feat: Better search tool (#739)

# Motivation

<!-- Why is this change necessary? -->

# Content

<!-- Please include a summary of the change -->

# Testing

<!-- How was the change tested? -->

# Please check the following before marking your PR as ready for review

- [ ] I have added tests for my changes
- [ ] I have updated the documentation or added new documentation as
needed

---------

Co-authored-by: kopekC <28070492+kopekC@users.noreply.github.com>

Author: kopekC

src/codegen/extensions/index/file_index.py

@@ -2,7 +2,9 @@
 
 import pickle
 from pathlib import Path
+from typing import Optional
 
+import modal
 import numpy as np
 import tiktoken
 from openai import OpenAI
@@ -26,6 +28,7 @@ class FileIndex(CodeIndex):
     EMBEDDING_MODEL = "text-embedding-3-small"
     MAX_TOKENS = 8000
     BATCH_SIZE = 100
+    USE_MODAL_DICT = True  # Flag to control whether to use Modal Dict
 
     def __init__(self, codebase: Codebase):
         """Initialize the file index.
@@ -37,10 +40,87 @@ def __init__(self, codebase: Codebase):
         self.client = OpenAI()
         self.encoding = tiktoken.get_encoding("cl100k_base")
 
+    def set_use_modal_dict(self, use_modal: bool) -> None:
+        """Set whether to use Modal Dict for storage.
+
+        Args:
+            use_modal: Whether to use Modal Dict for storage
+        """
+        self.USE_MODAL_DICT = use_modal
+        logger.info(f"Modal Dict storage {'enabled' if use_modal else 'disabled'}")
+
     @property
     def save_file_name(self) -> str:
         return "file_index_{commit}.pkl"
 
+    @property
+    def modal_dict_id(self) -> str:
+        """Get the Modal Dict ID based on the same naming convention as the pickle file."""
+        if not self.commit_hash:
+            return "file_index_latest"
+        return f"file_index_{self.commit_hash}"
+
+    def delete_modal_dict(self) -> bool:
+        """Delete the Modal Dict storage for this index.
+
+        Returns:
+            bool: True if successfully deleted, False otherwise
+        """
+        if not self.USE_MODAL_DICT:
+            logger.warning("Modal Dict storage is disabled")
+            return False
+
+        try:
+            dict_id = self.modal_dict_id
+            logger.info(f"Deleting Modal Dict: {dict_id}")
+
+            # Check if the dict exists before trying to delete
+            try:
+                # Use modal.Dict.delete to properly delete the dict
+                modal.Dict.delete(dict_id)
+                logger.info(f"Successfully deleted Modal Dict: {dict_id}")
+                return True
+            except Exception as e:
+                logger.info(f"Modal Dict {dict_id} does not exist or cannot be deleted: {e}")
+                return False
+        except Exception as e:
+            logger.exception(f"Failed to delete Modal Dict: {e}")
+            return False
+
+    def modal_dict_exists(self, commit_hash: Optional[str] = None) -> bool:
+        """Check if a Modal Dict exists for a specific commit.
+
+        Args:
+            commit_hash: The commit hash to check, or None to use the current commit
+
+        Returns:
+            bool: True if the Modal Dict exists, False otherwise
+        """
+        if not self.USE_MODAL_DICT:
+            return False
+
+        try:
+            # Use provided commit hash or current one
+            old_commit = self.commit_hash
+            if commit_hash is not None:
+                self.commit_hash = commit_hash
+
+            dict_id = self.modal_dict_id
+
+            # Restore original commit hash
+            if commit_hash is not None:
+                self.commit_hash = old_commit
+
+            try:
+                # Try to access the dict - this will raise an exception if it doesn't exist
+                modal_dict = modal.Dict.from_name(dict_id, create_if_missing=False)
+                # Check if our data is in the dict
+                return "index_data" in modal_dict
+            except Exception:
+                return False
+        except Exception:
+            return False
+
     def _split_by_tokens(self, text: str) -> list[str]:
         """Split text into chunks that fit within token limit."""
         tokens = self.encoding.encode(text)
@@ -135,17 +215,69 @@ def _get_changed_items(self) -> set[File]:
         return changed_files
 
     def _save_index(self, path: Path) -> None:
-        """Save index data to disk."""
+        """Save index data to disk and optionally to Modal Dict."""
+        # Save to local pickle file
         with open(path, "wb") as f:
             pickle.dump({"E": self.E, "items": self.items, "commit_hash": self.commit_hash}, f)
 
+        # Save to Modal Dict if enabled
+        if self.USE_MODAL_DICT:
+            try:
+                dict_id = self.modal_dict_id
+                logger.info(f"Saving index to Modal Dict: {dict_id}")
+
+                # Convert numpy arrays to lists for JSON serialization
+                modal_data = {"E": self.E.tolist() if self.E is not None else None, "items": self.items.tolist() if self.items is not None else None, "commit_hash": self.commit_hash}
+
+                # Create or update Modal Dict
+                # Note: from_name is lazy, so we need to explicitly set the data
+                modal_dict = modal.Dict.from_name(dict_id, create_if_missing=True)
+                modal_dict["index_data"] = modal_data
+
+                logger.info(f"Successfully saved index to Modal Dict: {dict_id}")
+            except Exception as e:
+                logger.exception(f"Failed to save index to Modal Dict: {e}")
+
     def _load_index(self, path: Path) -> None:
-        """Load index data from disk."""
-        with open(path, "rb") as f:
-            data = pickle.load(f)
-            self.E = data["E"]
-            self.items = data["items"]
-            self.commit_hash = data["commit_hash"]
+        """Load index data from disk or Modal Dict."""
+        # Try loading from Modal Dict first if enabled
+        if self.USE_MODAL_DICT:
+            try:
+                dict_id = self.modal_dict_id
+                logger.info(f"Attempting to load index from Modal Dict: {dict_id}")
+
+                # from_name is lazy, so we need to check if the dict exists first
+                try:
+                    modal_dict = modal.Dict.from_name(dict_id, create_if_missing=False)
+                    # Check if the dict contains our data
+                    if "index_data" in modal_dict:
+                        data = modal_dict["index_data"]
+
+                        # Convert lists back to numpy arrays
+                        self.E = np.array(data["E"]) if data["E"] is not None else None
+                        self.items = np.array(data["items"]) if data["items"] is not None else None
+                        self.commit_hash = data["commit_hash"]
+
+                        logger.info(f"Successfully loaded index from Modal Dict: {dict_id}")
+                        return
+                    else:
+                        logger.info(f"No index data found in Modal Dict: {dict_id}")
+                except Exception as e:
+                    logger.warning(f"Modal Dict {dict_id} not found or error accessing it: {e}")
+            except Exception as e:
+                logger.warning(f"Failed to load index from Modal Dict, falling back to local file: {e}")
+
+        # Fall back to loading from local file
+        try:
+            with open(path, "rb") as f:
+                data = pickle.load(f)
+                self.E = data["E"]
+                self.items = data["items"]
+                self.commit_hash = data["commit_hash"]
+                logger.info(f"Loaded index from local file: {path}")
+        except Exception as e:
+            logger.exception(f"Failed to load index from local file: {e}")
+            raise
 
     def similarity_search(self, query: str, k: int = 5) -> list[tuple[File, float]]:
         """Find the k most similar files to a query.
@@ -216,3 +348,20 @@ def update(self) -> None:
 
         # Update commit hash
         self.commit_hash = self._get_current_commit()
+
+        # Save updated index to Modal Dict if enabled
+        if self.USE_MODAL_DICT and (num_updated > 0 or num_added > 0):
+            try:
+                dict_id = self.modal_dict_id
+                logger.info(f"Updating index in Modal Dict: {dict_id}")
+
+                # Convert numpy arrays to lists for JSON serialization
+                modal_data = {"E": self.E.tolist() if self.E is not None else None, "items": self.items.tolist() if self.items is not None else None, "commit_hash": self.commit_hash}
+
+                # Create or update Modal Dict
+                modal_dict = modal.Dict.from_name(dict_id, create_if_missing=True)
+                modal_dict["index_data"] = modal_data
+
+                logger.info(f"Successfully updated index in Modal Dict: {dict_id}")
+            except Exception as e:
+                logger.exception(f"Failed to update index in Modal Dict: {e}")

src/codegen/extensions/langchain/tools.py

@@ -110,25 +110,111 @@ def _run(self, dirpath: str = "./", depth: int = 1) -> str:
 
 
 class SearchInput(BaseModel):
-    """Input for searching the codebase."""
-
     query: str = Field(
         ...,
-        description="The search query to find in the codebase. When ripgrep is available, this will be passed as a ripgrep pattern. "
-        "For regex searches, set use_regex=True. Ripgrep is the preferred method.",
+        description="""The text or pattern to search for in the codebase.
+
+        For simple text search (use_regex=False):
+        - Uses ripgrep's fixed-strings mode (--fixed-strings)
+        - Case-insensitive matching (--ignore-case)
+        - All characters are treated literally, including special regex characters
+        - Exact string matching (no regex interpretation)
+
+        For regex search (use_regex=True):
+        - Full regex pattern support
+        - Case-sensitive by default
+        - Special characters have regex meaning and need proper escaping
+        - Uses ripgrep's default regex mode
+
+        If no exact matches are found, automatically falls back to semantic search
+        to find relevant code even without exact text matches.""",
+    )
+
+    target_directories: Optional[list[str]] = Field(
+        default=None,
+        description="""Optional list of directories to limit the search scope.
+
+        - Paths should be relative to the workspace root
+        - Multiple directories are searched in parallel
+        - If None, searches the entire codebase
+
+        Example: ["src/frontend", "tests/unit"]""",
+    )
+
+    file_extensions: Optional[list[str]] = Field(
+        default=None,
+        description="""Optional list of file extensions to filter the search.
+
+        - Include the dot in extensions (e.g. ['.py', '.ts'])
+        - Multiple extensions are combined with OR logic
+        - If None, searches all file types
+        - Binary files are automatically excluded
+
+        Example: [".py", ".tsx", ".md"]""",
+    )
+
+    page: int = Field(
+        default=1,
+        description="""Page number for paginated results (1-based indexing).
+
+        - Use with files_per_page to navigate large result sets
+        - If page exceeds available pages, returns last available page
+        - Note: When falling back to semantic search, pagination is not supported
+
+        Example: page=2 with files_per_page=10 shows files 11-20""",
+    )
+
+    files_per_page: int = Field(
+        default=10,
+        description="""Number of files to show per page.
+
+        - Each file can contain multiple matching lines
+        - Reasonable values are between 5 and 50
+        - Larger values may impact performance
+        - When falling back to semantic search, this becomes the number of semantic results
+
+        Example: files_per_page=20 shows up to 20 files with matches""",
+    )
+
+    use_regex: bool = Field(
+        default=False,
+        description="""Whether to treat the query as a regex pattern.
+
+        - False (default): Simple text search, case-insensitive
+        - True: Full regex syntax, case-sensitive
+        - Invalid regex patterns will return an error
+        - Note: Semantic fallback is used regardless of this setting when no matches found
+
+        Example: Set to True to use patterns like "test_.*_func.*" """,
     )
-    target_directories: Optional[list[str]] = Field(default=None, description="Optional list of directories to search in")
-    file_extensions: Optional[list[str]] = Field(default=None, description="Optional list of file extensions to search (e.g. ['.py', '.ts'])")
-    page: int = Field(default=1, description="Page number to return (1-based, default: 1)")
-    files_per_page: int = Field(default=10, description="Number of files to return per page (default: 10)")
-    use_regex: bool = Field(default=False, description="Whether to treat query as a regex pattern (default: False)")
 
 
 class SearchTool(BaseTool):
     """Tool for searching the codebase."""
 
     name: ClassVar[str] = "search"
-    description: ClassVar[str] = "Search the codebase using text search or regex pattern matching"
+    description: ClassVar[str] = r"""Search the codebase using text search or regex pattern matching.
+
+    This tool provides powerful text-based search capabilities across your codebase,
+    with support for both simple text matching and regular expressions. It uses ripgrep
+    when available for high-performance searches.
+
+    If no exact matches are found, automatically falls back to semantic search to find
+    relevant code even without exact text matches.
+
+    Features:
+    - Plain text or regex pattern matching
+    - Directory and file type filtering
+    - Paginated results for large codebases
+    - Case-insensitive by default for simple text searches
+    - Semantic fallback for finding related code
+
+    Example queries:
+    1. Simple text: "function calculateTotal" (matches exactly, case-insensitive)
+    2. Regex: "def.*calculate.*\(.*\)" (with use_regex=True)
+    3. File-specific: "TODO" with file_extensions=[".py", ".ts"]
+    4. Directory-specific: "api" with target_directories=["src/backend"]
+    """
     args_schema: ClassVar[type[BaseModel]] = SearchInput
     codebase: Codebase = Field(exclude=True)
 
@@ -151,7 +237,27 @@ class EditFileTool(BaseTool):
     """Tool for editing files."""
 
     name: ClassVar[str] = "edit_file"
-    description: ClassVar[str] = "Edit a file by replacing its entire content. This tool should only be used for replacing entire file contents."
+    description: ClassVar[str] = r"""
+Edit a file by replacing its entire content. This tool should only be used for replacing entire file contents.
+Input for searching the codebase.
+
+    This tool provides powerful text-based search capabilities across your codebase,
+    with support for both simple text matching and regular expressions. It uses ripgrep
+    when available for high-performance searches, falling back to Python's regex engine
+    when necessary.
+
+    Features:
+    - Plain text or regex pattern matching
+    - Directory and file type filtering
+    - Paginated results for large codebases
+    - Case-insensitive by default for simple text searches
+
+    Example queries:
+    1. Simple text: "function calculateTotal" (matches exactly, case-insensitive)
+    2. Regex: "def.*calculate.*\(.*\)" (with use_regex=True)
+    3. File-specific: "TODO" with file_extensions=[".py", ".ts"]
+    4. Directory-specific: "api" with target_directories=["src/backend"]
+    """
     args_schema: ClassVar[type[BaseModel]] = EditFileInput
     codebase: Codebase = Field(exclude=True)
 
@@ -741,7 +847,7 @@ def get_workspace_tools(codebase: Codebase) -> list["BaseTool"]:
         RunBashCommandTool(),  # Note: This tool doesn't need the codebase
         SearchTool(codebase),
         # SemanticEditTool(codebase),
-        SemanticSearchTool(codebase),
+        # SemanticSearchTool(codebase),
         ViewFileTool(codebase),
         RelaceEditTool(codebase),
         ReflectionTool(codebase),
@@ -761,14 +867,26 @@ def get_workspace_tools(codebase: Codebase) -> list["BaseTool"]:
 
 
 class ReplacementEditInput(BaseModel):
-    """Input for regex-based replacement editing."""
-
-    filepath: str = Field(..., description="Path to the file to edit")
-    pattern: str = Field(..., description="Regex pattern to match")
-    replacement: str = Field(..., description="Replacement text (can include regex groups)")
-    start: int = Field(default=1, description="Starting line number (1-indexed, inclusive). Default is 1.")
-    end: int = Field(default=-1, description="Ending line number (1-indexed, inclusive). Default is -1 (end of file).")
-    count: Optional[int] = Field(default=None, description="Maximum number of replacements. Default is None (replace all).")
+    filepath: str = Field(..., description="Path to the file to edit relative to the workspace root. The file must exist and be a text file.")
+    pattern: str = Field(
+        ...,
+        description="Regular expression pattern to match text that should be replaced. Supports all Python regex syntax including capture groups (\1, \2, etc). The pattern is compiled with re.MULTILINE flag by default.",
+    )
+    replacement: str = Field(
+        ...,
+        description="Text to replace matched patterns with. Can reference regex capture groups using \1, \2, etc. If using regex groups in pattern, make sure to preserve them in replacement if needed.",
+    )
+    start: int = Field(
+        default=1, description="Starting line number (1-indexed, inclusive) to begin replacements from. Use this with 'end' to limit changes to a specific region. Default is 1 (start of file)."
+    )
+    end: int = Field(
+        default=-1,
+        description="Ending line number (1-indexed, inclusive) to stop replacements at. Use -1 to indicate end of file. Use this with 'start' to limit changes to a specific region. Default is -1 (end of file).",
+    )
+    count: Optional[int] = Field(
+        default=None,
+        description="Maximum number of replacements to make. Use None to replace all occurrences (default), or specify a number to limit replacements. Useful when you only want to replace the first N occurrences.",
+    )
 
 
 class ReplacementEditTool(BaseTool):