Add persian-tutor: Gradio-based GCSE Persian language learning app

Vocabulary study with FSRS spaced repetition, AI tutoring (Ollama/Claude),
essay marking, idioms browser, Anki export, and dashboard. 918 vocabulary
entries across 39 categories. 41 tests passing.

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

python/persian-tutor/ai.py

@@ -0,0 +1,44 @@
+"""Dual AI backend: Ollama (fast/local) and Claude CLI (smart)."""
+
+import subprocess
+
+import ollama
+
+DEFAULT_OLLAMA_MODEL = "qwen2.5:7b"
+
+
+def ask_ollama(prompt, system=None, model=DEFAULT_OLLAMA_MODEL):
+    """Query Ollama with an optional system prompt."""
+    messages = []
+    if system:
+        messages.append({"role": "system", "content": system})
+    messages.append({"role": "user", "content": prompt})
+    response = ollama.chat(model=model, messages=messages)
+    return response.message.content
+
+
+def ask_claude(prompt):
+    """Query Claude via the CLI subprocess."""
+    result = subprocess.run(
+        ["claude", "-p", prompt],
+        capture_output=True,
+        text=True,
+    )
+    return result.stdout.strip()
+
+
+def ask(prompt, system=None, quality="fast"):
+    """Unified interface. quality='fast' uses Ollama, 'smart' uses Claude."""
+    if quality == "smart":
+        return ask_claude(prompt)
+    return ask_ollama(prompt, system=system)
+
+
+def chat_ollama(messages, system=None, model=DEFAULT_OLLAMA_MODEL):
+    """Multi-turn conversation with Ollama."""
+    all_messages = []
+    if system:
+        all_messages.append({"role": "system", "content": system})
+    all_messages.extend(messages)
+    response = ollama.chat(model=model, messages=all_messages)
+    return response.message.content