feat:Introduced new tools for getting the directory structure and git status, earlier agent had to loop three times to get it

Author: ayusrjn

lambda_agent/subagent.py

@@ -47,18 +47,34 @@
 # ---------------------------------------------------------------------------
 
 # Default tools — the main agent can override per-task
-_DEFAULT_TOOL_NAMES = ["read_file", "search_repo", "run_command", "write_file"]
+_DEFAULT_TOOL_NAMES = [
+    "read_file",
+    "search_repo",
+    "run_command",
+    "write_file",
+    "list_directory",
+    "get_git_status",
+]
 
 
 def _get_tool_set() -> dict:
     """Lazily import tool functions from tools.py to avoid circular imports."""
-    from .tools import read_file, search_repo, run_command, write_file
+    from .tools import (
+        read_file,
+        search_repo,
+        run_command,
+        write_file,
+        list_directory,
+        get_git_status,
+    )
 
     return {
         "read_file": read_file,
         "search_repo": search_repo,
         "run_command": run_command,
         "write_file": write_file,
+        "list_directory": list_directory,
+        "get_git_status": get_git_status,
     }
 
 

lambda_agent/tools.py

@@ -64,34 +64,205 @@ def run_command(command: str) -> str:
         return f"Error executing command: {str(e)}"
 
 
-def get_workspace_summary() -> str:
-    """Gathers git context, branch, status, recent commits, and project documentation (like README.md or rule files) to help the agent understand the whole project."""
-    summary_parts = []
+# ---------------------------------------------------------------------------
+# Dedicated directory listing tool
+# ---------------------------------------------------------------------------
+
+# Directories to always skip when walking the tree
+_IGNORE_DIRS = {
+    ".git",
+    "__pycache__",
+    "node_modules",
+    ".venv",
+    "venv",
+    "env",
+    ".ruff_cache",
+    "dist",
+    "build",
+    ".next",
+    ".cache",
+    ".tox",
+    ".mypy_cache",
+    ".pytest_cache",
+    "*.egg-info",
+}
+
+
+def _should_skip_dir(name: str) -> bool:
+    """Return True if a directory name matches any ignore pattern."""
+    if name in _IGNORE_DIRS:
+        return True
+    # Handle glob-style suffix patterns like *.egg-info
+    for pat in _IGNORE_DIRS:
+        if pat.startswith("*") and name.endswith(pat[1:]):
+            return True
+    return False
+
+
+def list_directory(path: str = ".", max_depth: int = 3, git_aware: bool = True) -> str:
+    """Lists directory contents as a tree structure with smart filtering.
+
+    Returns a compact, indented tree of files and directories.  By default it
+    respects .gitignore (via `git ls-files`) so the model never wastes tokens
+    on build artefacts or vendored deps.
+
+    Args:
+        path: Root directory to list (defaults to current directory '.').
+        max_depth: How many levels deep to recurse (defaults to 3).
+        git_aware: If True, use git ls-files to respect .gitignore (defaults to True).
+    """
+    abs_path = os.path.abspath(path)
+    if not os.path.isdir(abs_path):
+        return f"Error: '{path}' is not a directory."
+
+    # ---- Fast path: use git ls-files when inside a repo ----
+    if git_aware:
+        try:
+            tracked = subprocess.check_output(
+                ["git", "ls-files", "--cached", "--others", "--exclude-standard", path],
+                text=True,
+                stderr=subprocess.DEVNULL,
+            ).strip()
+            if tracked:
+                lines = tracked.splitlines()
+                if len(lines) > 300:
+                    return (
+                        "\n".join(lines[:300])
+                        + f"\n\n... and {len(lines) - 300} more files."
+                    )
+                return "\n".join(lines)
+        except Exception:
+            pass  # Not a git repo or git not available — fall through to manual walk
+
+    # ---- Fallback: manual os.scandir walk ----
+    output_lines: list[str] = []
+
+    def _walk(dir_path: str, prefix: str, depth: int):
+        if depth > max_depth:
+            return
+        try:
+            entries = sorted(
+                os.scandir(dir_path), key=lambda e: (not e.is_dir(), e.name.lower())
+            )
+        except PermissionError:
+            return
+
+        visible_dirs = [
+            e
+            for e in entries
+            if e.is_dir()
+            and not _should_skip_dir(e.name)
+            and not e.name.startswith(".")
+        ]
+        files = [e for e in entries if e.is_file()]
+        combined = visible_dirs + files
+
+        for i, entry in enumerate(combined):
+            is_last = i == len(combined) - 1
+            connector = "└── " if is_last else "├── "
+            suffix = "/" if entry.is_dir() else ""
+            output_lines.append(f"{prefix}{connector}{entry.name}{suffix}")
+            if entry.is_dir():
+                extension = "    " if is_last else "│   "
+                _walk(entry.path, prefix + extension, depth + 1)
+
+    _walk(abs_path, "", 1)
+    result = "\n".join(output_lines)
+    if len(result) > 4000:
+        result = result[:4000] + "\n...[TRUNCATED]"
+    return result or "Empty directory."
+
+
+# ---------------------------------------------------------------------------
+# Dedicated git status tool
+# ---------------------------------------------------------------------------
+
+
+def get_git_status(include_diff: bool = False) -> str:
+    """Returns a comprehensive git status summary in a single call.
+
+    Bundles branch name, porcelain status, and recent commits so the agent
+    does not need multiple run_command calls.
+
+    Args:
+        include_diff: If True, also include a condensed diff stat of staged and unstaged changes (defaults to False).
+    """
+    parts: list[str] = []
 
-    # 1. Gather Git Context
+    # Branch
     try:
         branch = subprocess.check_output(
-            ["git", "branch", "--show-current"], text=True, stderr=subprocess.STDOUT
+            ["git", "branch", "--show-current"], text=True, stderr=subprocess.DEVNULL
         ).strip()
-        status = subprocess.check_output(
-            ["git", "status", "-s"], text=True, stderr=subprocess.STDOUT
-        )
-        log = subprocess.check_output(
-            ["git", "log", "-n", "5", "--oneline"], text=True, stderr=subprocess.STDOUT
-        )
-
-        summary_parts.append(
-            f"### Git Context\n**Branch**: {branch}\n**Status**:\n{status if status else 'Clean'}\n**Recent Commits**:\n{log}"
-        )
+        parts.append(f"Branch: {branch}")
     except Exception:
-        summary_parts.append("### Git Context\nNot a git repository or git error.")
+        return "Not a git repository."
 
-    # 2. Gather Directory Structure (limited to root)
+    # Status (porcelain for compact, stable output)
     try:
-        files = os.listdir(".")
-        summary_parts.append(f"### Root Directory Files\n{', '.join(files)}")
+        status = subprocess.check_output(
+            ["git", "status", "--porcelain=v2", "--branch"],
+            text=True,
+            stderr=subprocess.DEVNULL,
+        ).strip()
+        parts.append(f"Status:\n{status}" if status else "Status: Clean working tree")
     except Exception as e:
-        summary_parts.append(f"### Directory Listing Error\n{e}")
+        parts.append(f"Status error: {e}")
+
+    # Recent commits
+    try:
+        log = subprocess.check_output(
+            ["git", "log", "-n", "5", "--oneline", "--no-decorate"],
+            text=True,
+            stderr=subprocess.DEVNULL,
+        ).strip()
+        if log:
+            parts.append(f"Recent commits:\n{log}")
+    except Exception:
+        pass
+
+    # Optional diff stat
+    if include_diff:
+        try:
+            diff = subprocess.check_output(
+                ["git", "diff", "--stat", "--no-color"],
+                text=True,
+                stderr=subprocess.DEVNULL,
+            ).strip()
+            if diff:
+                parts.append(f"Unstaged changes:\n{diff}")
+        except Exception:
+            pass
+        try:
+            staged = subprocess.check_output(
+                ["git", "diff", "--cached", "--stat", "--no-color"],
+                text=True,
+                stderr=subprocess.DEVNULL,
+            ).strip()
+            if staged:
+                parts.append(f"Staged changes:\n{staged}")
+        except Exception:
+            pass
+
+    return "\n\n".join(parts)
+
+
+# ---------------------------------------------------------------------------
+# Workspace summary (used at session start)
+# ---------------------------------------------------------------------------
+
+
+def get_workspace_summary() -> str:
+    """Gathers git context, project structure, and documentation to help the agent understand the whole project."""
+    summary_parts = []
+
+    # 1. Git context — reuse the dedicated tool
+    git_info = get_git_status()
+    summary_parts.append(f"### Git Context\n{git_info}")
+
+    # 2. Project structure — reuse the dedicated tool (depth 2 to keep it compact)
+    tree = list_directory(".", max_depth=2)
+    summary_parts.append(f"### Project Structure\n{tree}")
 
     # 3. Read important docs
     docs_to_check = [
@@ -214,6 +385,8 @@ def finish_task(message: str) -> str:
     "read_file": read_file,
     "write_file": write_file,
     "run_command": run_command,
+    "list_directory": list_directory,
+    "get_git_status": get_git_status,
     "search_repo": search_repo,
     "ask_user": ask_user,
     "finish_task": finish_task,
@@ -227,6 +400,8 @@ def finish_task(message: str) -> str:
     read_file,
     write_file,
     run_command,
+    list_directory,
+    get_git_status,
     search_repo,
     ask_user,
     finish_task,