Optimization/func

README.md

@@ -12,6 +12,8 @@ A lightweight CLI tool that analyzes your staged changes and generates professio
 
 ## Features
 
+- **Hybrid Intelligence** - Merges deterministic heuristic algorithms with deep semantic understanding via Local LLMs
+- **Local AI Integration (Ollama)** - Seamlessly integrates with Ollama to generate context-rich commit messages offline
 - **Intelligent Analysis** - Analyzes git status and diff to understand your changes using advanced pattern detection
 - **Conventional Commits** - Follows the Conventional Commits specification for standardized messages
 - **Configuration Hierarchy** - Local (`.gitmit.json`) → Global (`~/.gitmit.json`) → Default (Embedded) config support
@@ -30,10 +32,46 @@ A lightweight CLI tool that analyzes your staged changes and generates professio
 - **Pattern Detection** - Detects error handling, tests, API changes, database operations, and more
 - **Multiple Commit Types** - Supports feat, fix, refactor, chore, test, docs, style, perf, ci, build, security, and more
 - **Zero Configuration** - Works out of the box with sensible defaults
-- **Offline First** - Complete offline operation, no AI or external dependencies required
+- **Hybrid Offline Approach** - Operates completely locally; chooses between heuristic rules or local AI
 - **History Tracking** - Learns from your commit history to avoid repetitive suggestions
 
 
+## Hybrid Intelligence: Local AI via Ollama
+
+Gitmit evolves from a rule-based utility into a **Hybrid Intelligence** tool by integrating with **Ollama**. This allows you to use powerful local LLMs (like Qwen2.5-Coder) to generate highly descriptive and contextually accurate commit messages without sacrificing privacy or performance.
+
+### Prerequisites
+
+1.  **Install Ollama**: Download from [ollama.com](https://ollama.com)
+2.  **Pull the Model**:
+    ```bash
+    ollama pull qwen2.5-coder:3b
+    ```
+
+### Enabling the AI Engine
+
+To switch from the default heuristic engine to the AI engine, update your `.gitmit.json`:
+
+```json
+{
+  "engine": "ollama",
+  "ollama": {
+    "model": "qwen2.5-coder:3b",
+    "url": "http://localhost:11434",
+    "temperature": 0.2
+  }
+}
+```
+
+### How the Hybrid Pipeline Works
+
+Gitmit uses a tiered approach to ensure you always get a high-quality suggestion:
+
+1.  **Refined Context Engineering**: Instead of piping raw diffs, Gitmit extracts structured metadata (modified files, symbols, change ratios, project type) and feeds it into a strict system prompt.
+2.  **Local LLM Processing**: Ollama processes the context using your specified model to generate a precise Conventional Commit message.
+3.  **Deterministic Fallback**: If the Ollama daemon is unreachable, the model is missing, or the output is malformatted, Gitmit **instantly falls back** to the Phase 1 Heuristic engine. Your workflow remains uninterrupted.
+
+
 ## Installation
 
 ### From Releases
@@ -110,15 +148,18 @@ Actions:
   n - Reject and exit
   e - Edit message manually
   r - Regenerate different suggestion
+  a - Upgrade suggestion with Local AI (Ollama)
 
-Choice [y/n/e/r]:
+Choice [y/n/e/r/a]:
 ```
 
 **Interactive Options:**
 - **`y`** (or press Enter) - Accept the suggestion and commit
 - **`n`** - Reject and exit without committing
 - **`e`** - Edit the message manually with your own text
 - **`r`** - Regenerate a completely different suggestion using intelligent variation algorithms
+- **`a`** - **Upgrade to Local AI**: If you are using the heuristic engine, this attempts to connect to Ollama for a more semantic suggestion
+- **`h`** - **Fallback to Heuristic**: If you are in AI mode, this switches back to the classic rule-based engine
 
 ### Command-Line Options
 
@@ -237,6 +278,12 @@ Gitmit uses intelligent offline algorithms to analyze your changes:
     - Special case detection
     - Diversity algorithms for variations
 
+11. **Local AI Generation (Optional)** - When enabled:
+    - Serializes extracted context into a structured LLM prompt
+    - Dispatches generation requests to local Ollama daemon
+    - Validates AI output against Conventional Commit standards
+    - Provides deterministic fallback to heuristic engine on error
+
 ## Commit Types
 
 Gitmit supports the following commit types (automatically detected):

assets/assets.go

@@ -0,0 +1,45 @@
+package assets
+
+import (
+	"bytes"
+	"embed"
+	"text/template"
+)
+
+//go:embed prompts/* messages/*
+var Files embed.FS
+
+// GetPrompt returns the system prompt template
+func GetPrompt() (string, error) {
+	b, err := Files.ReadFile("prompts/system_prompt.txt")
+	return string(b), err
+}
+
+// GetOllamaWarning returns the Ollama warning message
+func GetOllamaWarning() (string, error) {
+	b, err := Files.ReadFile("messages/ollama_warning.txt")
+	return string(b), err
+}
+
+// GetInitSuccess returns the initialization success message
+func GetInitSuccess() (string, error) {
+	b, err := Files.ReadFile("messages/init_success.txt")
+	return string(b), err
+}
+
+// RenderOllamaWarning renders the Ollama warning message with the provided context
+func RenderOllamaWarning(url, model string) (string, error) {
+	warningTmpl, err := GetOllamaWarning()
+	if err != nil {
+		return "", err
+	}
+	tmpl, err := template.New("warning").Parse(warningTmpl)
+	if err != nil {
+		return "", err
+	}
+	var buf bytes.Buffer
+	if err := tmpl.Execute(&buf, struct{ URL, Model string }{URL: url, Model: model}); err != nil {
+		return "", err
+	}
+	return buf.String(), nil
+}

assets/messages/init_success.txt

@@ -0,0 +1,7 @@
+
+You can now customize the configuration to fit your project's needs.
+
+Configuration hierarchy:
+  1. Local (.gitmit.json) - project-specific settings
+  2. Global (~/.gitmit.json) - user-wide settings
+  3. Default (embedded) - built-in defaults

assets/messages/ollama_warning.txt

@@ -0,0 +1,7 @@
+⚠️  Ollama connection not detected on {{.URL}}
+To enable Local AI generation, please ensure:
+  1. Ollama is running locally (`ollama serve`)
+  2. The required model is pulled (`ollama pull {{.Model}}`)
+  3. Your .gitmit.json sets "engine": "ollama"
+
+Falling back to interactive options...

assets/prompts/system_prompt.txt

@@ -0,0 +1,18 @@
+You are an expert developer assistant. Analyze the provided structured git diff metadata and generate a single-line commit message following the Conventional Commits specification.
+
+Guidelines:
+1. Format MUST be: <type>(<scope>): <short description in present tense>
+2. Allowed types: feat, fix, refactor, chore, test, docs, style, perf, ci, build, security
+3. Do NOT include any markdown, backticks, quotes, or introductory text like "Here is your commit message:".
+4. Output ONLY the raw string of the commit message.
+
+Metadata Context:
+- Project Type: {{.ProjectType}}
+- Active Branch Name: {{.CurrentBranch}}
+- Detected Intent/Type Bonus: {{.RecommendedType}}
+- Modified Files: {{range .Files}}{{.}}, {{end}}
+- Key Code Symbols Altered: {{range .CodeSymbols}}{{.}}, {{end}}
+- Dependency Changes: {{.DependencyAlert}}
+- Added/Deleted Line Ratio: {{printf "%.2f" .DiffSummary.Ratio}}
+
+Output:

cmd/init.go

@@ -8,6 +8,7 @@ import (
 	"github.com/fatih/color"
 	"github.com/spf13/cobra"
 
+	"gitmit/assets"
 	"gitmit/internal/config"
 )
 
@@ -152,11 +153,9 @@ func runInit(cmd *cobra.Command, args []string) error {
 
 	color.Green("✅ Created config file: %s", configPath)
 	color.Blue("\n📝 Detected project type: %s", projectType)
-	fmt.Println("\nYou can now customize the configuration to fit your project's needs.")
-	fmt.Println("\nConfiguration hierarchy:")
-	fmt.Println("  1. Local (.gitmit.json) - project-specific settings")
-	fmt.Println("  2. Global (~/.gitmit.json) - user-wide settings")
-	fmt.Println("  3. Default (embedded) - built-in defaults")
+
+	msg, _ := assets.GetInitSuccess()
+	fmt.Println(msg)
 
 	return nil
 }