snapshot-query 0.1.2

Efficiently query and analyze snapshot log files generated by cursor-ide-browser

snapshot-query

Efficiently query and analyze snapshot log files generated by cursor-ide-browser.

Features

• ✅ Multiple Query Methods:

  • Find by name (fuzzy/exact matching)
  • Find by role
  • Find by reference identifier
  • Regular expression queries (grep syntax)
  • CSS/jQuery selector queries
  • BM25 relevance-ranked queries

• ✅ Data Validation:

  • Pydantic-based data validation
  • Type-safe data models

• ✅ MCP Server Interface:

  • Model Context Protocol support
  • Query snapshot files via MCP tools

• ✅ Command Line Tool:

  • Rich CLI commands
  • Multiple output formats

Installation

Using uvx (Recommended, no installation needed)

# Run directly without installation
uvx snapshot-query <file_path> <command> [args]

Using pip

pip install snapshot-query

After installation, you can use:

snapshot-query <file_path> <command> [args]

Using Python Module

python -m snapshot_query <file_path> <command> [args]

Quick Start

Basic Queries

# Find elements containing "search"
uvx snapshot-query snapshot.log find-name "search"

# Find all buttons
uvx snapshot-query snapshot.log find-role button

# List all interactive elements
uvx snapshot-query snapshot.log interactive

# Count elements
uvx snapshot-query snapshot.log count

Command Reference

find-name

Find elements by name (fuzzy matching)

uvx snapshot-query snapshot.log find-name "search"

find-name-exact

Find elements by name (exact matching)

uvx snapshot-query snapshot.log find-name-exact "search"

find-name-bm25

Find elements using BM25 algorithm (ranked by relevance)

# Return all relevant results
uvx snapshot-query snapshot.log find-name-bm25 "search"

# Return top 10 most relevant results
uvx snapshot-query snapshot.log find-name-bm25 "search" 10

find-role

Find elements by role

uvx snapshot-query snapshot.log find-role button

find-ref

Find element by reference identifier

uvx snapshot-query snapshot.log find-ref ref-b9k8zlttiah

find-text

Find elements containing specified text

uvx snapshot-query snapshot.log find-text "login"

find-grep

Find elements using regular expressions (grep syntax)

# Search name field (default)
uvx snapshot-query snapshot.log find-grep "^search$" name

# Search role field
uvx snapshot-query snapshot.log find-grep "button|link" role

# Search ref field
uvx snapshot-query snapshot.log find-grep "ref-.*abc" ref

find-selector

Find elements using CSS/jQuery selector syntax

# Tag selector (role)
uvx snapshot-query snapshot.log find-selector "button"

# ID selector (ref)
uvx snapshot-query snapshot.log find-selector "#ref-rv7cgg62t9g"

# Attribute selector
uvx snapshot-query snapshot.log find-selector "[name='search']"

# Combined selector
uvx snapshot-query snapshot.log find-selector "button[name='search']"

# Contains match
uvx snapshot-query snapshot.log find-selector "[name*='Google']"

interactive

List all interactive elements

uvx snapshot-query snapshot.log interactive

count

Count elements by type

uvx snapshot-query snapshot.log count

path

Show element path in tree

uvx snapshot-query snapshot.log path ref-b9k8zlttiah

all-refs

List all reference identifiers

uvx snapshot-query snapshot.log all-refs

convert-to-markdown

Convert snapshot log file to Markdown format

# Convert and output to console
uvx snapshot-query snapshot.log convert-to-markdown

# Convert and save to file
uvx snapshot-query snapshot.log convert-to-markdown output.md

# Convert without ref identifiers
uvx snapshot-query snapshot.log convert-to-markdown output.md --no-ref

# Convert with maximum depth limit
uvx snapshot-query snapshot.log convert-to-markdown output.md --max-depth 3

Options:

• output_file (optional): Output file path. If not provided, outputs to console
• --no-ref: Exclude ref identifiers from output
• --max-depth <number>: Limit rendering depth (useful for large snapshots)

Output Format: The generated Markdown document includes:

• Document Header: Source file name and generation timestamp
• Overview Section: Brief introduction and total element count
• Statistics Section:
  • Element count by role in a table format with percentages
  • Interactive elements summary
• Accessibility Tree Structure:
  • Hierarchical tree view with proper headings
  • Each element shows: role, name (if available), and reference identifier
  • Child elements count for each parent
• Interactive Elements Reference:
  • Tables listing all interactive elements (links, buttons, textboxes, etc.)
  • Each table shows name and reference identifier for quick lookup
• Notes Section: Additional information about the document and usage

The document is structured as a coherent, readable Markdown file suitable for documentation, sharing, or further processing.

Common Query Scenarios

Scenario 1: Find button ref

Problem: I want to click a "search" button and need to find its ref

Solution:

uvx snapshot-query snapshot.log find-name "search"

Output Example:

Found 1 matching element:
role: button
ref: ref-b9k8zlttiah
name: search

Usage:

browser_click(
  element="Search button",
  ref="ref-b9k8zlttiah"
)

Scenario 2: Find all clickable elements

Problem: I want to see what interactive elements are on the page

Solution:

uvx snapshot-query snapshot.log interactive

Scenario 3: Find input boxes

Problem: I need to find the ref of a username input box

Solution:

# Method 1: Find all textboxes
uvx snapshot-query snapshot.log find-role textbox

# Method 2: Find by name
uvx snapshot-query snapshot.log find-name "username"

Scenario 4: Understand page structure

Problem: I want to know where an element is located in the page

Solution:

uvx snapshot-query snapshot.log path ref-b9k8zlttiah

Output Example:

Element path:

Level 0:
  role: generic
  ref: ref-zketxgetcys

Level 1:
  role: generic
  ref: ref-p37ecs217hp

Level 2:
  role: generic
  ref: ref-zhx4wavxy6q

Level 3:
  role: button
  ref: ref-b9k8zlttiah
  name: search

Scenario 5: Batch find links

Problem: I want to find all links containing "news"

Solution:

# Use find-text to find elements containing "news"
uvx snapshot-query snapshot.log find-text "news"

# Or use find-selector for combined query
uvx snapshot-query snapshot.log find-selector "link[name*='news']"

Advanced Usage

Using Python API

from snapshot_query import SnapshotQuery

# Load snapshot file
query = SnapshotQuery("snapshot.log")

# Find all buttons
buttons = query.find_by_role("button")
print(f"Found {len(buttons)} buttons")

# Find elements containing "login"
login_elements = query.find_by_name("login")
for elem in login_elements:
    print(f"Name: {elem.name}, ref: {elem.ref}")

# Use BM25 search
results = query.find_by_name_bm25("search", top_k=5)
for elem in results:
    print(f"Name: {elem.name}, ref: {elem.ref}")

# Use selector
results = query.find_by_selector("button[name='search']")
for elem in results:
    print(f"Name: {elem.name}, ref: {elem.ref}")

# Get element path
path = query.get_element_path("ref-b9k8zlttiah")
print(f"Element path depth: {len(path)}")

Using grep for quick search (suitable for large files)

# Windows PowerShell
Select-String -Path "snapshot.log" -Pattern "name: search" -Context 0,2

# Linux/Mac
grep -A 2 "name: search" snapshot.log

Batch processing multiple snapshot files

from pathlib import Path
from snapshot_query import SnapshotQuery

log_dir = Path(r"C:\Users\{username}\.cursor\browser-logs")

for log_file in log_dir.glob("snapshot-*.log"):
    print(f"\nProcessing file: {log_file.name}")
    query = SnapshotQuery(log_file)
    # Execute queries...

MCP Server Interface

snapshot-query provides an MCP (Model Context Protocol) server interface that allows AI assistants to query snapshot files through a standardized protocol.

Installation

Ensure MCP SDK is installed:

pip install mcp

Starting MCP Server

Method 1: Using command entry point (requires package installation)

pip install snapshot-query
snapshot-query-mcp

Method 2: Using Python module

python -m snapshot_query.mcp_server

Client Configuration

Configure the server in Cursor IDE or other MCP clients. Edit the MCP configuration file (usually at ~/.cursor/mcp.json or similar location):

Using installed package (recommended):

{
  "mcpServers": {
    "snapshot-query": {
      "command": "snapshot-query-mcp",
      "args": []
    }
  }
}

Using Python module:

{
  "mcpServers": {
    "snapshot-query": {
      "command": "python",
      "args": ["-m", "snapshot_query.mcp_server"]
    }
  }
}

Using uvx (recommended, no local installation needed):

{
  "mcpServers": {
    "snapshot-query": {
      "command": "uvx",
      "args": ["--from", "snapshot-query", "python", "-m", "snapshot_query.mcp_server"]
    }
  }
}

Note: When using uvx, MCP SDK needs to be installed separately. It's recommended to use the installed package method.

Available Tools

The MCP server provides the following tools:

1. find_by_name - Find elements by name (supports fuzzy and exact matching)
2. find_by_name_bm25 - Find elements using BM25 algorithm (ranked by relevance)
3. find_by_role - Find elements by role type
4. find_by_ref - Find element by reference identifier
5. find_by_text - Find elements containing specified text
6. find_by_regex - Find elements using regular expressions
7. find_by_selector - Find elements using CSS/jQuery selectors
8. find_interactive_elements - Find all interactive elements
9. count_elements - Count elements by type
10. get_element_path - Get element path in tree
11. extract_all_refs - Extract all reference identifiers from snapshot file

Usage Scenarios

In AI Assistants: After configuring the MCP server, AI assistants can directly call these tools to query snapshot files.

Working with cursor-ide-browser:

1. Use browser_snapshot() to get page snapshot
2. Snapshot is saved as log file
3. Query snapshot file via MCP interface to find element ref
4. Use ref for browser interaction operations

Real-world Workflow Example

# 1. Get page snapshot (in browser)
browser_snapshot()

# 2. Find target element
uvx snapshot-query snapshot.log find-name "login"

# 3. Get ref: ref-xxxxx

# 4. Use ref for operation
browser_click(element="Login button", ref="ref-xxxxx")

Tips

1. Combine queries: Use find-name to find approximate location, then use find-ref to get detailed information
2. Use path: The path command helps understand page structure
3. Batch processing: Write scripts to process multiple snapshot files
4. Cache results: Save frequently queried refs to configuration files
5. BM25 ranking: Use find-name-bm25 for smarter relevance-ranked results

FAQ

Q: Why can't I find an element?
A: Check if the element name is correct, try using fuzzy search find-name instead of exact search, or use find-name-bm25 for better matching results.

Q: Do ref values change?
A: Yes, ref values change with each page refresh or new snapshot. You need to get a new snapshot to get new ref values.

Q: How to batch process multiple snapshot files?
A: Use scripts to loop through files:

from pathlib import Path
from snapshot_query import SnapshotQuery

for log_file in Path(".cursor/browser-logs").glob("snapshot-*.log"):
    query = SnapshotQuery(log_file)
    # Execute queries...

Q: MCP SDK import error
A: Ensure MCP SDK is installed: pip install mcp

Dependencies

• Python >= 3.8
• PyYAML >= 6.0
• Pydantic >= 2.0.0
• MCP >= 1.0.0

More Information

For detailed documentation, see:

• cursor-ide-browser.md - cursor-ide-browser tool documentation
• AGENTS.md - Project development guide