Reset branch

.gitignore

@@ -18,3 +18,13 @@ undefined_syms_auto.*.txt
 build.ninja
 objdiff.json
 /permute*
+
+automation/functions.db
+automation/agent.log
+automation/agent_summary.json
+automation/__pycache__/
+automation/compile_errors_report.json
+automation/CONTINUE_SETUP.md
+automation/HANDOFF.md
+automation/QUICK_START_TYPE_LEARNING.md
+automation/TYPE_LEARNING.md

automation/CONTINUE_SETUP.md

@@ -0,0 +1,223 @@
+# Continue + Qwen Setup Guide
+
+## ✅ Setup Complete!
+
+Your system is configured to use **Continue** with **Qwen2.5-Coder 7B** locally via Ollama.
+
+## What's Installed
+
+- ✅ **Ollama** v0.17.4 (`/usr/local/bin/ollama`)
+- ✅ **qwen2.5-coder:7b** model (4.7 GB)
+- ✅ **Continue config** at `~/.continue/config.json`
+
+## Quick Start
+
+### 1. Install Continue Extension (If Not Already)
+
+If Continue isn't installed in VS Code:
+1. Open VS Code
+2. Go to Extensions (Ctrl+Shift+X)
+3. Search for "Continue"
+4. Install "Continue - Codestral, Claude, and more"
+5. Reload VS Code
+
+### 2. Start Ollama Service
+
+Ollama needs to be running in the background:
+
+```bash
+# Start ollama service (if not already running)
+ollama serve
+```
+
+Or better yet, run it in the background:
+
+```bash
+# Check if ollama is running
+pgrep ollama || ollama serve &
+```
+
+### 3. Using Continue
+
+Once Continue is installed and Ollama is running:
+
+**Open Continue Panel:**
+- Click the Continue icon in the left sidebar (or Ctrl+L)
+
+**Chat with Qwen:**
+- Type questions in the Continue chat
+- Select code and ask questions about it
+- Use @ to reference files/symbols
+
+**Custom Commands (for decompilation):**
+
+We've added special commands for your decompilation workflow:
+
+1. **`/explain-asm`** - Analyze MIPS assembly
+   - Select assembly code
+   - Type `/explain-asm`
+   - Get detailed explanation
+
+2. **`/decompile-hint`** - Get decompilation suggestions
+   - Select assembly function
+   - Type `/decompile-hint`
+   - Get C equivalent suggestions
+
+3. **`/review-decomp`** - Review decompiled C code
+   - Select your C code
+   - Type `/review-decomp`
+   - Get accuracy check
+
+**Code Completion:**
+- Just start typing - Qwen will suggest completions
+- Press Tab to accept
+
+## Example Workflow
+
+### Decompile a Function with AI Help
+
+1. **Find a function to decompile:**
+   ```bash
+   cd ~/GitHub/ff7-decomp-armstrca/automation
+   python3 discover.py --list-todo | head -10
+   ```
+
+2. **Get the assembly:**
+   Look at the corresponding .s file in `asm/us/`
+
+3. **Ask Qwen for help:**
+   - Open the .s file in VS Code
+   - Select the assembly
+   - Open Continue (Ctrl+L)
+   - Type: `/decompile-hint`
+
+4. **Implement the C code:**
+   - Use Qwen's suggestions
+   - Write C in the appropriate src/ file
+
+5. **Run decompilation:**
+   ```bash
+   python3 decompile.py --function func_name
+   ```
+
+6. **Verify it matches:**
+   ```bash
+   make build
+   # Check for matching
+   ```
+
+## Configuration Details
+
+Your Continue config (`~/.continue/config.json`) includes:
+
+- **Main model**: Qwen2.5-Coder 7B (32K context)
+- **Autocomplete**: Optimized for fast suggestions
+- **Temperature**: 0.2 (balanced creativity/accuracy)
+- **Custom commands**: Decompilation-specific prompts
+
+## Performance Expectations
+
+With your RTX 2060 (6GB VRAM) and i5-9300H:
+
+- **Chat responses**: ~20-40 tokens/sec
+- **Autocomplete**: Nearly instant (<100ms)
+- **Context window**: 32K tokens (~24K lines of code)
+- **Model size**: 4.7 GB (fits in VRAM + RAM)
+
+## Troubleshooting
+
+### Ollama not responding:
+```bash
+# Restart ollama service
+pkill ollama
+ollama serve &
+```
+
+### Model not loading:
+```bash
+# Verify model is installed
+ollama list
+
+# Pull again if needed
+ollama pull qwen2.5-coder:7b
+```
+
+### Continue can't connect:
+- Check Ollama is running: `pgrep ollama`
+- Verify port 11434 is open: `curl http://localhost:11434`
+- Restart VS Code
+
+### Slow responses:
+- Close other applications to free RAM
+- Use smaller batch sizes in automation
+- Consider running only Ollama during heavy AI use
+
+## Advanced: Running Ollama as a Service
+
+To auto-start Ollama on boot:
+
+```bash
+# Create systemd service
+sudo tee /etc/systemd/system/ollama.service > /dev/null <<EOF
+[Unit]
+Description=Ollama Service
+After=network.target
+
+[Service]
+ExecStart=/usr/local/bin/ollama serve
+User=$USER
+Restart=always
+
+[Install]
+WantedBy=multi-user.target
+EOF
+
+# Enable and start
+sudo systemctl enable ollama
+sudo systemctl start ollama
+
+# Check status
+sudo systemctl status ollama
+```
+
+## Useful Commands
+
+```bash
+# List all models
+ollama list
+
+# Remove a model
+ollama rm model-name
+
+# Pull additional models (if you want variety)
+ollama pull codellama:7b
+ollama pull deepseek-coder:6.7b
+
+# Check Ollama logs
+journalctl -u ollama -f
+
+# Test model directly
+ollama run qwen2.5-coder:7b "Explain MIPS assembly: lui $v0, 0x8001"
+```
+
+## Integration with Automation
+
+You can enhance the automation scripts to use Qwen:
+
+**For Chunk 5 (LLM Integration)**, we'll add:
+- Automatic assembly analysis before decompilation
+- Refinement suggestions for non-matching functions
+- Code quality checks on decompiled C
+
+This will be in the next chunk!
+
+## Resources
+
+- [Continue Documentation](https://continue.dev/docs)
+- [Ollama Models](https://ollama.ai/library)
+- [Qwen2.5-Coder](https://huggingface.co/Qwen/Qwen2.5-Coder-7B-Instruct)
+- [MIPS Assembly Reference](https://www.mips.com/products/architectures/mips32-2/)
+
+---
+
+**You're all set!** Continue is configured and ready to help with decompilation. Open VS Code and try it out!