feat(i18n): inject user language into AI agent system prompts

- Add _language_instruction() to deep_agent.py, reads language from core memory
- Append language directive to all 4 run_* functions (task/project/checkpoint/note)
- Minor fixes: alembic env, route imports, test cleanup

app/core/agent_runner.py

@@ -30,7 +30,6 @@ import asyncio
 import json
 import logging
 import os
-import uuid
 from datetime import datetime, timedelta, timezone
 from typing import Any
 

app/core/deep_agent.py

@@ -27,6 +27,34 @@ logger = logging.getLogger(__name__)
 FloatingDomainType = Literal["task", "timeline", "project", "node"]
 FloatingDomainSection = Literal["task", "timeline", "note"]
 
+# Mapping of core-memory language values to natural-language names for prompts.
+_LANGUAGE_NAMES: dict[str, str] = {
+    "en": "English", "it": "Italian", "es": "Spanish",
+    "fr": "French", "de": "German",
+    "english": "English", "italian": "Italian", "italiano": "Italian",
+    "spanish": "Spanish", "español": "Spanish",
+    "french": "French", "français": "French",
+    "german": "German", "deutsch": "German",
+}
+
+
+def _language_instruction(context: dict[str, Any]) -> str:
+    """Return a system-prompt suffix that tells the LLM to respond in the user's language.
+
+    Returns an empty string when the language is English or unknown — saves tokens.
+    """
+    core = context.get("core_memory") or {}
+    raw = (core.get("language") or "").strip().lower()
+    if not raw:
+        return ""
+    lang = _LANGUAGE_NAMES.get(raw, raw.title())  # best-effort capitalisation
+    if lang.lower() == "english":
+        return ""
+    return (
+        f"\n\nIMPORTANT: Always respond in {lang}. "
+        f"All your output text must be written in {lang}."
+    )
+
 _HOME_SYSTEM_PROMPT = (
     "You are the home assistant with direct access to all tools: tasks, projects, notes, timelines, and memory tools. "
     "Always use tools for factual data retrieval before answering. "
@@ -876,6 +904,7 @@ async def run_home(user_id: str, message: str, context: dict[str, Any]) -> str:
     system_prompt, langfuse_prompt = get_prompt_or_fallback(
         "home_system", _HOME_SYSTEM_PROMPT
     )
+    system_prompt += _language_instruction(context)
     response = await _run_single_agent(
         user_id=user_id,
         system_prompt=system_prompt,
@@ -893,6 +922,7 @@ async def run_floating(user_id: str, message: str, context: dict[str, Any]) -> t
     system_prompt, langfuse_prompt = get_prompt_or_fallback(
         "floating_system", _FLOATING_SYSTEM_PROMPT
     )
+    system_prompt += _language_instruction(context)
     response = await _run_single_agent(
         user_id=user_id,
         system_prompt=system_prompt,
@@ -916,6 +946,7 @@ async def run_home_stream(
     system_prompt, langfuse_prompt = get_prompt_or_fallback(
         "home_system", _HOME_SYSTEM_PROMPT
     )
+    system_prompt += _language_instruction(context)
     text_chunks: list[str] = []
     async for event in _run_single_agent_stream(
         user_id=user_id,
@@ -948,6 +979,7 @@ async def run_floating_stream(
     system_prompt, langfuse_prompt = get_prompt_or_fallback(
         "floating_system", _FLOATING_SYSTEM_PROMPT
     )
+    system_prompt += _language_instruction(context)
     sanitizer = _FloatingStreamSanitizer()
     emitted_sanitized = False
     raw_chunks: list[str] = []