Make worker lint-clean at vgi-lint-check 0.26.0 (strict profile, 100/100)

Add the strict-profile metadata the 0.26.0 linter requires on every object,
and make all examples execute cleanly against the worker (execute-on-by-default).

- Add vgi_nlp/meta.py with object_tags()/source_url() helpers (mirrors the
  vgi-units reference): vgi.title (distinct from the machine name), narrative
  vgi.doc_llm and vgi.doc_md (distinct Markdown), vgi.keywords, vgi.source_url.
- scalars.py: per-object tags on all 11 scalar functions; rewrite every example
  to be self-contained (string literals, no backend tables) so VGI901 examples
  bind/execute; switch the explicit-model examples to the installed
  en_core_web_sm.
- tables.py: per-object tags on entities/tokens/sentences/noun_chunks; rename
  vgi.columns_md -> vgi.result_columns_md (VGI405); self-contained per-function
  examples; add vgi.executable_examples (VGI509/VGI906) with inline one-row
  input relations.
- nlp_worker.py: rename vgi.description_llm/_md -> vgi.doc_llm/_md; add
  vgi.title + vgi.keywords on the catalog; add vgi.title/keywords/source_url/
  example_queries plus bare-key domain/category/topic (VGI123) on the schema;
  use the LicenseRef- SPDX form for the source-available license (VGI013).

Lint: 100/100, 0 findings at --fail-on info. pytest, ruff, mypy, pydoclint green.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: rustyconover

nlp_worker.py

@@ -70,8 +70,24 @@
 )
 
 _SCHEMA_DESCRIPTION_MD = (
-    "Classical NLP functions: language ID, sentiment, cleaning scalars, and "
-    "entity/token/sentence/noun-chunk table functions."
+    "# main\n\n"
+    "Classical NLP functions exposed to SQL.\n\n"
+    "- **Scalars:** language ID/confidence, sentiment score/label, lemmatize, "
+    "strip stop-words, normalize.\n"
+    "- **Table functions:** entities, tokens, sentences, noun_chunks "
+    "(one text row in, N rows out, with an optional `id :=` passthrough)."
+)
+
+# Representative, catalog-qualified example queries for the schema (VGI506).
+# All are self-contained so they bind/execute against the worker.
+_SCHEMA_EXAMPLE_QUERIES = (
+    "SELECT nlp.main.detect_lang('Bonjour tout le monde');\n"
+    "SELECT nlp.main.sentiment('I absolutely love this product!');\n"
+    "SELECT nlp.main.sentiment_label('This was a terrible experience');\n"
+    "SELECT nlp.main.lemmatize('The cats were running', 'en');\n"
+    "SELECT nlp.main.normalize('  Café   DELUXE  ');\n"
+    "SELECT * FROM nlp.main.entities("
+    "(SELECT 1 AS id, 'Apple is based in California.' AS body), id := 'id', lang := 'en');"
 )
 
 _NLP_CATALOG = Catalog(
@@ -80,11 +96,17 @@
     comment="Classical NLP (spaCy + fastText + VADER): language ID, sentiment, NER, tokenization for SQL.",
     source_url="https://github.com/Query-farm/vgi-nlp",
     tags={
-        "vgi.description_llm": _CATALOG_DESCRIPTION_LLM,
-        "vgi.description_md": _CATALOG_DESCRIPTION_MD,
+        "vgi.title": "Classical NLP for SQL",
+        "vgi.keywords": (
+            "nlp, natural language processing, language detection, sentiment analysis, "
+            "named entity recognition, ner, tokenization, lemmatize, stop words, "
+            "noun chunks, spacy, fasttext, vader, text enrichment"
+        ),
+        "vgi.doc_llm": _CATALOG_DESCRIPTION_LLM,
+        "vgi.doc_md": _CATALOG_DESCRIPTION_MD,
         "vgi.author": "Query.Farm",
         "vgi.copyright": "Copyright 2026 Query Farm LLC - https://query.farm",
-        "vgi.license": "Query Farm Source-Available License, Version 1.0",
+        "vgi.license": "LicenseRef-QueryFarm-Source-Available-1.0",
         "vgi.support_contact": "https://github.com/Query-farm/vgi-nlp/issues",
         "vgi.support_policy_url": "https://github.com/Query-farm/vgi-nlp/blob/main/README.md",
     },
@@ -93,8 +115,20 @@
             name="main",
             comment="Classical NLP: language ID, sentiment, NER, tokenization for SQL",
             tags={
-                "vgi.description_llm": _SCHEMA_DESCRIPTION_LLM,
-                "vgi.description_md": _SCHEMA_DESCRIPTION_MD,
+                "vgi.title": "NLP Functions (main)",
+                "vgi.keywords": (
+                    "nlp, language detection, sentiment, ner, entities, tokens, "
+                    "sentences, noun chunks, lemmatize, strip stopwords, normalize, "
+                    "spacy, fasttext, vader"
+                ),
+                # VGI123 classifying tags use BARE keys (not vgi.-namespaced).
+                "domain": "text-analytics",
+                "category": "natural-language-processing",
+                "topic": "language-detection-sentiment-ner",
+                "vgi.source_url": "https://github.com/Query-farm/vgi-nlp/blob/main/nlp_worker.py",
+                "vgi.doc_llm": _SCHEMA_DESCRIPTION_LLM,
+                "vgi.doc_md": _SCHEMA_DESCRIPTION_MD,
+                "vgi.example_queries": _SCHEMA_EXAMPLE_QUERIES,
             },
             functions=[*SCALAR_FUNCTIONS, *TABLE_FUNCTIONS],
         ),

vgi_nlp/meta.py

@@ -0,0 +1,49 @@
+"""Shared helpers for the per-object discovery/description metadata that the
+``vgi-lint`` strict profile expects on **every** function and table.
+
+Each function/table surfaces these in its ``Meta.tags``:
+
+- ``vgi.title`` (VGI124)        -- human-friendly display name (must differ from
+  the machine name; add an extra word so VGI125 stays quiet)
+- ``vgi.doc_llm`` (VGI112)      -- Markdown narrative aimed at LLMs/agents
+- ``vgi.doc_md`` (VGI113)       -- Markdown narrative for human docs (distinct
+  content from ``doc_llm``)
+- ``vgi.keywords`` (VGI126)     -- comma-separated search terms/synonyms
+- ``vgi.source_url`` (VGI128)   -- link to the implementing source file
+
+``source_url(file)`` builds the canonical GitHub blob URL so every object points
+at exactly where it is implemented.
+"""  # noqa: D205
+
+from __future__ import annotations
+
+#: Base GitHub blob URL for source files in this repo (pinned to ``main``).
+SOURCE_BASE = "https://github.com/Query-farm/vgi-nlp/blob/main"
+
+
+def source_url(relative_path: str) -> str:
+    """Build the implementation ``vgi.source_url`` for a repo-relative file.
+
+    e.g. ``source_url("vgi_nlp/scalars.py")``.
+    """
+    return f"{SOURCE_BASE}/{relative_path}"
+
+
+def object_tags(
+    title: str,
+    doc_llm: str,
+    doc_md: str,
+    keywords: str,
+    relative_path: str,
+) -> dict[str, str]:
+    """Build the five standard per-object discovery/description tags.
+
+    ``relative_path`` is the implementing file relative to the repo root.
+    """
+    return {
+        "vgi.title": title,
+        "vgi.doc_llm": doc_llm,
+        "vgi.doc_md": doc_md,
+        "vgi.keywords": keywords,
+        "vgi.source_url": source_url(relative_path),
+    }