deps: move tree-sitter-language-pack to 1.x, and read Dart type arguments from the call

CHANGELOG.md

@@ -84,6 +84,14 @@ uses [Semantic Versioning](https://semver.org/).
   `--update` keeps it, so existing flat docs stay flat. The GitHub Pages
   viewer, the web app and the MCP tools (`doc_path` in
   `processing_order.json`) follow the layout.
+- **tree-sitter-language-pack 1.x.** The pin moves from `>=0.8.0,<0.9.0` to
+  `>=1.23.1,<2.0.0`, which brings grammars 0.8 does not ship (Lean, among
+  others). 1.x downloads a grammar on first use and caches it under
+  `tree_sitter_language_pack.cache_dir()`, so the first run of the Dart, Go,
+  JavaScript/TypeScript-in-Vue, Vue and Zig analyzers needs network access;
+  the Docker image fetches them at build time. The 1.x Dart grammar parses
+  `Foo<T>(x)` as a call instead of a relational expression, and the Dart
+  analyzer now takes the type-argument edges from that call too.
 
 ## [2.0.0] - 2026-09-18
 

codewiki/src/be/dependency_analyzer/analyzers/dart.py

@@ -762,6 +762,17 @@ def _process_chain(self, target, head, selectors, local_types, lift_to):
             return
         caller, owner = target.caller, target.owner
         line = head.start_point[0] + 1
+        # `Foo<T>(x)` / `foo<T>(x)`: the type arguments sit in the call's
+        # argument_part (grammars that misparse it as a relational expression
+        # are handled by _handle_generic_invocation).
+        for token_kind, _, token_node in tokens:
+            if token_kind != "call":
+                continue
+            type_args = _first_child(token_node, "type_arguments")
+            if type_args is not None:
+                for ref in _descendants(type_args):
+                    if ref.type == "type_identifier":
+                        self._add_type_edge(caller, _text(ref), line)
         name = _text(head)
         if head.type == "identifier" and name in self.import_prefixes and tokens[0][0] == "member":
             name, tokens = tokens[0][1] or "", tokens[1:]

docker/Dockerfile

@@ -18,6 +18,10 @@ COPY requirements.txt .
 # Install Python dependencies
 RUN pip install --no-cache-dir -r requirements.txt
 
+# tree-sitter-language-pack 1.x downloads each grammar on first use. Fetch the
+# ones the analyzers load at build time, so the image runs offline.
+RUN python -c "import tree_sitter_language_pack as t; t.download(['dart', 'go', 'javascript', 'typescript', 'vue', 'zig'])"
+
 # Copy application code
 COPY codewiki ./codewiki
 COPY img ./img

guides/development.md

@@ -123,6 +123,8 @@ Analyzers are standalone classes, one file per language. Follow
    and `requirements.txt`, or take the grammar from
    `tree-sitter-language-pack` (already a dependency) with
    `get_parser("<lang>")`, as `analyzers/dart.py` and `analyzers/go.py` do.
+   The pack (1.x) downloads a grammar on first use, so also add the name to
+   the `t.download([...])` list in `docker/Dockerfile`.
 2. **Analyzer.** Create `analyzers/<lang>.py` with a
    `TreeSitter<Lang>Analyzer` class that extracts components and
    relationships, plus a module-level `analyze_<lang>_file(...)` function.

pyproject.toml

@@ -29,7 +29,7 @@ dependencies = [
     "GitPython>=3.1.40",
     "Jinja2>=3.1.6",
     "tree-sitter>=0.25",
-    "tree-sitter-language-pack>=0.8.0,<0.9.0",
+    "tree-sitter-language-pack>=1.23.1,<2.0.0",
     "tree-sitter-python>=0.23.6",
     "tree-sitter-java>=0.23.5",
     "tree-sitter-javascript>=0.21.4",

requirements.txt

@@ -148,7 +148,7 @@ tree-sitter-embedded-template==0.23.2
 tree-sitter-java==0.23.5
 tree-sitter-javascript==0.21.4
 tree-sitter-kotlin==1.1.0
-tree-sitter-language-pack==0.8.0
+tree-sitter-language-pack==1.23.1
 tree-sitter-python==0.23.6
 tree-sitter-ruby==0.23.1
 tree-sitter-rust==0.23.2