Jump to content

Connect SuperML | Leeroopedia MCP: Equip your AI agents with best practices, code verification, and debugging knowledge. Powered by Leeroo — building Organizational Superintelligence. Contact us at founders@leeroo.com.

Implementation:Langchain ai Langchain Anthropic File Search

From Leeroopedia
Revision as of 11:23, 16 February 2026 by Admin (talk | contribs) (Auto-imported from implementations/Langchain_ai_Langchain_Anthropic_File_Search.md)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Knowledge Sources
Domains LLM Framework, Anthropic, Agent Middleware, File Search
Last Updated 2026-02-11 00:00 GMT

Overview

The file_search.py module provides Glob and Grep search middleware that operates on virtual files stored in LangGraph state, enabling agents to search through files managed by the Anthropic text editor and memory tool middleware.

Description

This module implements StateFileSearchMiddleware, an AgentMiddleware class that adds two search tools to an agent: glob_search (file pattern matching by path) and grep_search (content search using regular expressions). These tools operate on files stored in the AnthropicToolsState virtual file system, making them complementary to the StateClaudeTextEditorMiddleware and StateClaudeMemoryMiddleware. The module also contains internal helper functions for brace-expansion of glob patterns and include-pattern validation.

Usage

Import StateFileSearchMiddleware and include it in the middleware list alongside a state-based text editor or memory middleware. The search tools will automatically have access to the virtual files in state.

Code Reference

Source Location

  • Repository: Langchain_ai_Langchain
  • File: libs/partners/anthropic/langchain_anthropic/middleware/file_search.py
  • Lines: 1-354

Signature

class StateFileSearchMiddleware(AgentMiddleware):
    state_schema = AnthropicToolsState

    def __init__(self, *, state_key: str = "text_editor_files") -> None: ...

Import

from langchain_anthropic.middleware.file_search import StateFileSearchMiddleware

I/O Contract

Inputs

Name Type Required Description
state_key str No State key to search. Default "text_editor_files". Use "memory_files" to search memory tool files

Tool: glob_search Inputs

Name Type Required Description
pattern str Yes The glob pattern to match files against (e.g., **/*.py)
path str No The directory to search in (default: "/")

Tool: grep_search Inputs

Name Type Required Description
pattern str Yes Regular expression pattern to search for in file contents
path str No The directory to search in (default: "/")
include str or None No File pattern filter (e.g., "*.js", "*.{ts,tsx}")
output_mode Literal No Output format: "files_with_matches" (default), "content", or "count"

Outputs

Name Type Description
result str Newline-separated search results. Returns "No files found" or "No matches found" if empty

Key Mechanisms

Glob Search

The glob search uses PurePosixPath.match() to match file paths against glob patterns. It handles the **/ prefix specially to also match files in the base directory. Results are sorted by modification time (most recently modified first).

Grep Search

The grep search compiles the pattern as a Python regular expression and searches line-by-line through file contents. It supports three output modes:

  • files_with_matches: Returns only the file paths containing matches (sorted)
  • content: Returns file:line_number:content format for each matching line
  • count: Returns file:count format showing match counts per file

Brace Expansion

The internal _expand_include_patterns function handles brace expansion for include filters (e.g., *.{py,pyi} becomes ["*.py", "*.pyi"]). This enables users to specify multiple file extensions in a single include pattern.

Pattern Validation

Include patterns are validated by _is_valid_include_pattern, which:

  • Rejects empty patterns
  • Rejects patterns containing null bytes, newlines, or carriage returns
  • Verifies brace expansion succeeds
  • Confirms each expanded pattern compiles to a valid regex via fnmatch.translate

Usage Examples

Basic Usage

from langchain.agents import create_agent
from langchain_anthropic.middleware.anthropic_tools import (
    StateClaudeTextEditorMiddleware,
)
from langchain_anthropic.middleware.file_search import StateFileSearchMiddleware

agent = create_agent(
    model=model,
    tools=[],
    middleware=[
        StateClaudeTextEditorMiddleware(),
        StateFileSearchMiddleware(),  # Searches text_editor_files by default
    ],
)

Searching Memory Files

from langchain_anthropic.middleware.file_search import StateFileSearchMiddleware

# Search memory tool files instead of text editor files
search_middleware = StateFileSearchMiddleware(state_key="memory_files")

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.
Principle
Implementation
Heuristic
Environment