FAQ (Frequently Asked Questions)¶

Installation / Running¶

Q: What Python version do I need?¶

Python 3.9+. 3.12+ recommended.

Q: I installed via pip but `salmalm` command doesn't work¶

It's likely not on your PATH. Use python -m salmalm instead.

Q: Errors on Windows¶

Use cmd.exe (not PowerShell). Avoid Korean/non-ASCII characters in folder paths.

python -m salmalm

Q: How do I run with Docker?¶

git clone https://github.com/hyunjun6928-netizen/salmalm.git
cd salmalm
docker compose up -d
# → http://localhost:18800

API Keys¶

Q: Can I use it without an API key?¶

Yes. Install Ollama for a fully local, free LLM experience.

ollama pull llama3.2
python -m salmalm
# During onboarding, set Ollama URL: http://localhost:11434/v1

Q: Where do I get API keys?¶

Provider	URL
Anthropic (Claude)	https://console.anthropic.com/settings/keys
OpenAI (GPT)	https://platform.openai.com/api-keys
xAI (Grok)	https://console.x.ai
Google (Gemini)	https://aistudio.google.com/apikey
Brave Search	https://brave.com/search/api/ (free)

Q: Where are API keys stored?¶

Two options: 1. .env file — plaintext in project root (simple) 2. Vault — AES-256-GCM encrypted storage (secure)

.env takes priority over vault when both exist.

Password / Vault¶

Q: I forgot my password¶

Delete the vault file to reset: - Windows: del %USERPROFILE%\.salmalm\vault.enc - Linux/Mac: rm ~/.salmalm/vault.enc

Then run python -m salmalm to start fresh. ⚠️ This also deletes API keys stored in vault. Back them up in .env first.

Q: I don't want a password¶

Choose "No, start right away" during first-run setup. Already set one? Settings → 🔒 Master Password → "Remove Password"

Q: Can I add a password later?¶

Yes. Settings → 🔒 Master Password → Set New Password

Models / Usage¶

Q: Which model should I use?¶

General chat: Claude Sonnet or GPT-4.1 (best value)
Complex coding: Claude Opus or o3 (top performance)
Fast responses: Gemini Flash or GPT-4.1-nano (low cost)
Free: Ollama + llama3.2 (local)

Q: How do I switch models?¶

Web UI → Settings → Model dropdown, or type /model anthropic/claude-opus-4-6 in chat.

Q: How much does it cost?¶

Check Settings → Token Usage for real-time tracking. Typical usage: $0.5–$2/day.

Images¶

Q: How do I generate images?¶

Just ask naturally in chat: "Draw a cat". Uses DALL-E or xAI Aurora. Requires an OpenAI API key.

Q: How about image analysis?¶

Upload an image file or send a URL — the image_analyze tool handles it automatically. Uses GPT-4o or Claude Vision.

Troubleshooting¶

Q: "Connection refused" error¶

Make sure the server is running. Run python -m salmalm, then visit http://localhost:18800.

Q: Telegram bot not responding¶

Check .env for TELEGRAM_TOKEN and TELEGRAM_OWNER_ID
Another process polling with the same bot token will cause conflicts

Q: How do I update?¶

python -m pip install --upgrade salmalm
python -m salmalm

Or use Web UI → Settings → Update → Check for Updates

Q: Where are logs?¶

~/.salmalm/salmalm.log (or salmalm.log in the project directory)