From 4b55a41ece6c7a01627f1fdc7a7ab89079189a32 Mon Sep 17 00:00:00 2001
From: endergebnis <endergebnis@hanime.zip>
Date: Tue, 8 Sep 2026 22:48:45 +0200
Subject: [PATCH] Add --exclude, detect/inspect_bundle tools, Spring/test
 entry-point detection

- Global `--exclude <glob>` (repeatable) on search symbols/text and
  find dead/dead-symbols, with built-in defaults (vendor/, node_modules/,
  *.min.*, dist/, build/). Minified/vendored assets stay indexed for import
  edges but are filtered from query results (bootstrap.min.css noise: 90% of
  dead symbols -> gone).
- `detect` CLI / `codebase_detect` MCP tool: probe a directory's languages and
  stack from file extensions and manifest files before indexing.
- `inspect bundle` CLI / `inspect_bundle` MCP tool: symbol source + direct
  callers + direct importers in one call (3 round-trips -> 1).
- find dead-symbols: skip CSS files and structural kinds (package, heading*)
  as non-callable; extend framework-wired detection with JUnit/Spring
  annotations (@Test, @Service, @Component, ...) and pytest/JUnit test-class
  naming (Test*, test_*, *Test, *Tests), plus `test_` methods reported as
  `member` kind.
- README: correct CSS statement, document new tools/limits.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
---
 Cargo.lock        |   1 +
 Cargo.toml        |   1 +
 README.md         |  39 ++++++----
 src/cli.rs        |  13 ++++
 src/config.rs     | 101 ++++++++++++++++++++++++
 src/main.rs       |  20 +++--
 src/mcp/mod.rs    |  87 +++++++++++++++++++--
 src/ops.rs        | 190 +++++++++++++++++++++++++++++++++++++++++++---
 src/search/mod.rs |  41 ++++++++--
 9 files changed, 452 insertions(+), 41 deletions(-)

diff --git a/Cargo.lock b/Cargo.lock
index c92d91a..4f25651 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -399,6 +399,7 @@ version = "0.1.0"
 dependencies = [
  "anyhow",
  "clap",
+ "globset",
  "ignore",
  "rayon",
  "regex",
diff --git a/Cargo.toml b/Cargo.toml
index 2b24cdb..509db8b 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -8,6 +8,7 @@ license = "MIT"
 clap = { version = "4", features = ["derive"] }
 rusqlite = { version = "0.31", features = ["bundled"] }
 ignore = "0.4"
+globset = "0.4"
 rayon = "1"
 xxhash-rust = { version = "0.8", features = ["xxh3"] }
 serde = { version = "1", features = ["derive"] }
diff --git a/README.md b/README.md
index 2790d0e..5d32792 100644
--- a/README.md
+++ b/README.md
@@ -8,8 +8,10 @@ context tokens.
 Language coverage (via ctags): Rust, Java, Python, C/C++, TypeScript,
 JavaScript, HTML, C#. HTML import edges are extracted directly (not via ctags),
 and `<link>`/`<script>`/`<img>` references to local `.css`/`.js`/image files are
-tracked as imports. CSS files are discovered and import-tracked, but ctags has
-no CSS parser — `search symbols` returns nothing for `.css` files.
+tracked as imports. universal-ctags (5.9.0+) *does* index CSS selectors as
+symbols, so `search symbols` can return them — `find dead-symbols` skips CSS
+files and structural kinds (selectors/`package`/`heading*` are not callable
+symbols) and the default `--exclude` set filters minified/generated assets.
 
 `rustytokens` is an independent, clean-room implementation. It does not share
 code with any AGPL-licensed project.
@@ -42,22 +44,27 @@ cargo build --release
 | `index create <path>` | Index a project from scratch |
 | `index update <path>` | Re-index only changed files (content-hash diff) |
 | `index file <file>` | Re-index a single file |
-| `search symbols <query> [--include-source] [--limit n]` | Symbol search (exact > prefix > full-text > substring) |
-| `search text <query> [--limit n]` | Full-text search with line numbers, read live from disk |
+| `detect [path]` | Detect languages/stack at a path before indexing (no index needed) |
+| `search symbols <query> [--include-source] [--limit n] [--exclude glob]` | Symbol search (exact > prefix > full-text > substring) |
+| `search text <query> [--limit n] [--exclude glob]` | Full-text search with line numbers, read live from disk |
 | `inspect symbol <id\|name>` | Source of a single symbol (id or exact name) |
+| `inspect bundle <id\|name>` | Symbol source + direct callers + direct importers in one call |
 | `inspect outline <file>` | Symbol outline of a file |
 | `inspect file <file> [--lines a-b]` | Read a file or line range |
 | `inspect tree [--depth n]` | Indexed file tree |
 | `inspect blast-radius <file> [--depth n]` | Files transitively affected by a change |
 | `find importers <file>` | Who imports this file |
-| `find dead` | Files nobody imports (entry points excluded) |
-| `find dead-symbols` | Symbols never referenced (framework-wired entry points excluded) |
+| `find dead [--exclude glob]` | Files nobody imports (entry points excluded) |
+| `find dead-symbols [--exclude glob]` | Symbols never referenced (framework-wired entry points excluded) |
 | `find callers <id\|name>` | Symbols that call the given symbol |
 | `find references <identifier>` | Code lines referencing an identifier (definitions/strings/comments excluded) |
 | `inspect cycles` | Circular imports |
 | `serve` | MCP server over stdio |
 | `stats` | Index statistics |
 
+`--exclude` is a repeatable glob; it adds to built-in defaults (`vendor/`,
+`node_modules/`, `*.min.*`, `dist/`, `build/`).
+
 Global options: `--project <dir>` (default `.`), `--format json|table`
 (default `json`), `--compact`.
 
@@ -150,11 +157,11 @@ Notes:
 ## Scope and known gaps
 
 `rustytokens` covers the same *core* retrieval surface for Java/Python/Rust/
-C/C++/HTML as `toktoken`, but intentionally implements a smaller tool set: 17
+C/C++/HTML as `toktoken`, but intentionally implements a smaller tool set: 19
 tools vs `toktoken`'s 27. The following `toktoken` tools have **no equivalent
 here yet** (deliberately, not an oversight):
 
-- `inspect_bundle`, `inspect_dependencies`, `inspect_hierarchy` (graph views)
+- `inspect_dependencies`, `inspect_hierarchy` (graph views)
 - `search_similar`, `search_cooccurrence` (similarity search)
 - GitHub indexing and centrality ranking
 
@@ -163,17 +170,21 @@ for the "roughly cover Java/Python/Rust/C/C++/HTML" goal.
 
 Known limits that are by design, not bugs:
 
-- `.css` files are discovered and import-tracked, but `ctags` has no CSS
-  parser — `search symbols` returns nothing for `.css`.
+- universal-ctags (5.9.0+) indexes CSS selectors as symbols, so `search
+  symbols` may return them. `find dead-symbols` skips CSS files and structural
+  kinds (`package`, `heading*`) — selectors/declarations are not callable
+  symbols — and the default `--exclude` set drops minified assets.
 - `find dead` over-counts for C/C++ header-only entry points (documented
   conservatism); treat results as candidates, not verdicts.
 - `find callers` / `find references` are word-boundary heuristics: they match a
   symbol's name in text, so same-name symbols in unrelated files can produce
   false positives (`find callers` narrows this via the import graph).
-- `find dead-symbols` treats pytest `test_*` functions, `#[test]`-wired Rust
-  tests and framework-decorated symbols (FastAPI/Flask routes, Celery tasks,
-  click/typer commands, …) as live entry points. A pure name-scan cannot see
-  these call sites, so this is a documented heuristic.
+- `find dead-symbols` treats pytest `test_*` functions, `Test*`/`test_*` test
+  classes, `#[test]`-wired Rust tests, JUnit/Spring annotations (`@Test`,
+  `@Service`, `@Component`, …) and other framework-decorated symbols
+  (FastAPI/Flask routes, Celery tasks, click/typer commands, …) as live entry
+  points. A pure name-scan cannot see these call sites, so this is a documented
+  heuristic.
 - `find references` filters the definition line and occurrences inside strings
   and comments; multi-line docstrings/block comments are approximated
   line-by-line, so a symbol name inside a long string block may still appear.
diff --git a/src/cli.rs b/src/cli.rs
index c7761d1..a444d83 100644
--- a/src/cli.rs
+++ b/src/cli.rs
@@ -23,6 +23,11 @@ pub struct Cli {
     #[arg(long, global = true)]
     pub compact: bool,
 
+    /// Exclude files matching a glob from query results (repeatable; adds to
+    /// built-in defaults: vendor/, node_modules/, *.min.*, dist/, build/)
+    #[arg(long, global = true, action = clap::ArgAction::Append)]
+    pub exclude: Vec<String>,
+
     #[command(subcommand)]
     pub command: Command,
 }
@@ -49,6 +54,12 @@ pub enum Command {
         #[command(subcommand)]
         sub: FindCommand,
     },
+    /// Detect languages/stack at a path before indexing
+    Detect {
+        /// Path to probe (defaults to current directory)
+        #[arg(default_value = ".")]
+        path: PathBuf,
+    },
     /// Start the MCP server over stdio
     Serve,
     /// Show index statistics
@@ -90,6 +101,8 @@ pub enum SearchCommand {
 pub enum InspectCommand {
     /// Source code of a symbol by id or exact name
     Symbol { target: String },
+    /// Symbol source + direct callers + direct importers in one call
+    Bundle { target: String },
     /// Symbol outline of a file
     Outline { file: String },
     /// Read a file or a line range (e.g. --lines 10-40)
diff --git a/src/config.rs b/src/config.rs
index aa428bb..2615f8e 100644
--- a/src/config.rs
+++ b/src/config.rs
@@ -2,6 +2,69 @@
 
 use std::path::{Path, PathBuf};
 
+/// Default `--exclude` globs applied to query results (`find dead`,
+/// `find dead-symbols`, `search symbols`, `search text`). Minified/generated
+/// assets and vendored trees are indexed (so import edges resolve) but are
+/// almost pure noise for symbol/call-site queries, so they are filtered out by
+/// default.
+pub const DEFAULT_EXCLUDES: &[&str] = &["vendor/", "node_modules/", "*.min.*", "dist/", "build/"];
+
+/// Compiled glob matcher for `--exclude`. A path matches if any pattern (the
+/// defaults plus user-supplied globs) matches the slash-normalized relative
+/// path. A trailing `/` marks a directory: it is expanded to `**/<dir>/**` so
+/// it matches at any depth. A bare filename glob (`*.min.*`) is expanded to
+/// `**/*.min.*` for the same reason.
+#[derive(Clone)]
+pub struct ExcludeSet {
+    set: globset::GlobSet,
+}
+
+impl ExcludeSet {
+    pub fn new(extra: &[String]) -> ExcludeSet {
+        let mut builder = globset::GlobSetBuilder::new();
+        for pat in DEFAULT_EXCLUDES
+            .iter()
+            .copied()
+            .chain(extra.iter().map(|s| s.as_str()))
+        {
+            let normalized = normalize_glob(pat);
+            if normalized.is_empty() {
+                continue;
+            }
+            if let Ok(g) = globset::Glob::new(&normalized) {
+                builder.add(g);
+            }
+        }
+        let set = builder.build().unwrap_or_default();
+        ExcludeSet { set }
+    }
+
+    /// True if `rel` (slash-normalized relative path) is excluded.
+    pub fn matches(&self, rel: &str) -> bool {
+        self.set.is_match(rel)
+    }
+}
+
+/// Expand a user glob into a form that matches at any depth.
+fn normalize_glob(pat: &str) -> String {
+    let p = pat.trim();
+    let dir = p.strip_suffix('/');
+    if let Some(d) = dir {
+        if d.is_empty() {
+            return String::new();
+        }
+        return format!("**/{d}/**");
+    }
+    if !p.contains('/') && !p.contains('*') && !p.contains('?') {
+        // A bare name with no slash or glob meta is a directory name.
+        return format!("**/{p}/**");
+    }
+    if !p.contains('/') {
+        return format!("**/{p}");
+    }
+    p.to_string()
+}
+
 /// Directory names skipped during discovery (own list).
 pub const SKIP_DIRS: &[&str] = &[
     ".git",
@@ -112,3 +175,41 @@ pub fn index_db_path(project: &Path) -> PathBuf {
         .unwrap_or_else(|| "project".to_string());
     cache_dir().join(format!("{name}-{key}.db"))
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn normalize_glob_expands_dirs_and_bare_names() {
+        assert_eq!(normalize_glob("vendor/"), "**/vendor/**");
+        assert_eq!(normalize_glob("build"), "**/build/**");
+        assert_eq!(normalize_glob("*.min.*"), "**/*.min.*");
+        assert_eq!(normalize_glob("src/generated/"), "**/src/generated/**");
+        // A slash-containing glob is kept verbatim.
+        assert_eq!(normalize_glob("a/b/*.c"), "a/b/*.c");
+        // Trailing slash alone is dropped.
+        assert_eq!(normalize_glob("/"), "");
+    }
+
+    #[test]
+    fn exclude_set_matches_defaults_at_any_depth() {
+        let set = ExcludeSet::new(&[]);
+        assert!(set.matches("node_modules/left-pad/index.js"));
+        assert!(set.matches("web/bootstrap.min.css"));
+        assert!(set.matches("dist/bundle.js"));
+        assert!(set.matches("sub/build/out.txt"));
+        assert!(set.matches("vendor/foo.rs"));
+        assert!(!set.matches("src/main.rs"));
+        assert!(!set.matches("web/app.css"));
+    }
+
+    #[test]
+    fn exclude_set_merges_extra_globs() {
+        let extra = vec!["generated/".to_string(), "*.pb.go".to_string()];
+        let set = ExcludeSet::new(&extra);
+        assert!(set.matches("generated/foo.rs"));
+        assert!(set.matches("api/gen.pb.go"));
+        assert!(!set.matches("src/foo.rs"));
+    }
+}
diff --git a/src/main.rs b/src/main.rs
index 23e1d70..682ec39 100644
--- a/src/main.rs
+++ b/src/main.rs
@@ -29,6 +29,7 @@ fn main() {
 
 fn run() -> Result<()> {
     let cli = Cli::parse();
+    let excludes = config::ExcludeSet::new(&cli.exclude);
     match &cli.command {
         Command::Index { sub } => match sub {
             IndexCommand::Create { path } => emit(&ops::index_create(path)?, &cli)?,
@@ -41,17 +42,21 @@ fn run() -> Result<()> {
                 include_source,
                 limit,
             } => emit(
-                &ops::search_symbols(&cli.project, query, *include_source, *limit)?,
+                &ops::search_symbols(&cli.project, query, *include_source, *limit, &excludes)?,
+                &cli,
+            )?,
+            SearchCommand::Text { query, limit } => emit(
+                &ops::search_text(&cli.project, query, *limit, &excludes)?,
                 &cli,
             )?,
-            SearchCommand::Text { query, limit } => {
-                emit(&ops::search_text(&cli.project, query, *limit)?, &cli)?
-            }
         },
         Command::Inspect { sub } => match sub {
             InspectCommand::Symbol { target } => {
                 emit(&ops::inspect_symbol(&cli.project, target)?, &cli)?
             }
+            InspectCommand::Bundle { target } => {
+                emit(&ops::inspect_bundle(&cli.project, target)?, &cli)?
+            }
             InspectCommand::Outline { file } => {
                 emit(&ops::inspect_outline(&cli.project, file)?, &cli)?
             }
@@ -72,8 +77,10 @@ fn run() -> Result<()> {
             FindCommand::Importers { file } => {
                 emit(&ops::find_importers(&cli.project, file)?, &cli)?
             }
-            FindCommand::Dead => emit(&ops::find_dead(&cli.project)?, &cli)?,
-            FindCommand::DeadSymbols => emit(&ops::find_dead_symbols(&cli.project)?, &cli)?,
+            FindCommand::Dead => emit(&ops::find_dead(&cli.project, &excludes)?, &cli)?,
+            FindCommand::DeadSymbols => {
+                emit(&ops::find_dead_symbols(&cli.project, &excludes)?, &cli)?
+            }
             FindCommand::Callers { target } => {
                 emit(&ops::find_callers(&cli.project, target)?, &cli)?
             }
@@ -81,6 +88,7 @@ fn run() -> Result<()> {
                 emit(&ops::find_references(&cli.project, identifier)?, &cli)?
             }
         },
+        Command::Detect { path } => emit(&ops::detect(path)?, &cli)?,
         Command::Serve => mcp::serve(&cli.project)?,
         Command::Stats => emit(&ops::stats(&cli.project)?, &cli)?,
     }
diff --git a/src/mcp/mod.rs b/src/mcp/mod.rs
index c2dc8fc..f0b4f64 100644
--- a/src/mcp/mod.rs
+++ b/src/mcp/mod.rs
@@ -125,6 +125,7 @@ fn dispatch(name: &str, args: &Value, project: &Path) -> Result<Value> {
     }
     let opt_str = |k: &str| args.get(k).and_then(|v| v.as_str());
     let opt_depth = |k: &str| args.get(k).and_then(|v| v.as_u64()).map(|d| d as usize);
+    let excludes = crate::config::ExcludeSet::new(&exclude_list(args));
 
     match name {
         "index_create" => ops::index_create(Path::new(req_str(args, "path", name)?)),
@@ -140,6 +141,7 @@ fn dispatch(name: &str, args: &Value, project: &Path) -> Result<Value> {
                 .and_then(|v| v.as_u64())
                 .map(|n| n as usize)
                 .unwrap_or(200),
+            &excludes,
         ),
         "search_text" => ops::search_text(
             project,
@@ -148,6 +150,7 @@ fn dispatch(name: &str, args: &Value, project: &Path) -> Result<Value> {
                 .and_then(|v| v.as_u64())
                 .map(|n| n as usize)
                 .unwrap_or(100),
+            &excludes,
         ),
         "inspect_symbol" => {
             if let Some(id) = args.get("id").and_then(|v| v.as_i64()) {
@@ -156,14 +159,21 @@ fn dispatch(name: &str, args: &Value, project: &Path) -> Result<Value> {
                 ops::inspect_symbol(project, req_str(args, "name", name)?)
             }
         }
+        "inspect_bundle" => {
+            if let Some(id) = args.get("id").and_then(|v| v.as_i64()) {
+                ops::inspect_bundle(project, &id.to_string())
+            } else {
+                ops::inspect_bundle(project, req_str(args, "name", name)?)
+            }
+        }
         "inspect_outline" => ops::inspect_outline(project, req_str(args, "file", name)?),
         "inspect_file" => {
             ops::inspect_file(project, req_str(args, "file", name)?, opt_str("lines"))
         }
         "inspect_tree" => ops::inspect_tree(project, opt_depth("depth")),
         "find_importers" => ops::find_importers(project, req_str(args, "file", name)?),
-        "find_dead" => ops::find_dead(project),
-        "find_dead_symbols" => ops::find_dead_symbols(project),
+        "find_dead" => ops::find_dead(project, &excludes),
+        "find_dead_symbols" => ops::find_dead_symbols(project, &excludes),
         "find_callers" => {
             if let Some(id) = args.get("id").and_then(|v| v.as_i64()) {
                 ops::find_callers(project, &id.to_string())
@@ -176,13 +186,37 @@ fn dispatch(name: &str, args: &Value, project: &Path) -> Result<Value> {
         "inspect_blast_radius" => {
             ops::inspect_blast_radius(project, req_str(args, "file", name)?, opt_depth("depth"))
         }
+        "codebase_detect" => {
+            let path = opt_str("path").map(Path::new).unwrap_or(project);
+            ops::detect(path)
+        }
         "stats" => ops::stats(project),
         _ => anyhow::bail!("unknown tool: {name}"),
     }
 }
 
+/// Parse the optional `exclude` array argument into a list of globs.
+fn exclude_list(args: &Value) -> Vec<String> {
+    args.get("exclude")
+        .and_then(|v| v.as_array())
+        .map(|a| {
+            a.iter()
+                .filter_map(|v| v.as_str().map(String::from))
+                .collect()
+        })
+        .unwrap_or_default()
+}
+
 fn tools() -> Value {
     json!([
+        tool(
+            "codebase_detect",
+            "Detect the languages and stack of a directory before indexing: scans file extensions and manifest files (Cargo.toml, package.json, pom.xml, etc.) without needing ctags or an index. Use first on an unknown project to decide whether and how to index it.",
+            json!({
+                "type": "object",
+                "properties": { "path": { "type": "string", "description": "Directory to probe (defaults to the served project)" } }
+            })
+        ),
         tool(
             "index_create",
             "Index a project from scratch: scan source files, extract symbols via ctags and imports, and write a SQLite index. Run before any other tool on a new project, or to rebuild a stale index.",
@@ -221,7 +255,12 @@ fn tools() -> Value {
                         "type": "boolean",
                         "description": "Include each symbol's source inline (default false)"
                     },
-                    "limit": { "type": "integer", "description": "Maximum results (default 200)" }
+                    "limit": { "type": "integer", "description": "Maximum results (default 200)" },
+                    "exclude": {
+                        "type": "array",
+                        "items": { "type": "string" },
+                        "description": "Extra globs to exclude from results (adds to built-in defaults: vendor/, node_modules/, *.min.*, dist/, build/)"
+                    }
                 },
                 "required": ["query"]
             })
@@ -233,7 +272,12 @@ fn tools() -> Value {
                 "type": "object",
                 "properties": {
                     "query": { "type": "string", "description": "Text or phrase to search for" },
-                    "limit": { "type": "integer", "description": "Maximum matching files (default 100)" }
+                    "limit": { "type": "integer", "description": "Maximum matching files (default 100)" },
+                    "exclude": {
+                        "type": "array",
+                        "items": { "type": "string" },
+                        "description": "Extra globs to exclude from results (adds to built-in defaults: vendor/, node_modules/, *.min.*, dist/, build/)"
+                    }
                 },
                 "required": ["query"]
             })
@@ -249,6 +293,17 @@ fn tools() -> Value {
                 }
             })
         ),
+        tool(
+            "inspect_bundle",
+            "Get a symbol's full source, direct callers, and direct importers in a single call. Replaces three round-trips (inspect_symbol + find_callers + find_importers). Resolve by id or exact name.",
+            json!({
+                "type": "object",
+                "properties": {
+                    "id": { "type": "integer", "description": "Symbol id from search_symbols (unambiguous)" },
+                    "name": { "type": "string", "description": "Exact symbol name (alternative to id; may match multiple)" }
+                }
+            })
+        ),
         tool(
             "inspect_outline",
             "Get the symbol outline of a file: names, kinds, signatures, and line ranges for every top-level symbol. Use to orient yourself in a file without reading it whole.",
@@ -290,12 +345,30 @@ fn tools() -> Value {
         tool(
             "find_dead",
             "List files with no importers (entry-point files like main/lib are excluded). Use to spot orphaned or unreferenced modules.",
-            json!({ "type": "object" })
+            json!({
+                "type": "object",
+                "properties": {
+                    "exclude": {
+                        "type": "array",
+                        "items": { "type": "string" },
+                        "description": "Extra globs to exclude from results (adds to built-in defaults)"
+                    }
+                }
+            })
         ),
         tool(
             "find_dead_symbols",
-            "List symbols never referenced by name outside their own definition. Heuristic: framework-wired symbols (routes, test functions) may appear here because they are invoked by convention, not by name. Use to find genuinely unused functions and methods.",
-            json!({ "type": "object" })
+            "List symbols never referenced by name outside their own definition. Heuristic: framework-wired symbols (routes, test functions, Spring/JUnit beans) are excluded; CSS selectors and vendored/minified files are skipped. Use to find genuinely unused functions and methods.",
+            json!({
+                "type": "object",
+                "properties": {
+                    "exclude": {
+                        "type": "array",
+                        "items": { "type": "string" },
+                        "description": "Extra globs to exclude from results (adds to built-in defaults)"
+                    }
+                }
+            })
         ),
         tool(
             "find_callers",
diff --git a/src/ops.rs b/src/ops.rs
index f0e8284..9ee7840 100644
--- a/src/ops.rs
+++ b/src/ops.rs
@@ -10,6 +10,7 @@ use serde::Serialize;
 use serde_json::Value;
 
 use crate::config;
+use crate::discovery;
 use crate::graph::Graph;
 use crate::index::{pipeline, store};
 use crate::search;
@@ -34,6 +35,72 @@ pub fn index_file(project: &Path, file: &str) -> Result<Value> {
     Ok(serde_json::to_value(stats)?)
 }
 
+// ---- detect ----
+
+/// Manifest files that identify a project's stack, independent of whether the
+/// language is indexed. Used by `detect` before `index create`.
+const MANIFESTS: &[(&str, &str)] = &[
+    ("Cargo.toml", "rust"),
+    ("package.json", "javascript"),
+    ("pom.xml", "java"),
+    ("build.gradle", "java"),
+    ("build.gradle.kts", "java"),
+    ("settings.gradle", "java"),
+    ("requirements.txt", "python"),
+    ("pyproject.toml", "python"),
+    ("setup.py", "python"),
+    ("Pipfile", "python"),
+    ("go.mod", "go"),
+    ("Gemfile", "ruby"),
+    ("composer.json", "php"),
+    ("CMakeLists.txt", "c"),
+];
+
+/// Lightweight pre-index onboarding probe: report the languages/stack present
+/// at `path` from file extensions and manifest files only (no ctags, no
+/// index). Lets an agent decide whether/how to index before paying for it.
+pub fn detect(path: &Path) -> Result<Value> {
+    let files = discovery::discover(path)?;
+    let mut counts: BTreeMap<&str, usize> = BTreeMap::new();
+    for f in &files {
+        *counts.entry(f.language).or_default() += 1;
+    }
+    let languages: Vec<Value> = counts
+        .iter()
+        .map(|(l, c)| serde_json::json!({"language": l, "files": c}))
+        .collect();
+    let manifests: Vec<Value> = detect_manifests(path)
+        .into_iter()
+        .map(|(file, stack)| serde_json::json!({"file": file, "stack": stack}))
+        .collect();
+    let root = path
+        .canonicalize()
+        .unwrap_or_else(|_| path.to_path_buf())
+        .to_string_lossy()
+        .into_owned();
+    Ok(serde_json::json!({
+        "root": root,
+        "languages": languages,
+        "manifests": manifests,
+        "total_files": files.len(),
+        "indexable": !files.is_empty(),
+    }))
+}
+
+fn detect_manifests(path: &Path) -> Vec<(String, &'static str)> {
+    let mut found = Vec::new();
+    let Ok(entries) = std::fs::read_dir(path) else {
+        return found;
+    };
+    for entry in entries.flatten() {
+        let name = entry.file_name().to_string_lossy().into_owned();
+        if let Some((_, stack)) = MANIFESTS.iter().find(|(n, _)| *n == name) {
+            found.push((name, stack));
+        }
+    }
+    found
+}
+
 // ---- search ----
 
 pub fn search_symbols(
@@ -41,9 +108,10 @@ pub fn search_symbols(
     query: &str,
     include_source: bool,
     limit: usize,
+    excludes: &config::ExcludeSet,
 ) -> Result<Value> {
     let conn = open_read(project)?;
-    let hits = search::search_symbols(&conn, query, limit)?;
+    let hits = search::search_symbols(&conn, query, limit, Some(excludes))?;
     if !include_source {
         return Ok(serde_json::to_value(hits)?);
     }
@@ -74,7 +142,12 @@ pub fn search_symbols(
 /// immediately, consistent with `inspect file`/`inspect symbol`. `limit` caps
 /// the number of files that contribute matches; within a file, identical
 /// snippets are collapsed to the first occurrence to reduce noise.
-pub fn search_text(project: &Path, query: &str, limit: usize) -> Result<Value> {
+pub fn search_text(
+    project: &Path,
+    query: &str,
+    limit: usize,
+    excludes: &config::ExcludeSet,
+) -> Result<Value> {
     let conn = open_read(project)?;
     let q = query.trim();
     if q.is_empty() {
@@ -87,6 +160,9 @@ pub fn search_text(project: &Path, query: &str, limit: usize) -> Result<Value> {
         if files_hit >= limit {
             break;
         }
+        if excludes.matches(&f.path) {
+            continue;
+        }
         let path = project.join(&f.path);
         let Ok(content) = std::fs::read_to_string(&path) else {
             continue;
@@ -174,6 +250,50 @@ pub fn inspect_outline(project: &Path, file: &str) -> Result<Value> {
     Ok(serde_json::to_value(symbols)?)
 }
 
+/// Symbol source + direct callers + direct importers in a single call. `target`
+/// is a numeric id or an exact (case-insensitive) name; a name matching several
+/// symbols returns one bundle per match (same shape as `inspect symbol`).
+pub fn inspect_bundle(project: &Path, target: &str) -> Result<Value> {
+    let conn = open_read(project)?;
+    let syms = resolve_symbols(&conn, target)?;
+    let bundles: Vec<Value> = syms
+        .iter()
+        .map(|sym| bundle_object(project, &conn, sym))
+        .collect::<Result<_>>()?;
+    if target.trim().parse::<i64>().is_ok() {
+        Ok(bundles
+            .into_iter()
+            .next()
+            .expect("resolve_symbols is non-empty"))
+    } else {
+        Ok(serde_json::json!(bundles))
+    }
+}
+
+fn bundle_object(project: &Path, conn: &Connection, sym: &store::SymbolRow) -> Result<Value> {
+    let base = symbol_object(project, conn, sym)?;
+    let callers = callers_of(project, conn, std::slice::from_ref(sym))?;
+    let callers_count = callers.len();
+    let importers = store::importers_of(conn, sym.file_id)?
+        .into_iter()
+        .map(|id| store::file_by_id(conn, id))
+        .collect::<std::result::Result<Vec<_>, _>>()?
+        .into_iter()
+        .flatten()
+        .map(|f| f.path)
+        .collect::<Vec<_>>();
+    let importers_count = importers.len();
+    let mut obj = match base {
+        Value::Object(m) => m,
+        _ => serde_json::Map::new(),
+    };
+    obj.insert("callers".into(), serde_json::json!(callers));
+    obj.insert("callers_count".into(), serde_json::json!(callers_count));
+    obj.insert("importers".into(), serde_json::json!(importers));
+    obj.insert("importers_count".into(), serde_json::json!(importers_count));
+    Ok(Value::Object(obj))
+}
+
 pub fn inspect_file(project: &Path, file: &str, lines: Option<&str>) -> Result<Value> {
     let conn = open_read(project)?;
     let fid = file_id(&conn, file)?;
@@ -222,7 +342,7 @@ pub fn find_importers(project: &Path, file: &str) -> Result<Value> {
     Ok(serde_json::to_value(paths)?)
 }
 
-pub fn find_dead(project: &Path) -> Result<Value> {
+pub fn find_dead(project: &Path, excludes: &config::ExcludeSet) -> Result<Value> {
     let conn = open_read(project)?;
     let files = store::all_files(&conn)?;
     let ids: Vec<i64> = files.iter().map(|f| f.id).collect();
@@ -233,7 +353,7 @@ pub fn find_dead(project: &Path) -> Result<Value> {
         .into_iter()
         .filter_map(|id| {
             let f = files.iter().find(|f| f.id == id)?;
-            if is_entry_point(&f.path, &f.language) {
+            if excludes.matches(&f.path) || is_entry_point(&f.path, &f.language) {
                 return None;
             }
             Some(serde_json::json!({"id": f.id, "path": f.path, "language": f.language}))
@@ -250,14 +370,26 @@ pub fn find_dead(project: &Path) -> Result<Value> {
 /// reference exists, so this errs toward under-reporting rather than false
 /// positives. Includes `kind` so callers can filter out structural kinds
 /// (fields, variants) whose names legitimately appear only via qualified paths.
-pub fn find_dead_symbols(project: &Path) -> Result<Value> {
+pub fn find_dead_symbols(project: &Path, excludes: &config::ExcludeSet) -> Result<Value> {
     let conn = open_read(project)?;
     let symbols = store::all_symbols(&conn)?;
     let files = store::all_files(&conn)?;
     let path_of: HashMap<i64, &str> = files.iter().map(|f| (f.id, f.path.as_str())).collect();
+    let lang_of: HashMap<i64, &str> = files.iter().map(|f| (f.id, f.language.as_str())).collect();
 
     let mut by_name: BTreeMap<String, Vec<&store::SymbolRow>> = BTreeMap::new();
     for s in &symbols {
+        let own = path_of.get(&s.file_id).copied().unwrap_or("");
+        // Vendored/generated assets are indexed for import edges but are noise
+        // for symbol queries; CSS selectors and structural kinds (Java
+        // `package`, HTML/Markdown `heading*`) are not callable symbols, so
+        // they can never be meaningfully "dead".
+        if excludes.matches(own)
+            || lang_of.get(&s.file_id).copied() == Some("css")
+            || is_structural_kind(&s.kind)
+        {
+            continue;
+        }
         by_name.entry(s.name.to_lowercase()).or_default().push(s);
     }
     // Lowercased symbol names, used to bound the per-file word index to tokens
@@ -330,6 +462,12 @@ fn index_words(project: &Path, rel: &str, names: &HashSet<&str>) -> HashMap<Stri
     m
 }
 
+/// Structural/documentation kinds that are never "called" and therefore can't
+/// be meaningfully reported as dead symbols.
+fn is_structural_kind(kind: &str) -> bool {
+    kind == "package" || kind.starts_with("heading")
+}
+
 /// Decorator keywords that wire a symbol to an external framework. Such symbols
 /// are referenced by convention/framework, not by a name call-site, so a pure
 /// name-scan wrongly flags them dead.
@@ -349,13 +487,36 @@ const WIRED_DECORATORS: &[&str] = &[
     "receiver",
     "signal",
     "listener",
+    // Java/JUnit & Spring: `@Test`, `@Before(Each)`, plus DI/lifecycle
+    // annotations that wire a bean/component without a name call-site.
+    "test",
+    "before",
+    "service",
+    "component",
+    "bean",
+    "eventlistener",
+    "restcontroller",
+    "scheduled",
+    "configuration",
 ];
 
-/// A symbol is externally wired if it is a pytest `test_*` function or its
+/// A symbol is externally wired if it is a test by naming convention or its
 /// definition is preceded by a framework decorator (Flask/FastAPI route,
-/// pytest fixture, Celery task, click/typer command, discord.py command, …).
+/// pytest fixture, Celery task, click/typer command, discord.py command,
+/// JUnit/Spring annotation, …).
 fn is_wired_entry_point(project: &Path, own: &str, s: &store::SymbolRow) -> bool {
-    if matches!(s.kind.as_str(), "function" | "method") && s.name.starts_with("test_") {
+    if matches!(s.kind.as_str(), "function" | "method" | "member") && s.name.starts_with("test_") {
+        return true;
+    }
+    // Test classes by convention: pytest `TestHealth`/`test_health`, JUnit
+    // `UserServiceTest` (and friends). The runner discovers them, not a name
+    // call-site, so a pure name-scan wrongly flags them dead.
+    if s.kind == "class"
+        && (s.name.starts_with("Test")
+            || s.name.starts_with("test_")
+            || s.name.ends_with("Test")
+            || s.name.ends_with("Tests"))
+    {
         return true;
     }
     let Ok(lines) = read_all_lines(project, own) else {
@@ -850,7 +1011,18 @@ fn raw_to_nodes(node: &RawNode, base: &str, depth: Option<usize>) -> Vec<TreeNod
 
 #[cfg(test)]
 mod tests {
-    use super::{code_only_line, is_entry_point};
+    use super::{code_only_line, is_entry_point, is_structural_kind};
+
+    #[test]
+    fn structural_kinds_are_never_dead() {
+        assert!(is_structural_kind("package"));
+        assert!(is_structural_kind("heading1"));
+        assert!(is_structural_kind("heading3"));
+        assert!(!is_structural_kind("function"));
+        assert!(!is_structural_kind("method"));
+        assert!(!is_structural_kind("class"));
+        assert!(!is_structural_kind("member"));
+    }
 
     #[test]
     fn entry_point_is_language_scoped() {
diff --git a/src/search/mod.rs b/src/search/mod.rs
index ecee61c..9421d1b 100644
--- a/src/search/mod.rs
+++ b/src/search/mod.rs
@@ -6,6 +6,8 @@ use anyhow::Result;
 use rusqlite::{Connection, params};
 use serde::Serialize;
 
+use crate::config::ExcludeSet;
+
 #[derive(Debug, Serialize)]
 pub struct SymbolHit {
     pub id: i64,
@@ -30,13 +32,26 @@ pub struct TextHit {
 const SYMBOL_COLS: &str =
     "s.id, s.name, s.qualified_name, s.kind, s.signature, f.path, s.line_start, s.line_end";
 
-pub fn search_symbols(conn: &Connection, query: &str, limit: usize) -> Result<Vec<SymbolHit>> {
+pub fn search_symbols(
+    conn: &Connection,
+    query: &str,
+    limit: usize,
+    excludes: Option<&ExcludeSet>,
+) -> Result<Vec<SymbolHit>> {
     let q = query.trim();
     let mut seen: HashSet<i64> = HashSet::new();
     let mut out: Vec<SymbolHit> = Vec::new();
     if q.is_empty() {
         return Ok(out);
     }
+    // Fetch a superset when excluding so filtering doesn't starve the output;
+    // the final cap is still `limit`.
+    let fetch_limit = if excludes.is_some() {
+        limit.saturating_mul(8).max(200)
+    } else {
+        limit
+    } as i64;
+    let excluded = |h: &SymbolHit| excludes.is_some_and(|e| e.matches(&h.file));
 
     // 1) exact name (case-insensitive)
     {
@@ -44,7 +59,11 @@ pub fn search_symbols(conn: &Connection, query: &str, limit: usize) -> Result<Ve
             "SELECT {SYMBOL_COLS} FROM symbols s JOIN files f ON f.id = s.file_id \
              WHERE s.name = ?1 COLLATE NOCASE ORDER BY s.id LIMIT ?2"
         ))?;
-        let rows = stmt.query_map(params![q, limit as i64], |r| row_to_hit(r, "exact"))?;
+        let rows = stmt.query_map(params![q, fetch_limit], |r| row_to_hit(r, "exact"))?;
+        let rows = rows.filter_map(|r| match r {
+            Ok(h) if excluded(&h) => None,
+            other => Some(other),
+        });
         push_hits(&mut out, &mut seen, rows, limit)?;
     }
 
@@ -55,7 +74,11 @@ pub fn search_symbols(conn: &Connection, query: &str, limit: usize) -> Result<Ve
             "SELECT {SYMBOL_COLS} FROM symbols s JOIN files f ON f.id = s.file_id \
              WHERE s.name LIKE ?1 COLLATE NOCASE ORDER BY s.id LIMIT ?2"
         ))?;
-        let rows = stmt.query_map(params![pat, limit as i64], |r| row_to_hit(r, "prefix"))?;
+        let rows = stmt.query_map(params![pat, fetch_limit], |r| row_to_hit(r, "prefix"))?;
+        let rows = rows.filter_map(|r| match r {
+            Ok(h) if excluded(&h) => None,
+            other => Some(other),
+        });
         push_hits(&mut out, &mut seen, rows, limit)?;
     }
 
@@ -68,7 +91,11 @@ pub fn search_symbols(conn: &Connection, query: &str, limit: usize) -> Result<Ve
              JOIN files f ON f.id = s.file_id \
              WHERE symbols_fts MATCH ?1 ORDER BY rank LIMIT ?2"
         ))?;
-        let rows = stmt.query_map(params![phrase, limit as i64], |r| row_to_hit(r, "fulltext"))?;
+        let rows = stmt.query_map(params![phrase, fetch_limit], |r| row_to_hit(r, "fulltext"))?;
+        let rows = rows.filter_map(|r| match r {
+            Ok(h) if excluded(&h) => None,
+            other => Some(other),
+        });
         push_hits(&mut out, &mut seen, rows, limit)?;
     }
 
@@ -79,7 +106,11 @@ pub fn search_symbols(conn: &Connection, query: &str, limit: usize) -> Result<Ve
             "SELECT {SYMBOL_COLS} FROM symbols s JOIN files f ON f.id = s.file_id \
              WHERE s.name LIKE ?1 COLLATE NOCASE ORDER BY s.id LIMIT ?2"
         ))?;
-        let rows = stmt.query_map(params![pat, limit as i64], |r| row_to_hit(r, "substring"))?;
+        let rows = stmt.query_map(params![pat, fetch_limit], |r| row_to_hit(r, "substring"))?;
+        let rows = rows.filter_map(|r| match r {
+            Ok(h) if excluded(&h) => None,
+            other => Some(other),
+        });
         push_hits(&mut out, &mut seen, rows, limit)?;
     }