whatmodel 0.0.2__tar.gz → 0.0.3__tar.gz

{whatmodel-0.0.2 → whatmodel-0.0.3}/CHANGELOG.md

@@ -7,6 +7,16 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Fixed
+
+- A name that says nothing about hosting no longer gets a "Runs locally" row. Where a model runs is shown only when the name states it or a lookup confirms it; a model installed in local Ollama is reported as local by that lookup.
+- The `latest` tag Ollama would fill in is no longer shown as a Build row read from the name. It is a note below the table, since a bare name may not be an Ollama name at all.
+- README links work on PyPI.
+
+### Changed
+
+- `whatmodel tokens` prints grid tables.
+
 ### Added
 
 - `whatmodel <name>...` explains Ollama names, Hugging Face repos and GGUF file names in plain English.

{whatmodel-0.0.2 → whatmodel-0.0.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: whatmodel
-Version: 0.0.2
+Version: 0.0.3
 Summary: Explain LLM model names in plain English.
 Project-URL: Homepage, https://github.com/ShekharBhardwaj/whatmodel
 Project-URL: Repository, https://github.com/ShekharBhardwaj/whatmodel
@@ -71,7 +71,7 @@ gemma4:e2b-it-qat
 ├──────────────┼────────────────────────────────────┼──────────────┤
 │ Compression  │ quantization-aware trained         │ name         │
 ├──────────────┼────────────────────────────────────┼──────────────┤
-│ Runs         │ locally                            │ name         │
+│ Runs         │ locally, installed in Ollama here  │ local Ollama │
 ├──────────────┼────────────────────────────────────┼──────────────┤
 │ Parameters   │ 4.6B                               │ local Ollama │
 ├──────────────┼────────────────────────────────────┼──────────────┤
@@ -114,8 +114,6 @@ bartowski/Qwen_Qwen3-30B-A3B-GGUF
 ├──────────────┼────────────────────────────────────────────┼──────────────┤
 │ Format       │ GGUF (for llama.cpp, Ollama and LM Studio) │ name         │
 ├──────────────┼────────────────────────────────────────────┼──────────────┤
-│ Runs         │ locally                                    │ name         │
-├──────────────┼────────────────────────────────────────────┼──────────────┤
 │ Parameters   │ 30.5B                                      │ Hugging Face │
 ├──────────────┼────────────────────────────────────────────┼──────────────┤
 │ Context      │ 32k tokens                                 │ Hugging Face │
@@ -146,8 +144,6 @@ Llama-3-8B-Instruct.Q4_K_M.gguf
 │ Compression │ 4-bit (Q4_K_M, the balanced variant)       │ name   │
 ├─────────────┼────────────────────────────────────────────┼────────┤
 │ Format      │ GGUF (for llama.cpp, Ollama and LM Studio) │ name   │
-├─────────────┼────────────────────────────────────────────┼────────┤
-│ Runs        │ locally                                    │ name   │
 └─────────────┴────────────────────────────────────────────┴────────┘
 ```
 
@@ -193,15 +189,14 @@ qwen-max-0725
 ├───────────┼───────────────────────────────────────┼────────┤
 │ Tier      │ max tier (the most capable option)    │ name   │
 ├───────────┼───────────────────────────────────────┼────────┤
-│ Build     │ latest (the default build)            │ name   │
-├───────────┼───────────────────────────────────────┼────────┤
-│ Runs      │ locally                               │ name   │
-├───────────┼───────────────────────────────────────┼────────┤
 │ Ambiguous │ 0725: dated 07-25 (month-day) [build] │ name   │
 │           │ or released 2025-07 [build]           │        │
 └───────────┴───────────────────────────────────────┴────────┘
+No tag in the name. Pulled with Ollama, it would mean latest.
 ```
 
+The same rule covers what a name leaves out. A name alone cannot show where a model runs or which build you would get, so whatmodel adds a Runs row only when the name says so (a `cloud` tag, or a family that is only offered as a hosted service) or a lookup confirms it.
+
 The same applies to a name such as `org/model` with no tag, which fits both a Hugging Face repo and an Ollama model. Both readings are named unless a lookup settles which one it is.
 
 ### Where the facts come from

{whatmodel-0.0.2 → whatmodel-0.0.3}/README.md

@@ -39,7 +39,7 @@ gemma4:e2b-it-qat
 ├──────────────┼────────────────────────────────────┼──────────────┤
 │ Compression  │ quantization-aware trained         │ name         │
 ├──────────────┼────────────────────────────────────┼──────────────┤
-│ Runs         │ locally                            │ name         │
+│ Runs         │ locally, installed in Ollama here  │ local Ollama │
 ├──────────────┼────────────────────────────────────┼──────────────┤
 │ Parameters   │ 4.6B                               │ local Ollama │
 ├──────────────┼────────────────────────────────────┼──────────────┤
@@ -82,8 +82,6 @@ bartowski/Qwen_Qwen3-30B-A3B-GGUF
 ├──────────────┼────────────────────────────────────────────┼──────────────┤
 │ Format       │ GGUF (for llama.cpp, Ollama and LM Studio) │ name         │
 ├──────────────┼────────────────────────────────────────────┼──────────────┤
-│ Runs         │ locally                                    │ name         │
-├──────────────┼────────────────────────────────────────────┼──────────────┤
 │ Parameters   │ 30.5B                                      │ Hugging Face │
 ├──────────────┼────────────────────────────────────────────┼──────────────┤
 │ Context      │ 32k tokens                                 │ Hugging Face │
@@ -114,8 +112,6 @@ Llama-3-8B-Instruct.Q4_K_M.gguf
 │ Compression │ 4-bit (Q4_K_M, the balanced variant)       │ name   │
 ├─────────────┼────────────────────────────────────────────┼────────┤
 │ Format      │ GGUF (for llama.cpp, Ollama and LM Studio) │ name   │
-├─────────────┼────────────────────────────────────────────┼────────┤
-│ Runs        │ locally                                    │ name   │
 └─────────────┴────────────────────────────────────────────┴────────┘
 ```
 
@@ -161,15 +157,14 @@ qwen-max-0725
 ├───────────┼───────────────────────────────────────┼────────┤
 │ Tier      │ max tier (the most capable option)    │ name   │
 ├───────────┼───────────────────────────────────────┼────────┤
-│ Build     │ latest (the default build)            │ name   │
-├───────────┼───────────────────────────────────────┼────────┤
-│ Runs      │ locally                               │ name   │
-├───────────┼───────────────────────────────────────┼────────┤
 │ Ambiguous │ 0725: dated 07-25 (month-day) [build] │ name   │
 │           │ or released 2025-07 [build]           │        │
 └───────────┴───────────────────────────────────────┴────────┘
+No tag in the name. Pulled with Ollama, it would mean latest.
 ```
 
+The same rule covers what a name leaves out. A name alone cannot show where a model runs or which build you would get, so whatmodel adds a Runs row only when the name says so (a `cloud` tag, or a family that is only offered as a hosted service) or a lookup confirms it.
+
 The same applies to a name such as `org/model` with no tag, which fits both a Hugging Face repo and an Ollama model. Both readings are named unless a lookup settles which one it is.
 
 ### Where the facts come from

{whatmodel-0.0.2 → whatmodel-0.0.3}/tests/test_cli.py

@@ -43,6 +43,7 @@ LOOKED_UP = ModelInfo(
         Fact("context_length", 131072, "128k tokens", "ollama"),
         Fact("tools", True, "yes", "ollama"),
         Fact("license", "Apache License, Version 2.0", "Apache License, Version 2.0", "ollama"),
+        Fact("hosting", "local", "locally, installed in Ollama here", "ollama"),
     ),
     {"ollama": "ok", "huggingface": "not_applicable"},
 )
@@ -67,7 +68,7 @@ def test_layout() -> None:
             "├─────────────┼────────────────────────────────────┼──────────────┤",
             "│ Compression │ quantization-aware trained         │ name         │",
             "├─────────────┼────────────────────────────────────┼──────────────┤",
-            "│ Runs        │ locally                            │ name         │",
+            "│ Runs        │ locally, installed in Ollama here  │ local Ollama │",
             "├─────────────┼────────────────────────────────────┼──────────────┤",
             "│ Compression │ 4-bit (Q4_0)                       │ local Ollama │",
             "├─────────────┼────────────────────────────────────┼──────────────┤",
@@ -86,9 +87,20 @@ def test_every_row_says_where_its_fact_came_from() -> None:
     assert rows[0] == ["Field", "Value", "Source"]
     sources = [row[2] for row in rows[1:]]
     # Name rows come first, looked-up rows after, and the two never interleave.
-    assert sources == ["name"] * 5 + ["local Ollama"] * 4
+    assert sources == ["name"] * 4 + ["local Ollama"] * 5
     assert ["Family", "Gemma 4 (Google)", "name"] in rows
     assert ["Context", "128k tokens", "local Ollama"] in rows
+    # Where it runs is only ever a looked-up fact here, never read off the name.
+    assert ["Runs", "locally, installed in Ollama here", "local Ollama"] in rows
+
+
+def test_assumed_default_tag_is_a_note_not_a_row() -> None:
+    # Whether a bare name is an Ollama name at all is not something it can prove.
+    info = ModelInfo(parse("qwen-max-0725"))
+    assert "Build" not in rows_of("qwen-max-0725")
+    assert "Runs" not in rows_of("qwen-max-0725")
+    assert "Pulled with Ollama, it would mean latest." in text_of(info)
+    assert "Pulled with Ollama" not in text_of(ModelInfo(parse("llama3:latest")))
 
 
 def test_offline_output_has_only_name_rows() -> None:
@@ -134,7 +146,7 @@ def test_lookup_facts_that_repeat_the_name_are_left_out() -> None:
         ("hf.co/bartowski/Qwen_Qwen3-30B-A3B-GGUF:Q4_K_M", "Family", "Qwen 3 (Alibaba)"),
         ("Qwen3-30B-A3B", "Size", "30B parameters | 3B active per token"),
         ("Qwen3-30B-A3B", "Architecture", "mixture of experts"),
-        ("llama3", "Build", "latest (the default build)"),
+        ("llama3:latest", "Build", "latest (the default build)"),
         ("mystery:v2", "Version", "v2"),
         ("some-weird-xyz:abc", "Unknown", "some-weird-xyz, abc"),
     ],
@@ -167,7 +179,7 @@ def test_name_with_two_families() -> None:
     rows = rows_of(name)
     assert rows["Family"] == "Qwen 3.5 (Alibaba) | Claude 4.6 (Anthropic)"
     assert rows["Version"] == "v2"  # the trailing version belongs to neither family
-    assert rows["Runs"] == "locally"  # a hosted family named as a source implies nothing
+    assert "Runs" not in rows  # a hosted family named as a source implies nothing
     cells = cells_of(ModelInfo(parse(name)))
     assert ["Family", "Qwen 3.5 (Alibaba)", "name"] in cells
     assert ["", "Claude 4.6 (Anthropic)", ""] in cells  # each family on its own line

{whatmodel-0.0.2 → whatmodel-0.0.3}/tests/test_dictionary.py

@@ -182,11 +182,17 @@ def test_implied_tokens_resolve(dictionary: Dictionary) -> None:
             assert dictionary.describe(token, category=category) is not None
 
 
-def test_category_default(dictionary: Dictionary) -> None:
-    default = dictionary.categories["hosting"].default
+def test_no_category_assumes_a_value(dictionary: Dictionary) -> None:
+    # A default would present an assumption as a fact read from the name.
+    assert [c.id for c in dictionary.categories.values() if c.default is not None] == []
+
+
+def test_category_default_is_still_supported() -> None:
+    data = _minimal()
+    data["categories"]["host"]["default"] = "somewhere"
+    default = build_dictionary(data).categories["host"].default
     assert default is not None
-    assert default.label == "locally"
-    assert default.value == "local"
+    assert (default.label, default.value) == ("somewhere", "somewhere")
 
 
 # ---------------------------------------------------------------- matching ----

{whatmodel-0.0.2 → whatmodel-0.0.3}/tests/test_lookup.py

@@ -142,6 +142,7 @@ def test_ollama_facts_come_from_show_and_tags() -> None:
         "thinking": (True, "ollama"),
         "license": ("Apache License, Version 2.0", "ollama"),
         "file_size": (4336358185, "ollama"),
+        "hosting": ("local", "ollama"),
     }
     labels = {fact.key: fact.label for fact in info.facts}
     assert labels["context_length"] == "128k tokens"
@@ -203,6 +204,15 @@ def test_remote_model_is_reported_as_hosted() -> None:
     assert [(fact.key, fact.value) for fact in facts] == [("hosting", "cloud")]
 
 
+def test_installed_weights_are_reported_as_local() -> None:
+    facts = ollama_facts({}, {"size": 4336358185})
+    assert ("hosting", "local") in [(fact.key, fact.value) for fact in facts]
+
+
+def test_hosting_is_not_reported_without_evidence() -> None:
+    assert "hosting" not in {fact.key for fact in ollama_facts({"details": {}})}
+
+
 def test_unknown_quantization_level_is_passed_through_unexplained() -> None:
     (fact,) = ollama_facts({"details": {"quantization_level": "nvfp4"}})
     assert (fact.value, fact.label, fact.explanation) == ("nvfp4", "nvfp4", None)

{whatmodel-0.0.2 → whatmodel-0.0.3}/tests/test_parser.py

@@ -116,7 +116,6 @@ def test_default_tag_is_implied_for_a_bare_ollama_name() -> None:
     assert parsed.build == "latest"
     assert [(item.reading.category, item.reading.value) for item in parsed.implied] == [
         ("build", "latest"),
-        ("hosting", "local"),
     ]
     assert "no tag" in parsed.implied[0].reason
 
@@ -336,15 +335,18 @@ def test_unambiguous_date_stamp() -> None:
 # ----------------------------------------------------------------- implied ----
 
 
-def test_hosting_defaults_to_local_and_says_why() -> None:
-    parsed = parse("qwen3.5:4b")
-    assert parsed.hosting == "local"
-    (item,) = parsed.implied
-    assert item.reading.label == "locally"
-    assert item.reason == "the name has no hosting token"
+@pytest.mark.parametrize(
+    "name",
+    ["qwen3.5:4b", "qwen-max-0725", "deepseek/deepseek-v4.1-flash", "llama3", "model.gguf"],
+)
+def test_hosting_is_not_assumed_when_the_name_is_silent(name: str) -> None:
+    parsed = parse(name)
+    assert parsed.hosting is None
+    assert parsed.values("hosting") == ()
+    assert all(item.reading.category != "hosting" for item in parsed.implied)
 
 
-def test_cloud_tag_overrides_the_default() -> None:
+def test_cloud_tag_says_where_the_model_runs() -> None:
     parsed = parse("gpt-oss:120b-cloud")
     assert parsed.hosting == "cloud"
     assert parsed.implied == ()
@@ -358,7 +360,7 @@ def test_token_can_imply_another_category() -> None:
     assert implied["architecture"].reason == "implied by 'A3B' in the name"
 
 
-def test_hosted_family_implies_cloud_instead_of_the_local_default() -> None:
+def test_hosted_family_implies_cloud() -> None:
     parsed = parse("claude-3-5-sonnet-20241022")
     assert parsed.hosting == "cloud"
     assert "local" not in parsed.values("hosting")
@@ -367,7 +369,7 @@ def test_hosted_family_implies_cloud_instead_of_the_local_default() -> None:
 def test_rival_families_imply_nothing() -> None:
     parsed = parse("Jackrong/Qwen3.5-9B-Claude-4.6-Opus-Reasoning-Distilled-v2")
     assert parsed.values("family") == ("qwen", "claude")
-    assert parsed.hosting == "local"
+    assert parsed.hosting is None
     assert parsed.unknown == ()
 
 

{whatmodel-0.0.2 → whatmodel-0.0.3}/whatmodel/lookup.py

@@ -292,6 +292,9 @@ def ollama_facts(show: Mapping[str, Any], summary: Mapping[str, Any] | None = No
     remote = _text(show.get("remote_host")) or _text(summary.get("remote_host"))
     if remote is not None:
         facts.append(Fact("hosting", "cloud", f"remotely, on {remote}", source))
+    elif size is not None:
+        # The weights are on this machine's disk, so this is observed, not assumed.
+        facts.append(Fact("hosting", "local", "locally, installed in Ollama here", source))
     return facts
 
 

{whatmodel-0.0.2 → whatmodel-0.0.3}/whatmodel/parser.py

@@ -26,6 +26,8 @@ from whatmodel.dictionary import (
 )
 
 #: The tag Ollama uses when a name has none.
+#: Why the default tag is implied. It only holds if the name is used with Ollama.
+DEFAULT_TAG_REASON = "the name has no tag, so Ollama uses its default"
 DEFAULT_TAG = "latest"
 
 #: Hosts that mean "this is a Hugging Face repo". Used for routing, not meaning.
@@ -411,7 +413,7 @@ def _implied(
     # 2. The tag Ollama fills in when the name has none.
     if structure.tag_is_default and structure.kinds == (KIND_OLLAMA,):
         implied.extend(
-            Implied(reading, "the name has no tag, so Ollama uses its default")
+            Implied(reading, DEFAULT_TAG_REASON)
             for reading in dictionary.match(normalize(DEFAULT_TAG), ignore_context=True)
         )
 

{whatmodel-0.0.2 → whatmodel-0.0.3}/whatmodel/render.py

@@ -26,7 +26,13 @@ from whatmodel.lookup import (
     ModelInfo,
     format_count,
 )
-from whatmodel.parser import KIND_FILE, KIND_HUGGINGFACE, KIND_OLLAMA, ParsedName
+from whatmodel.parser import (
+    DEFAULT_TAG_REASON,
+    KIND_FILE,
+    KIND_HUGGINGFACE,
+    KIND_OLLAMA,
+    ParsedName,
+)
 
 #: Row order and headings for looked-up facts.
 FACT_HEADINGS = {
@@ -126,7 +132,9 @@ def name_rows(parsed: ParsedName, dictionary: Dictionary | None = None) -> list[
         for position, token in enumerate(parsed.tokens)
         if token.reading is not None and position not in attached
     ]
-    readings += [item.reading for item in parsed.implied]
+    # The default tag is left out: it holds only if the name is used with Ollama,
+    # which the name cannot establish. render_info says so in a note instead.
+    readings += [item.reading for item in parsed.implied if item.reason != DEFAULT_TAG_REASON]
     for category in dictionary.categories.values():
         if category.id == "family":
             if family_entries:
@@ -259,6 +267,8 @@ def render_info(info: ModelInfo, *, verbose: bool = False) -> RenderableType:
     kinds = _kinds_line(info)
     if kinds:
         captions.append(f"Read as {kinds}.")
+    if parsed.tag_is_default and parsed.kinds == (KIND_OLLAMA,):
+        captions.append(f"No tag in the name. Pulled with Ollama, it would mean {parsed.tag}.")
     note = lookup_note(info)
     if note and not looked_up and STATUS_OK not in info.lookups.values():
         captions.append(f"Nothing looked up: {note}.")

{whatmodel-0.0.2 → whatmodel-0.0.3}/whatmodel/tokens.yaml

@@ -93,11 +93,9 @@ categories:
     description: Which snapshot or release stage of the model this is.
   hosting:
     heading: Runs
-    description: Where the model actually runs.
-    default:
-      label: locally
-      value: local
-      explanation: Nothing in the name marks it as hosted, so the weights download and run on your own machine.
+    # No default on purpose: a name that says nothing about hosting proves
+    # nothing about where the model runs. Only a token or a lookup may say so.
+    description: Where the model runs, when the name or a lookup establishes it.
 
 tokens:
   # ---------------------------------------------------------------- host ----