+
+from __future__ import annotations
+
+import os
+
+import requests
+
+MODEL = "typesafe/jev-1.13"
+DECISIONS_URL = "https://openrouter.ai/api/alpha/decisions"
+TIMEOUT = 30
+CONFIDENCE_THRESHOLD = 0.5
+NOT_SURE = "not sure"
+
+TEXT_CLASSIFIER_CRITERIA = {
+    "job listing": "A vacancy with role, requirements, and how to apply.",
+    "ecommerce": "A store homepage or shop section, not one product and not a search-results page.",
+    "product page": "One product for sale, with price and a buy action.",
+    "forum post": "A single user message on a forum.",
+    "porn": "Sexually explicit adult content or an age-gated adult video page.",
+    "forum discussion": "A thread made of an opening post and replies.",
+    "podcast or audio program": "An audio episode or podcast with listen duration and subscribe links.",
+    "horoscope": "Daily or weekly zodiac predictions and lucky numbers.",
+    "recipe": "Cooking instructions with ingredients and steps.",
+    "advertisement": "Sponsored or promotional marketing copy for a product or service.",
+    "real estate listing": "A property for sale or rent with price, beds, and tour details.",
+    "ecommerce listing": "A search or category page listing many products for sale.",
+    "pet adoption listing": "An adoptable pet profile with adoption fee and application.",
+    "crossword puzzle": "A crossword or word puzzle with clues and a grid.",
+    "gambling": "Sports betting odds, wagers, or a sportsbook page.",
+    "casino content": "Online casino games such as slots, roulette, or blackjack.",
+    "sports statistics": "League tables, match stats, or player scoring records.",
+    "glossary": "A dictionary or glossary entry defining a term.",
+    "weather": "A forecast with temperature, conditions, and multi-day outlook.",
+    "news article": "A reported story about an event, with who, what, and when.",
+    NOT_SURE: "Too little text, or none of the other categories fit.",
+}
+
+PARSEER_CRITERIA = {
+    "news article": "A reported story about an event, with who, what, and when.",
+    NOT_SURE: "Too little text, or none of the other categories fit.",
+    "other": "None of the other categories fit, and there is enough text to tell.",
+    "product page": "One product for sale, with price and a buy action.",
+    "advertisement": "Sponsored or promotional marketing copy for a product or service.",
+    "job listing": "A vacancy with role, requirements, and how to apply.",
+    "forum post": "A single user message on a forum.",
+    "wiki page": "An encyclopedia-style article with sections, see-also links, and references.",
+}
+
+PARSEER_INSTRUCTIONS = (
+    "Which category best describes this web page text? "
+    "For newsroom or press-release content, be slightly more forgiving: "
+    "if there is a reasonable chance it is editorial news, prefer 'news article' over 'advertisement'."
+)
+
+def classify_text_classifier(text: str) -> str:
+    """text_classifier labels. Returns a category, or 'not sure' when the call fails."""
+    return _classify(text, "Which category best describes this web page text?", TEXT_CLASSIFIER_CRITERIA)
+
+
+def classify_parseer(text: str) -> str:
+    """Parseer page labels. Returns a category, or 'not sure' when the call fails."""
+    return _classify(text, PARSEER_INSTRUCTIONS, PARSEER_CRITERIA)
+
+
+def _classify(text: str, instructions: str, criteria: dict[str, str]) -> str:
+    cleaned = (text or "").strip()
+    if not cleaned:
+        return NOT_SURE
+    try:
+        body = _post({
+            "model": MODEL,
+            "state": {"text": cleaned},
+            "questions": {
+                "category": {"type": "choice", "instructions": instructions, "criteria": criteria},
+            },
+        })
+        answer = body["answers"]["category"]
+        return _resolve(answer.get("choice"), answer.get("confidence"), criteria)
+    except (requests.RequestException, KeyError, ValueError):
+        return NOT_SURE
+
+
+def _resolve(choice: str | None, confidence: float | None, criteria: dict[str, str]) -> str:
+    if choice not in criteria or confidence is None or confidence < CONFIDENCE_THRESHOLD:
+        return NOT_SURE
+    return choice
+
+
+def _post(payload: dict) -> dict:
+    key = os.environ.get("JEV_API_KEY")
+    if not key:
+        raise ValueError("JEV_API_KEY is not set")
+    response = requests.post(
+        DECISIONS_URL,
+        headers={"Authorization": f"Bearer {key}", "Content-Type": "application/json"},
+        json=payload,
+        timeout=TIMEOUT,
+    )
+    response.raise_for_status()
+    return response.json()

nbd_jev3-0.1.0/nbd_jev3.egg-info/PKG-INFO

@@ -0,0 +1,7 @@
+Metadata-Version: 2.4
+Name: nbd-jev3
+Version: 0.1.0
+Summary: Jev classifiers for text_classifier and Parseer pages
+Author-email: Shaked Degani <shaked@webz.io>
+Requires-Python: >=3.12
+Requires-Dist: requests

nbd_jev3-0.1.0/nbd_jev3.egg-info/SOURCES.txt

@@ -0,0 +1,10 @@
+README.md
+pyproject.toml
+nbd_jev/__init__.py
+nbd_jev/check.py
+nbd_jev/classify.py
+nbd_jev3.egg-info/PKG-INFO
+nbd_jev3.egg-info/SOURCES.txt
+nbd_jev3.egg-info/dependency_links.txt
+nbd_jev3.egg-info/requires.txt
+nbd_jev3.egg-info/top_level.txt

nbd_jev3-0.1.0/nbd_jev3.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

nbd_jev3-0.1.0/nbd_jev3.egg-info/requires.txt

@@ -0,0 +1 @@
+requests

nbd_jev3-0.1.0/nbd_jev3.egg-info/top_level.txt

@@ -0,0 +1 @@
+nbd_jev

nbd_jev3-0.1.0/pyproject.toml

@@ -0,0 +1,14 @@
+[build-system]
+requires = ["setuptools>=68"]
+build-backend = "setuptools.build_meta"
+
+[project]
+name = "nbd-jev3"
+version = "0.1.0"
+description = "Jev classifiers for text_classifier and Parseer pages"
+requires-python = ">=3.12"
+authors = [{ name = "Shaked Degani", email = "shaked@webz.io" }]
+dependencies = ["requests"]
+
+[tool.setuptools.packages.find]
+include = ["nbd_jev*"]

nbd_jev3-0.1.0/setup.cfg

@@ -0,0 +1,4 @@
+[egg_info]
+tag_build = 
+tag_date = 0
+