Smart Tree - ST

# Smart Tree Claude Wishlist ## Efficiency Enhancements - **Digest-First Workflow**: Prioritize digest checks to minimize unnecessary processing and data transfer. - **Pagination and Fields Selector**: Implement pagination and fields selector to reduce payload size and improve response times. - **Summary-AI Mode**: Default to summary-ai for compact overviews, reducing token usage. ## Compatibility Improvements - **Compression Capability Probe**: Implement a capability probe to determine client support for compression, ensuring compatibility. - **Strict Mode Enhancements**: Enhance strict mode to ensure JSON-only outputs and deterministic ordering. ### 1. Show Line Content in Search Results ⭐⭐⭐⭐⭐ (IMPLEMENTED ✅) **Status**: Implemented in v4.0.0 **Description**: `search_in_files` now returns actual matching lines with content! **New Features Added**: - `include_content` parameter (default: true for AI) - `max_matches_per_file` parameter (default: 20) - `context_lines` parameter (prepared for future enhancement) - Returns line number, content, and column for each match ```json // Example output: { "path": "/src/main.rs", "matches": 3, "lines": [ { "line_number": 42, "content": "// TODO: Add better error handling", "column": 3 }, { "line_number": 156, "content": "fn process_todo_items() {", "column": 14 } ] } ``` **Use Case**: When fixing imports or TODOs, you can now see the context without opening each file! ### 2. Batch File Read Tool ⭐⭐⭐⭐ **Tool Name**: `read_files_from_search` **Description**: Read multiple files based on search results **Use Case**: After finding all files with a specific pattern, read them all in one operation ``` # Step 1: Search results = search_in_files(path="/project", keyword="StoredVector") # Step 2: Read all matching files contents = read_files_from_search(results, max_files=10) ``` ### 3. Find and Replace Tool ⭐⭐⭐⭐⭐ **Tool Name**: `find_and_replace` **Description**: Replace text across multiple files with preview **Parameters**: - `path`: Directory to search - `find_pattern`: Text or regex to find - `replace_with`: Replacement text - `file_pattern`: Optional file filter - `preview`: Show changes before applying **Use Case**: Renaming functions, updating imports, fixing consistent typos ## Medium Priority Features ### 4. Dependency Graph Analysis ⭐⭐⭐ **Tool Name**: `analyze_dependencies` **Description**: Show module/crate dependencies as a graph **Output**: Mermaid diagram showing relationships **Use Case**: Understanding project structure, identifying circular dependencies ### 5. Import Analysis Tool ⭐⭐⭐ **Tool Name**: `analyze_imports` **Description**: Show what each file imports and exports **Use Case**: Refactoring module structure, understanding dependencies ### 6. Symbol Search ⭐⭐⭐⭐ **Tool Name**: `find_symbol` **Description**: Find type/function/trait definitions **Example**: `find_symbol(name="StoredVector", type="struct")` **Use Case**: Quickly locating type definitions without grep ## Quality of Life Improvements ### 7. Relative Path Options ⭐⭐⭐ **Enhancement**: Add `path_display` option to all tools **Options**: `absolute`, `relative`, `from_root` **Use Case**: Cleaner output for documentation and reports ### 8. File Type Groups ⭐⭐⭐ **Enhancement**: Predefined file type groups **Groups**: - `rust_src`: `.rs` files excluding tests - `config_all`: All config files - `tests_all`: All test files **Use Case**: `find_files(path="/", type_group="rust_src")` ### 9. Context-Aware Search ⭐⭐⭐ **Enhancement**: Search with semantic understanding **Example**: `search_in_files(keyword="error handling", context="functions")` **Use Case**: Find error handling code without matching comments ### 10. Cached Workspace Analysis ⭐⭐⭐⭐ **Enhancement**: Cache analysis results with TTL **Benefits**: - Instant results for large codebases - Incremental updates on changes - Reduced token usage **Use Case**: Repeatedly analyzing large monorepos ## Bug Fixes ### 11. Empty Directory Handling **Issue**: `analyze_directory` sometimes fails on empty directories **Fix**: Gracefully handle empty directories with clear message ### 12. Large File Streaming **Issue**: Memory spike when processing very large files (>100MB) **Fix**: Implement proper streaming for all file operations ### 13. Directory-Only Filtering (FIXED ✅) **Issue**: `--type d` was misleading - it filters by file extension, not entry type **Fix**: Added `--entry-type f|d` to filter files vs directories **Example**: `st --find ".*" --entry-type d` now correctly shows only directories **Status**: Fixed in v3.3.0 ### 14. LS Mode with Filtered Results (FIXED ✅) **Issue**: When using `--find` with `-m ls`, only parent directories were shown **Fix**: LS formatter now detects filtered results and shows full paths for matches **Example**: `st --find "MiniLM" --entry-type d -m ls` now shows all matching directories with full paths **Note**: The `-a` flag only shows hidden files but doesn't override default ignores (like `.cache`). Use `--everything` for that. **Status**: Fixed in v3.3.0 ### 15. Time-Aware MCP Tools (FIXED ✅) **Issue**: AI assistants don't know the current date when using date filters **Fix**: - Added current date/time to `server_info` response - Added new `find_in_timespan` tool for searching files in a date range **Example**: `find_in_timespan(path="/home", start_date="2025-07-10", end_date="2025-07-13")` **Status**: Fixed in v3.3.1 ### 16. Intuitive Sort Options & LS Mode Default Depth (FIXED ✅) **Issue**: - Sort options were confusing (e.g., `--sort size` didn't clarify if it was ascending or descending) - LS mode showed full tree instead of defaulting to depth 1 like real `ls` command **Fix**: - Added intuitive sort options: `largest/smallest`, `newest/oldest`, `a-to-z/z-to-a` - LS mode now defaults to depth 1 (shows only immediate children) - Legacy options (`name`, `size`, `date`) still work for backward compatibility **Examples**: - `st . --mode ls --sort largest --top 10` - Shows 10 largest files - `st . --mode ls --sort oldest` - Shows files oldest first - `st . --mode ls` - Now shows only depth 1 by default **Status**: Fixed in v3.3.1 ### 17. Auto-Detection and Installation of Shell Completions (FIXED ✅) **Issue**: Users had to manually generate and install shell completions **Fix**: - Install script now auto-detects shell (bash/zsh/fish) - Offers to install completions during installation - Enhanced zsh completions with tips and SQL-like examples - Standalone `setup-completions.sh` script for existing installations **Features**: - Auto-detects user's shell from $SHELL or /etc/passwd - Finds appropriate completion directories - Offers to add sourcing to shell config files - Downloads enhanced completions for zsh with tips **Example**: ```bash # During install curl -sSL https://raw.githubusercontent.com/8b-is/smart-tree/main/scripts/install.sh | bash # Prompts: "Would you like to install shell completions?" # For existing installations ./scripts/setup-completions.sh ``` **Status**: Fixed in v3.3.1 ### 18. Updated MCP Tools List & Built-in MCP Support (FIXED ✅) **Issue**: - `--mcp-tools` showed outdated, incomplete list (only 2 tools) - manage.sh still referenced MCP as a feature when it's now built-in **Fix**: - Updated `--mcp-tools` to show all 20+ available tools organized by category - Removed MCP feature references from manage.sh since MCP is now standard - Added helpful tips and pro-tips to the tools list **Changes**: - `--mcp-tools` now shows: Core Tools, File Discovery, Content Search, Analysis, Advanced, and Feedback categories - manage.sh: `mcp-build` command now just runs regular build - Removed "Build with MCP support" - it's always included **Status**: Fixed in v3.3.1 ### 19. Hex Mode Search Results in Hex Format (FIXED ✅) **Issue**: Search match positions showed decimal values in hex mode (e.g., `[SEARCH:L670:C5]`) **Fix**: Updated hex formatter to display search positions in hexadecimal **Example**: - Before: `[SEARCH:L670:C5,4x]` (line 670, column 5, 4 matches) - After: `[SEARCH:L29e:C5,4x]` (line 0x29e, column 0x5, 0x4 matches) **Rationale**: Maintains consistency with hex output format **Status**: Fixed in v3.3.1 ### 20. Elegant Project Renaming - Identity Transition (IMPLEMENTED ✅) **Feature**: Context-aware project renaming that understands code semantics **Command**: `st --rename-project "OldName" "NewName"` **Capabilities**: - Detects and converts between naming conventions (snake_case, camelCase, etc.) - Context-aware replacements (function names, strings, comments, configs) - Handles multiple file types (Rust, Python, JS, TOML, YAML, JSON, Markdown) - Interactive preview mode with confidence scores - Safety features: backup, dry-run, selective editing - Optional logo generation **Example**: ```bash st --rename-project "BobsAmazingGame" "F1 Racing" # Shows: # ✅ Found 41 matches across: # - 5 source files (.rs, .py) # - 3 config files (Cargo.toml, package.json) # - 2 docs (README.md) # 🎨 Context-aware replacements: # • Identifiers → `f1_racing` # • Strings → "F1 Racing" # • Titles → `F1 Racing` ``` **Status**: Implemented in v3.3.1 ### 21. Per-Directory Sorting in Classic Mode (FIXED ✅) **Issue**: `--sort` option didn't work with classic tree mode **Root Cause**: Classic formatter re-sorted all nodes by path for tree structure **Fix**: Implemented per-directory sorting that maintains tree hierarchy **Solution**: - Each directory's children are sorted independently - Sort options (largest/smallest, newest/oldest, a-to-z/z-to-a) work within each directory - Maintains proper parent-child relationships for tree visualization **Examples**: ```bash st --sort largest -d 2 src # Shows largest files first in each directory st --sort oldest src # Shows oldest files first, per directory st --sort z-to-a src # Reverse 4.0.0betical within each directory ``` **Note**: This mimics how file explorers sort - each folder has its own sort order **Status**: Fixed in v3.3.1 ### 22. Smart Auto-Switch for --top Option (IMPLEMENTED ✅) **Feature**: Automatically switch to ls mode when using --top **Issue**: `--top` doesn't work with classic tree mode (needs all entries for structure) **Solution**: - Auto-detects when `--top` is used without explicit mode selection - Switches to `ls` mode automatically for useful results - Respects explicit mode choices with helpful note **Behavior**: ```bash # Auto-switches to ls mode: st --sort largest --top 5 # Output: Top 5 largest files in ls format # Respects explicit mode: st --mode classic --sort largest --top 5 # Output: Full tree with note about --top limitation ``` **Smart Detection**: Only auto-switches if user didn't specify --mode or ST_DEFAULT_MODE **Status**: Implemented in v3.3.1 ### 23. LS Mode Sorting Preservation (FIXED ✅) **Issue**: LS formatter was re-sorting nodes 4.0.0betically, overriding user's --sort preference **Root Cause**: ls.rs had hardcoded 4.0.0betical sorting at line 296 **Fix**: Removed automatic sorting in ls formatter to preserve scanner's sort order **Impact**: All --sort options now work correctly in ls mode **Examples**: ```bash st --sort largest --mode ls # Shows files sorted by size st --sort newest # Shows newest files first st --sort z-to-a --mode ls # Reverse 4.0.0betical order ``` **Status**: Fixed in v3.3.1 ### 24. Smart Depth Auto-Detection (IMPLEMENTED ✅) **Feature**: Each mode gets its ideal default depth when not specified **Issue**: Default depth 5 was hardcoded, preventing users from explicitly requesting depth 5 **Solution**: - Changed default depth to 0 (auto) - Each mode picks its optimal depth when depth = 0 - Users can explicitly set any depth including 5 **Mode Defaults**: - LS mode: 1 (shows only immediate children like real ls) - Classic: 3 (balanced tree view) - AI/Hex: 5 (more detail for analysis) - Stats/Digest/Waste/Relations: 10 (comprehensive scan) - Others: 4 (reasonable default) **Examples**: ```bash st # Classic mode with depth 3 st --mode ls # LS mode with depth 1 st -d 5 # Explicit depth 5 (works now!) st --mode stats # Stats with depth 10 ``` **Status**: Implemented in v3.3.1 ### 25. Enhanced File Type Detection & Semantic Emojis (IMPLEMENTED ✅) **Feature**: Rich, context-aware file type detection with beautiful emoji mapping **Issue**: Limited file type categories - missing databases, media variants, web assets, etc. **Solution**: - Expanded FileCategory enum with 40+ categories - Created centralized emoji_mapper module - Added support for .mem8/.m8 (MEM|8 files with 🧠 emoji!) - Rich categorization: databases, office docs, fonts, 3D models, certificates, etc. **New Categories Added**: - **Databases**: .db, .sqlite, .mdb (🗄️) - **Office**: .docx, .xlsx, .pptx, .pdf, .epub (📄📊📕📚) - **Media**: .m4a audio support, .webp images (🎵🖼️) - **Security**: .cert, .pem, .gpg (🔐🔒) - **Scientific**: .ipynb, .rdata, .mat (📓📊📐) - **Web**: .wasm, .map (🌐) - **3D/CAD**: .stl, .obj, .blend (🎲) - **Fonts**: .ttf, .woff, .woff2 (🔤) - **Special**: .mem8, .m8 - MEM|8 memory files (🧠) **Example Output**: ``` 📁 test_files ├── 🦀 test.rs (Rust) ├── 🐍 test.py (Python) ├── 🗄️ database.db (Database) ├── 🧠 memory.mem8 (MEM|8!) ├── 📓 notebook.ipynb (Jupyter) ├── 🔐 server.cert (Certificate) └── 🎲 model.stl (3D Model) ``` **Status**: Implemented in v3.3.1 ### 26. ST_DEFAULT_MODE Environment Variable Precedence Bug (FIXED ✅) **Issue**: ST_DEFAULT_MODE environment variable incorrectly overrides explicit --mode command line arguments **Current Behavior**: - When ST_DEFAULT_MODE is set, it takes precedence over user-provided --mode flag - Example: `ST_DEFAULT_MODE=hex st --mode classic` results in hex mode, not classic **Expected Behavior**: Command line arguments should always override environment variables **Root Cause**: - The mode selection logic in main.rs checks ST_DEFAULT_MODE before using args.mode - Since args.mode always has a value (default "classic"), the code can't distinguish between explicit and default values **Code Location**: src/main.rs lines 519-534 **Current precedence (incorrect)**: 1. AI_TOOLS environment variable 2. --semantic flag 3. ST_DEFAULT_MODE environment variable ❌ 4. --mode command line argument 5. Default value **Expected precedence**: 1. AI_TOOLS environment variable 2. --semantic flag 3. --mode command line argument 4. ST_DEFAULT_MODE environment variable 5. Default value **Status**: Fixed in v3.3.1 ✅ **Fix Applied**: - Changed default mode to "auto" to detect explicit --mode usage - Updated mode selection logic to prioritize command line args - Now `ST_DEFAULT_MODE=hex st --mode ls` correctly shows ls mode ### 27. File History Tracking - The Ultimate Context-Driven System (IMPLEMENTED ✅) **Feature**: Comprehensive AI file manipulation tracking system **Description**: Logs all AI file operations to `~/.mem8/.filehistory/` with hash-based change detection **Implementation**: - Created `file_history` module with operation tracking - Added MCP tools: `track_file_operation`, `get_file_history`, `get_project_history_summary` - 10-minute resolution timestamps for efficient log grouping - Append-first preference for least intrusive operations **Features**: - **Operation Codes**: A=Append, P=Prepend, I=Insert, D=Delete, R=Replace, C=Create, X=Remove, M=Relocate, N=Rename, r=Read - **Hash Tracking**: Before/after SHA256 hashes for every change - **Project Organization**: Logs stored by project ID under `~/.mem8/.filehistory/project_id/YYYYMMDD_HHMM.flg` - **Session Grouping**: Related operations tracked with session IDs **Example Usage**: ```bash # Track a file operation mcp.callTool('track_file_operation', { file_path: '/src/main.rs', old_content: 'fn main() {}', new_content: 'fn main() {\n println!("Hello!");\n}', agent: 'claude' }) # Get file history mcp.callTool('get_file_history', { file_path: '/src/main.rs' }) # Shows: timestamp, operation, agent, session, bytes affected, hashes # Get project summary mcp.callTool('get_project_history_summary', { project_path: '/my/project' }) # Shows: total operations, files modified, operation breakdown ``` **Status**: Implemented in v4.0.0 ## Performance Enhancements ### 25. Parallel Search Operations ⭐⭐⭐⭐ **Enhancement**: Accept multiple search patterns in one call **Example**: `search_in_files(patterns=["TODO", "FIXME", "HACK"])` **Benefit**: 3x faster than sequential searches ### 26. Smart Tree Index ⭐⭐⭐⭐⭐ **Feature**: Background indexer for instant searches **Commands**: - `st --index /project` - Build index - `st --search-indexed "pattern"` - Instant search **Use Case**: Near-instant searches in massive codebases ## MCP-Specific Enhancements ### 27. Tool Chaining Support ⭐⭐⭐ **Feature**: Output of one tool as input to another **Example**: `find_files` → `read_files` → `analyze_content` **Benefit**: Reduce round trips, faster operations ### 28. Batch Operations Tool ⭐⭐⭐⭐ **Tool Name**: `batch_operations` **Description**: Execute multiple operations in one call **Use Case**: Reduce API calls, improve performance ## Integration Features ### 29. Git-Aware Operations ⭐⭐⭐ **Enhancement**: All tools respect `.gitignore` by default **Option**: `--include-ignored` to override **Use Case**: Avoid noise from build artifacts ### 30. Language Server Protocol Info ⭐⭐⭐ **Tool Name**: `get_lsp_info` **Description**: Extract LSP-like information (definitions, references) **Use Case**: Smarter code navigation and refactoring ## Working Features (Confirmations) ### Auto-Detection for Ignored Directories ✅ **Status**: Confirmed working correctly **Behavior**: When explicitly requesting ignored directories (e.g., `st .git` or `st node_modules`), smart-tree automatically: - Disables default ignore patterns - Shows hidden files for directories starting with `.` - Displays full directory content **Test Results**: - `st .git` - Shows complete git directory structure - `st node_modules` - Shows all packages and files **Note**: This is ideal for AI tools that need to inspect these directories when specifically requested ## Wishlist Metadata **Last Updated**: 2025-07-21 **Total Requests**: 32 **High Priority**: 3 **Fixed Issues**: 12 **Implemented Features**: 5 **Bug Reports**: 1 **Working Features Confirmed**: 1 **Submitted By**: Claude (Opus 4) --- ## How to Contribute AI assistants using Smart Tree can append to this file with new feature requests. Please include: 1. Clear description of current vs desired behavior 2. Practical use case demonstrating the value 3. Priority rating (⭐ to ⭐⭐⭐⭐⭐) 4. Example code/command showing how it would work The Smart Tree team reviews this file regularly to prioritize development!