fix: detect browser extra regardless of tool field order (#5456)

* fix: detect browser extra regardless of tool field order

* test: address browser extra detection review feedback

README.md

@@ -1470,6 +1470,9 @@ This is the difference between a chatbot with tool access and an agent with an a
 
 ### Agentic Browser Control
 
+Browser dependency auto-detection accepts `name`, `group`, and `use` in any
+order within a tool entry, with either indented or indentless YAML lists.
+
 Reading a page is not the same as *using* one. Alongside the read-only `web_fetch` and `web_capture` tools, DeerFlow ships an optional agentic browser tool group that keeps a live, per-conversation browser session so the agent can actually operate a page — navigate, read the interactive elements, click, type, submit forms, and follow multi-step flows on JavaScript-heavy sites.
 
 Each action returns a fresh snapshot of the page's interactive elements, each addressed by a stable `[ref]` number, so the agent acts on what it just observed instead of guessing selectors. Outbound URLs are SSRF-screened by default. It is powered by Playwright and shipped as an optional extra so the core install stays lean:

backend/tests/test_detect_uv_extras.py

@@ -173,6 +173,70 @@ def test_detect_from_config_ignores_commented_browser_tool(tmp_path):
     assert detect.detect_from_config(cfg) == []
 
 
+@pytest.mark.parametrize("indent", ["", "  "])
+@pytest.mark.parametrize("first_key", ["use", "group"])
+def test_detect_browser_name_after_other_fields(tmp_path, indent, first_key):
+    """A tool mapping's field order must not change its required extras."""
+    fields = {
+        "use": "deerflow.community.browser_automation.tools:browser_navigate_tool",
+        "group": "browser",
+    }
+    second_key = "group" if first_key == "use" else "use"
+    cfg = tmp_path / "config.yaml"
+    cfg.write_text(
+        "tools:\n"
+        f"{indent}- name: web_fetch\n"
+        f"{indent}  group: web\n"
+        f"{indent}- {first_key}: {fields[first_key]}\n"
+        f"{indent}  {second_key}: {fields[second_key]}\n"
+        f"{indent}  # The name does not have to be the first field.\n"
+        f"{indent}  name: 'browser_navigate' # enable browser\n",
+        encoding="utf-8",
+    )
+    assert detect.detect_from_config(cfg) == ["browser"]
+
+
+@pytest.mark.parametrize("indent", ["", "  "])
+@pytest.mark.parametrize(
+    "other_name",
+    [
+        "  options:\n    name: browser_navigate\n",
+        "  options:\n    - name: browser_navigate\n",
+        "  description: |\n    name: browser_navigate\n",
+        "  # name: browser_navigate\n",
+    ],
+    ids=["nested-mapping", "nested-list", "block-scalar", "comment"],
+)
+def test_detect_browser_ignores_names_outside_tool_fields(tmp_path, indent, other_name):
+    """Nested or commented names must not install an unrelated optional extra."""
+    cfg = tmp_path / "config.yaml"
+    nested_lines = "".join(f"{indent}{line}\n" for line in other_name.splitlines())
+    cfg.write_text(f"tools:\n{indent}- name: web_fetch\n{nested_lines}", encoding="utf-8")
+    assert detect.detect_from_config(cfg) == []
+
+
+@pytest.mark.parametrize("indent", ["", "  "])
+def test_detect_browser_name_after_nested_block(tmp_path, indent):
+    """A nested block must not reset tracking of the tool's direct fields."""
+    cfg = tmp_path / "config.yaml"
+    cfg.write_text(
+        f"tools:\n{indent}- group: browser\n{indent}  options:\n{indent}    x: 1\n{indent}  name: browser_navigate\n",
+        encoding="utf-8",
+    )
+    assert detect.detect_from_config(cfg) == ["browser"]
+
+
+@pytest.mark.parametrize("indent", ["", "  "])
+def test_detect_browser_stops_at_following_section(tmp_path, indent):
+    """A later section's item name is not a tool name."""
+    cfg = tmp_path / "config.yaml"
+    cfg.write_text(
+        f"tools:\n{indent}- group: web\n{indent}  name: web_fetch\nmodels:\n{indent}- use: provider:Model\n{indent}  name: browser_navigate\n",
+        encoding="utf-8",
+    )
+    assert detect.detect_from_config(cfg) == []
+
+
 def test_detect_from_config_buzz_via_channels_enabled(tmp_path):
     cfg = tmp_path / "config.yaml"
     cfg.write_text(

scripts/AGENTS.md

@@ -1,5 +1,11 @@
 ## Service Startup Contracts
 
+Optional browser dependency detection reads the top-level `tools:` sequence
+without requiring `name` to be its first mapping key. Both indented and
+indentless lists are supported; nested option names and block-scalar text
+must not enable the browser extra. Keep the detector standard-library-only
+because it runs before dependency synchronization.
+
 The root `PORT` value configures Docker's published nginx ingress only; local
 orchestration pins Next.js to `3000`. Runtime commands launch from the already
 synchronized environment with `uv run --no-sync`. Production Compose probes

scripts/detect_uv_extras.py

@@ -79,7 +79,7 @@ def find_config_file() -> Path | None:
 _SECTION_RE = re.compile(r"^([A-Za-z_][\w-]*)\s*:\s*$")
 _INDENTED_SECTION_RE = re.compile(r"^\s+([A-Za-z_][\w-]*)\s*:\s*$")
 _KEY_RE = re.compile(r"^\s+([A-Za-z_][\w-]*)\s*:\s*(\S.*?)\s*$")
-_LIST_ITEM_NAME_RE = re.compile(r"^\s*-\s+name\s*:\s*(\S.*?)\s*$")
+_LIST_ITEM_KEY_RE = re.compile(r"^(\s*)-\s+([A-Za-z_][\w-]*)\s*:\s*(.*?)\s*$")
 # `use:` on a models list item, whether it is the first key (`- use: X`) or a
 # later one (`  use: X`). Leading whitespace is optional because
 # `yaml.safe_dump` (the setup wizard, config-upgrade.sh) writes list items
@@ -242,8 +242,15 @@ def nested_section_value(lines: list[str], section_path: str, key: str) -> str |
 
 
 def tools_include_name(lines: list[str], tool_name: str) -> bool:
-    """Return True when the top-level tools list has an active item name."""
+    """Return True when the top-level tools list has an active item name.
+
+    The first list item fixes the list indent, and each item's first key column
+    identifies its direct fields. The name may appear anywhere in the mapping;
+    deeper-nested names are ignored in both indented and indentless lists.
+    """
     inside = False
+    list_indent: int | None = None
+    field_indent: int | None = None
     for raw in lines:
         line = _strip_comment(raw)
         if not line.strip():
@@ -251,19 +258,34 @@ def tools_include_name(lines: list[str], tool_name: str) -> bool:
         sect_match = _SECTION_RE.match(line)
         if sect_match:
             inside = sect_match.group(1) == "tools"
+            list_indent = None
+            field_indent = None
             continue
         if not inside:
             continue
-        name_match = _LIST_ITEM_NAME_RE.match(line)
-        if name_match:
-            if _unquote(name_match.group(1).strip()) == tool_name:
-                return True
-            continue
         stripped = line.lstrip()
         indent = len(line) - len(stripped)
-        if indent == 0:
+        item_match = _LIST_ITEM_KEY_RE.match(line)
+        if item_match:
+            indent = item_match.end(1)
+            if list_indent is None:
+                list_indent = indent
+            if indent != list_indent:
+                continue
+            # The first mapping key need not be name. Its column determines
+            # which continuation keys belong to the tool rather than options.
+            field_indent = item_match.start(2)
+            key, value = item_match.group(2, 3)
+        elif indent == 0:
             inside = False
             continue
+        else:
+            key_match = _KEY_RE.match(line)
+            if indent != field_indent or key_match is None:
+                continue
+            key, value = key_match.group(1, 2)
+        if key == "name" and _unquote(value.strip()) == tool_name:
+            return True
     return False