feat(llm): direct Claude Haiku 4.5 backend with prompt caching

Adds a parallel LLM backend that bypasses OpenClaw and talks to
Anthropic Messages API directly. Selected via LLM_BACKEND=claude in
.env; default remains openclaw so nothing breaks for existing setup.

Why: OpenClaw gateway adds 500-1000ms overhead on every turn (auth,
memory fetch, routing). Direct Haiku 4.5 + prompt caching = faster
first token and -90% cost on cached chunks.

- satellite/llm_claude.py — Anthropic SDK streaming client, prompt
  caching on system prompt and all-but-last-2 history messages, per
  agent+date JSON history in HISTORY_DIR, reset_history() for the
  'сбрось' command, per-agent system prompts (Cosmo / Люся), fallback
  to error event if SDK/key missing.
- satellite/llm.py — dispatches to ask_claude_stream when backend=claude,
  exports LLM_BACKEND so modes.py can route reset too.
- satellite/modes.py — _handle_reset calls reset_history when backend
  is claude, keeps /new POST for openclaw.
- requirements.txt — anthropic >= 0.50.0
- .env.example — LLM_BACKEND, ANTHROPIC_API_KEY, ANTHROPIC_MODEL,
  HISTORY_DIR, MAX_HISTORY, HTTPS_PROXY block for non-RU egress.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

satellite/modes.py

@@ -3,34 +3,42 @@ import os
 from .config import GATEWAY_URL, AGENTS, FOLLOWUP_TIMEOUT, MAX_DURATION, log
 from .audio import record
 from .tts import speak, stop_speaking
-from .llm import ask_agent_stream, is_reset_command, VOICE_SESSION_KEY
+from .llm import ask_agent_stream, is_reset_command, VOICE_SESSION_KEY, LLM_BACKEND
 from . import notifier
 
 WAKE_THRESHOLD = float(os.getenv("WAKE_THRESHOLD", "0.5"))
 
 
 def _handle_reset(text: str, agent_id: str) -> bool:
-    """Команда сброса — отправляет slash-команду /new в OpenClaw (без озвучки ответа)."""
+    """Команда сброса. В зависимости от backend:
+    - claude:    удаляет локальный файл истории
+    - openclaw:  шлёт /new в gateway
+    """
     if not is_reset_command(text):
         return False
 
-    cfg = AGENTS.get(agent_id, AGENTS["cosmo"])
-    print("🔄 Отправляю /new в OpenClaw")
-    try:
-        cfg["session"].post(
-            f"{cfg['gateway_url']}/v1/chat/completions",
-            headers={
-                "x-ocplatform-model": cfg["voice_model"],
-                "x-openclaw-session-key": cfg.get("session_key", VOICE_SESSION_KEY),
-            },
-            json={
-                "stream": False,
-                "messages": [{"role": "user", "content": "/new"}],
-            },
-            timeout=30,
-        )
-    except Exception:
-        log.exception("Не удалось отправить /new")
+    if LLM_BACKEND == "claude":
+        from .llm_claude import reset_history
+        print("🔄 Сбрасываю локальную историю (Claude)")
+        reset_history(agent_id)
+    else:
+        cfg = AGENTS.get(agent_id, AGENTS["cosmo"])
+        print("🔄 Отправляю /new в OpenClaw")
+        try:
+            cfg["session"].post(
+                f"{cfg['gateway_url']}/v1/chat/completions",
+                headers={
+                    "x-ocplatform-model": cfg["voice_model"],
+                    "x-openclaw-session-key": cfg.get("session_key", VOICE_SESSION_KEY),
+                },
+                json={
+                    "stream": False,
+                    "messages": [{"role": "user", "content": "/new"}],
+                },
+                timeout=30,
+            )
+        except Exception:
+            log.exception("Не удалось отправить /new")
 
     msg = "Начинаю новую сессию."
     print(f"🔄 {msg}")