Add KI/LLM integration with OpenAI-compatible API support

Features:
- /ai setup - Setup wizard via private chat for security
- /ai status - Show current config (API key masked)
- /ai on/off - Enable/disable AI per chat
- /ai random <1-10> - Set random trigger probability (admins)
- /ai help - Detailed help
- /ask-ai [text] - Generate AI response

Security:
- API keys only entered in private chat
- API keys masked in status display
- Admin-only configuration

Architecture:
- OpenAI-compatible client (OpenAI, Ollama, OpenRouter)
- Global system prompt (.env) + group prompt (DB)
- 50 message context per chat
- SQLite tables: ai_settings, message_context, setup_sessions

Defaults:
- AI disabled by default
- Trigger word: ask-ai
- Random probability: 1%

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

CLAUDE.md

@@ -4,7 +4,7 @@ This file provides guidance to Claude Code (claude.ai/code) when working with co
 
 ## Project Overview
 
-Telegram bot built with TypeScript and [Telegraf](https://telegraf.js.org/). Uses Markov chains with trigrams to generate sentences from learned messages. Each chat has its own isolated Markov chain, creating unique "personalities" per group.
+Telegram bot built with TypeScript and [Telegraf](https://telegraf.js.org/). Uses Markov chains with trigrams to generate sentences from learned messages. Optional KI/LLM integration via OpenAI-compatible API. Each chat has its own isolated data.
 
 ## Commands
 
@@ -19,75 +19,97 @@ docker compose up -d --build  # Docker deployment
 
 ```
 src/
-├── index.ts      # Bot entry point, commands, message handling
+├── index.ts      # Bot entry point, commands, setup wizard
 ├── markov.ts     # Markov chain with trigram support
-└── database.ts   # SQLite persistence layer
+├── database.ts   # SQLite persistence layer
+└── ai.ts         # OpenAI-compatible client
 ```
 
-**Data flow:**
+**Data flow (Markov):**
 1. Message received → `chain.learn(text)` → SQLite
 2. Trigger check (reply/mention/random)
 3. If triggered → `chain.generate()` → reply
 
+**Data flow (AI):**
+1. `/ask-ai` or trigger word detected
+2. Load last 50 messages as context
+3. Build prompt (global + group prompt)
+4. Call OpenAI-compatible API → reply
+
 **Per-chat isolation:**
 - Each chat has separate Markov chain data
-- Each chat has its own response probability setting
-- In-memory cache (`chains` Map) loaded lazily from DB
+- Each chat has separate AI settings
+- Each chat has separate message context (50 messages)
 
 ## Key Implementation Details
 
 ### Markov Chain (markov.ts)
 - Uses **trigrams** (order=2): "word1 word2" → "word3"
-- Better grammatical coherence than bigrams
 - Weighted random selection based on frequency
 - Generates sentences up to 20 words
 
 ### Database (database.ts)
 - SQLite with `better-sqlite3` (synchronous API)
-- Three tables: `chat_settings`, `markov_transitions`, `markov_starts`
-- Transactions for atomic updates
+- Tables:
+  - `chat_settings` - Markov probability
+  - `markov_transitions` - Word transitions
+  - `markov_starts` - Sentence starts
+  - `ai_settings` - KI configuration per chat
+  - `message_context` - Last 50 messages per chat
+  - `setup_sessions` - Setup wizard state
 - Automatic cleanup every 24h
 
-### Cleanup Mechanism
-```typescript
-const MAX_TRANSITIONS_PER_CHAT = 10000;  // Upper limit
-const MIN_TRANSITION_COUNT = 2;          // Remove rare entries
-const MIN_TRANSITIONS_TO_CLEANUP = 100;  // Protect new chats
-```
+### AI Client (ai.ts)
+- OpenAI-compatible API (works with OpenAI, Ollama, OpenRouter, etc.)
+- Prompt system: Global (.env) + Group-specific (DB)
+- API key masking for security
+- Context: Last 50 messages
 
-### Response Triggers
-1. Reply to bot's message → always respond
-2. @username mention → always respond
-3. Random probability (default 10%) in groups
-4. Always respond in private chats
+### Setup Wizard
+- Started via `/ai setup` in group
+- Continues in private chat for security
+- API keys never shown in full
+- Steps: Provider → API Key → Model → URL → Trigger/Prob → Group Prompt
 
 ## Environment
 
-- `BOT_TOKEN` - Telegram bot token (required)
-- Database: `data/ulfbot.db` (auto-created)
-
-## Docker
-
-- Uses `node:20-slim` for better-sqlite3 compatibility
-- Volume `ulfbot-data` mounts to `/app/data`
-- Build locally: `npm run build` then `docker compose up -d --build`
-
-## Common Modifications
-
-**Change response probability default:**
-```typescript
-// database.ts
-return { probability: (row?.probability ?? 10) };  // Change default
+```env
+BOT_TOKEN=                    # Telegram bot token (required)
+AI_DEFAULT_API_KEY=           # Default API key (optional)
+AI_DEFAULT_BASE_URL=          # Default API URL (default: OpenAI)
+AI_DEFAULT_MODEL=             # Default model (default: gpt-4o-mini)
+AI_SYSTEM_PROMPT=             # Global system prompt
 ```
 
-**Change cleanup interval:**
-```typescript
-// index.ts
-const CLEANUP_INTERVAL_MS = 24 * 60 * 60 * 1000;  // Currently 24h
+## Database Schema
+
+```sql
+-- Markov
+chat_settings (chat_id, probability)
+markov_transitions (chat_id, key, next_word, count)
+markov_starts (chat_id, key, count)
+
+-- AI
+ai_settings (chat_id, enabled, trigger_word, random_prob, group_prompt, provider, base_url, model, api_key)
+message_context (chat_id, role, content, timestamp)
+setup_sessions (user_id, chat_id, step, data)
 ```
 
-**Change Markov order (affects grammar quality):**
-```typescript
-// markov.ts
-constructor(order: number = 2)  // Higher = better grammar, needs more data
-```
+## Response Triggers
+
+### Markov (always active)
+1. Reply to bot's message → always respond
+2. @username mention → always respond
+3. Random probability (configurable per chat)
+
+### AI (when enabled)
+1. `/ask-ai [text]` command
+2. Trigger word (configurable, default: "ask-ai")
+3. Random probability (configurable, default: 0%)
+
+## Security
+
+- API keys stored in database, never logged
+- API keys masked in `/ai status` output: `sk-***...***xyz`
+- Setup only via private chat
+- Admin-only configuration commands