neovim_config/START_HERE.md

🎉 Setup Complete - Start Here!

What You Now Have

Your CodeCompanion is now configured to work with Ollama across your Tailscale network. This means:

• ✅ Use local Ollama on your main machine
• ✅ Access Ollama from other machines via Tailscale (no local Ollama needed)
• ✅ Switch between Claude and Ollama instantly
• ✅ Secure, encrypted connections via Tailscale

🚀 Get Started in 5 Minutes

Step 1: Configure Your Ollama Server (5 min)

On the machine running Ollama:

# Make Ollama accessible from network
sudo systemctl edit ollama

Add this line in the [Service] section:

Environment="OLLAMA_HOST=0.0.0.0:11434"

Save and exit, then:

sudo systemctl restart ollama

# Pull a model
ollama pull mistral

# Find your Tailscale IP
tailscale ip -4
# You'll see something like: 100.123.45.67

Step 2: Configure Other Machines (2 min)

On each machine that needs to access Ollama:

# Add to ~/.zshrc (or ~/.bashrc)
echo 'export OLLAMA_ENDPOINT="http://100.123.45.67:11434"' >> ~/.zshrc

# Reload shell
source ~/.zshrc

# Test it works
curl $OLLAMA_ENDPOINT/api/tags

Step 3: Use in Neovim (1 min)

" Start Neovim
nvim

" Press <leader>cll to chat with Ollama
" Type a message and press Enter
" You should get a response!

📚 Documentation

Start with these in order:

1. README_OLLAMA_INTEGRATION.md ← Read this first for overview
2. docs/QUICK_REFERENCE.md ← Quick reference card
3. docs/OLLAMA_SETUP.md ← Full setup guide
4. docs/TROUBLESHOOTING.md ← If something doesn't work

⌨️ Keymaps

Keymap	What It Does
<leader>cll	Chat with Ollama
<leader>cc	Chat with Claude Haiku
<leader>cs	Chat with Claude Sonnet
<leader>co	Chat with Claude Opus
<leader>ca	Show all CodeCompanion actions

🔧 What Was Changed

Modified

• lua/shelbybark/plugins/codecompanion.lua - Added Ollama adapter and keymaps

Created

• 8 comprehensive documentation files in docs/
• 1 main README file

🎯 Common Tasks

Pull a Different Model

ollama pull neural-chat
ollama pull llama2
ollama pull dolphin-mixtral

Change Default Model

Edit lua/shelbybark/plugins/codecompanion.lua line 40:

default = "neural-chat",  -- Change this

Test Connection

# Local
curl http://localhost:11434/api/tags

# Remote
curl http://100.x.x.x:11434/api/tags

List Available Models

ollama list

🆘 Something Not Working?

1. Check Ollama is running: ps aux | grep ollama
2. Check it's listening: sudo netstat -tlnp | grep 11434
3. Check Tailscale: tailscale status
4. Read troubleshooting: docs/TROUBLESHOOTING.md

📋 Checklist

• Ollama server configured with OLLAMA_HOST=0.0.0.0:11434
• Ollama restarted: sudo systemctl restart ollama
• Model pulled: ollama pull mistral
• Tailscale IP found: tailscale ip -4
• Environment variable set on other machines
• Shell reloaded: source ~/.zshrc
• Connection tested: curl $OLLAMA_ENDPOINT/api/tags
• Neovim tested: Press <leader>cll

💡 Pro Tips

1. Use mistral - Fast, good quality, recommended
2. Monitor latency - ping 100.x.x.x should be < 50ms
3. Keep Tailscale updated - Better performance
4. Use GPU if available - Much faster inference
5. Try smaller models - orca-mini for quick answers

📞 Need Help?

• Setup issues: See docs/OLLAMA_SETUP.md
• Troubleshooting: See docs/TROUBLESHOOTING.md
• Architecture: See docs/ARCHITECTURE.md
• Quick reference: See docs/QUICK_REFERENCE.md

🎓 How It Works (Simple Version)

Your Machine A (Ollama Server)
    ↓
    Ollama Service (localhost:11434)
    ↓
    Tailscale Network (Encrypted)
    ↓
Your Machine B (Laptop)
    ↓
    Neovim + CodeCompanion
    ↓
    Press <leader>cll
    ↓
    Chat with Ollama!

🔐 Security

• All traffic encrypted via Tailscale
• Uses private Tailscale IPs (100.x.x.x)
• Not exposed to public internet
• Secure end-to-end

🚀 Next Steps

1. ✅ Read README_OLLAMA_INTEGRATION.md
2. ✅ Follow the 5-minute setup above
3. ✅ Test with <leader>cll in Neovim
4. ✅ Enjoy local LLM access across your network!

---

Everything is ready to go!

Start with: README_OLLAMA_INTEGRATION.md

Questions?: Check docs/QUICK_REFERENCE.md

Issues?: Check docs/TROUBLESHOOTING.md

---

Date: 2026-02-05 Status: ✅ Ready to Use