initial project wide agentic code update implemented with project and preview

Author: gcclinux

Makefile

@@ -2,7 +2,7 @@
 
 # Variables
 BINARY_NAME=ti
-VERSION=0.0.2.2
+VERSION=0.0.2.4
 BUILD_DIR=build
 GO=go
 BUILD_NUMBER=$(shell printf "%04d" $$(($(shell git rev-list --count HEAD 2>/dev/null || echo "0") + 1)))

README.md

@@ -132,6 +132,69 @@ The Git integration uses:
 
 For more technical details, see [internal/git/README.md](internal/git/README.md).
 
+## Agentic Project Assist
+
+The AI assistant can now operate across your **entire project** — not just the open file. Use `/project` to ask the AI to find, read, and modify multiple files at once based on a natural-language request.
+
+### How It Works
+
+When you send a `/project` command, the AI runs a three-phase pipeline:
+
+1. **Scanning** — recursively walks the workspace directory, collecting all text files (up to 500, skipping `.git`, `vendor`, `node_modules`, `.ti`, `build`)
+2. **Ranking** — sends the file list and your request to the AI model, which returns up to 20 files most likely to need changes
+3. **Modifying** — reads each relevant file (up to 2000 lines), generates `SEARCH/REPLACE` patches, validates them, and writes the changes to disk
+
+A change report is displayed when the operation completes, listing every file read, modified, or that encountered an error.
+
+### Commands
+
+```
+/project <request>          Run a project-wide change
+/preview /project <request> Dry-run: show what would change without writing files
+/proceed                    Apply the changes from the last dry-run
+```
+
+### Examples
+
+```
+/project add error handling to all HTTP client calls
+/project rename the Config struct to AppConfig everywhere
+/project improve the aichat pane height calculation
+/preview /project update all log.Printf calls to use structured logging
+```
+
+### Change Report
+
+After each operation you'll see a summary like:
+
+```
+Project-wide operation complete
+
+Files scanned: 42
+Files modified: 3
+  - internal/ui/aichat.go (+5 -2)
+  - internal/config/config.go (+1 -0)
+  - README.md (+3 -1)
+
+Patch failures: 1
+  - internal/executor/executor.go: search text not found in file content
+```
+
+If nothing was changed: `No files were modified.`
+
+### Safety
+
+- Only files inside the workspace root are ever read or written
+- Symlinks that resolve outside the workspace are rejected
+- Paths returned by the AI that don't exist on disk are discarded (logged as "hallucinated paths")
+- Preview mode (`/preview /project`) never writes anything to disk
+
+### Supported File Types
+
+The scanner considers: `.go`, `.md`, `.sh`, `.bash`, `.ps1`, `.py`, `.ts`, `.js`, `.json`, `.yaml`, `.yml`, `.toml`, `.txt`, `.html`, `.css`
+
+---
+
 ## Agentic Code Fixing
 
 The AI assistant can autonomously fix code issues in your open files. Simply describe what you want to change, and the AI will read your code, generate a fix, and apply it directly to the editor.

docs/USAGE.md

@@ -4,47 +4,22 @@
 
 ```bash
 # Run with config file (if ~/.ti/config.json exists)
-./ti
-
-# Override config with command-line flags
-./ti -workspace /path/to/workspace
-
-# Specify Ollama URL (overrides config)
-./ti -ollama http://localhost:11434
-
-# Specify model (overrides config)
-./ti -model llama2
-
-# Use Gemini instead of Ollama
-./ti -gemini YOUR_API_KEY -model gemini-2.0-flash-exp
+./ti-linux-amd64
 
 # Show version
-./ti -version
+./ti-linux-amd64 -version
 
-# Show help
-./ti -help
 
 ## Starting the Application
 
 ```bash
-./build/ti -model qwen2.5-coder:3b -ollama http://192.168.0.78:11434 -workspace ~/ti-workspace
-```
-
-```bash
-./build/ti -model qwen2.5-coder:1.5b -ollama http://localhost:11434 -workspace ~/ti-workspace
+./ti-linux-amd6
 ```
 
 ```powershell
-.\build\ti.exe -model qwen2.5-coder:3b -ollama http://localhost:11434 -workspace C:\Users\ricardo\Programming\test
+.\build\ti-windows-amd6.exe
 ``` 
 
-```bash
-./build/ti -model gemini-2.5-flash-lite -gemini api_key_goes_here -workspace ~/ti-workspace
-```
-```bash
-./build/ti -model gemini-3-flash-preview -gemini api_key_goes_here -workspace ~/ti-workspace
-```
-
 
 ## Key Features
 
@@ -81,6 +56,9 @@ The right pane is now split horizontally:
 - `/fix <request>` - Force agentic mode (AI will modify code)
 - `/ask <question>` - Force conversational mode (AI won't modify code)
 - `/preview <request>` - Preview changes before applying them
+- `/project <request>` - Run a project-wide change across all files in the workspace
+- `/preview /project <request>` - Dry-run a project-wide change without writing files
+- `/proceed` - Apply the changes from the last `/preview /project` dry-run
 - `/model` - Display current agent, model, and API key (for Gemini)
 - `/help` - Display keyboard shortcuts and agent commands
 
@@ -147,3 +125,5 @@ At the bottom of the screen, you'll see:
 - Unsaved files show a `*` in the title
 - AI responses show timestamps
 - Messages with code context are marked `[with context]`
+- `/project` works best with a focused, specific request — the AI picks up to 20 files to modify per run
+- Use `/preview /project` first on large or risky changes to see what would be affected before committing