feat(bot): tool-using chat agent over RAG, ЕМИАС record and drug-safety reference

Free-text treatment and medication questions go to an LLM agent with read-only tools bound to the current
patient: guideline search (RAG, doses stripped), ЕМИАС record mock (diagnoses, comorbidities, intolerances,
other specialists' drugs, current prescription), drug info (composition, contraindications, interactions)
and a deterministic patient-specific check. Providers: OpenRouter (temporary, synthetic data only) and an
offline mock; GigaChat to follow. Scrubbed input only, forbidden-output filter, any failure falls back to
the existing extractive RAG / instruction answers. Off unless OPENROUTER_API_KEY or CHAT_AGENT=mock.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

app/config.py

@@ -28,6 +28,16 @@ class Settings(BaseSettings):
     gigachat_model: str = "GigaChat"
     gigachat_verify_ssl: bool = False
 
+    # Chat agent (tool-using LLM over RAG + patient record). `auto` = OpenRouter if a key is set, else off
+    # (extractive RAG / instruction answers stay). OpenRouter is a TEMPORARY prototype provider on synthetic
+    # data only — prod must use GigaChat (152-FZ, foreign services are forbidden for real PII).
+    chat_agent: Literal["auto", "off", "mock", "openrouter", "gigachat"] = "auto"
+    openrouter_api_key: str = ""
+    openrouter_model: str = "anthropic/claude-sonnet-4.5"
+    openrouter_base_url: str = "https://openrouter.ai/api/v1"
+    chat_agent_timeout_s: float = 25.0
+    chat_agent_max_steps: int = 5
+
     # Telegram is used only for the prototype on synthetic data; prod channel is MAX / Gosuslugi.
     tg_bot_token: str = ""
     doctor_notify_chat_id: int | None = None