Add offline analysis and client event tracking to Assistent API

- Introduced `/assistent/analyze-reply` endpoint for offline extraction of patches and client predictions without VRAM.
- Added `/assistent/client-event` and `/assistent/client-events` endpoints for tracking UI interactions and retrieving event history.
- Updated Debug API documentation to reflect new endpoints and their functionalities.
- Enhanced Assistent session handling with improved client prediction logic and diagnostics.

Co-authored-by: Cursor <cursoragent@cursor.com>

tests/test_assistent_session.py

@@ -4,7 +4,12 @@ from __future__ import annotations
 
 import time
 
-from gpu_rent.debug_assistent import extract_assistent_patch
+from gpu_rent.debug_assistent import (
+    analyze_assistent_reply,
+    extract_assistent_patch,
+    predict_client_turn,
+    user_asks_generate,
+)
 from gpu_rent import debug_assistent_session as das
 
 
@@ -33,6 +38,51 @@ def test_extract_assistent_patch_ignores_non_patch_json():
     assert out["patch"] is None
 
 
+def test_predict_client_turn_civitai_question_no_patch():
+    pred = predict_client_turn("а ты знаешь какие то промпты с civitai", None)
+    assert pred["user_asks_generate"] is False
+    assert pred["would_generate"] is False
+    assert pred["http_starts_generate"] is False
+    assert pred["generate_reason"] == "none"
+
+
+def test_predict_client_turn_model_flag_without_user_phrase():
+    pred = predict_client_turn(
+        "а ты знаешь какие то промпты с civitai",
+        {"prompt": "a redhead", "generate": True},
+        persona="neutral",
+    )
+    assert pred["user_asks_generate"] is False
+    assert pred["model_generate"] is True
+    assert pred["would_generate"] is False
+    assert pred["generate_reason"] == "model_flag_ignored"
+    assert pred["persona_chip"] == "Нормальный"
+
+
+def test_predict_client_turn_narisuy_with_prompt():
+    pred = predict_client_turn("нарисуй лису", {"prompt": "a fox"})
+    assert pred["user_asks_generate"] is True
+    assert pred["would_generate"] is True
+    assert pred["generate_reason"] == "user_phrase"
+    assert user_asks_generate("нарисуй лису") is True
+    assert user_asks_generate("Давай сделаем изображение рыжей") is True
+    assert user_asks_generate("что ты умеешь") is False
+
+
+def test_analyze_assistent_reply_offline():
+    out = analyze_assistent_reply(
+        message="сгенерируй",
+        reply='ok\n```json\n{"prompt":"a cat","generate":true}\n```\n',
+        persona="leonid",
+        pack="ordinary",
+    )
+    assert out["ok"] is True
+    assert out["offline"] is True
+    assert out["patch"]["generate"] is True
+    assert out["client"]["would_generate"] is True
+    assert out["client"]["generate_reason"] == "user_phrase"
+
+
 def test_analyze_exact_merge_fills_omitted_params():
     patch = {"actions": ["generate"], "prompt": "x"}
     exact = {

tests/test_debug_api.py

@@ -145,7 +145,49 @@ def test_debug_server_routes(monkeypatch, tmp_path):
         assert "/assistent/extension" in data["paths"]
         assert "/assistent/api" in data["paths"]
         assert "/assistent/chat-eval" in data["paths"]
+        assert "/assistent/analyze-reply" in data["paths"]
         assert "post" in data["paths"]["/assistent/chat-eval"]
+        assert "post" in data["paths"]["/assistent/analyze-reply"]
+
+        req = urllib.request.Request(
+            f"http://127.0.0.1:{port}/assistent/analyze-reply",
+            data=json.dumps(
+                {
+                    "message": "а ты знаешь какие то промпты с civitai",
+                    "reply": '```json\n{"prompt":"x","generate":true}\n```',
+                    "persona": "neutral",
+                }
+            ).encode("utf-8"),
+            headers={"Content-Type": "application/json"},
+            method="POST",
+        )
+        with urllib.request.urlopen(req, timeout=5) as resp:
+            scored = json.loads(resp.read().decode("utf-8"))
+        assert scored["ok"] is True
+        assert scored["client"]["would_generate"] is False
+        assert scored["client"]["user_asks_generate"] is False
+        assert scored["client"]["model_generate"] is True
+        assert scored["client"]["persona_chip"] == "Нормальный"
+
+        beacon = urllib.request.Request(
+            f"http://127.0.0.1:{port}/assistent/client-event",
+            data=json.dumps(
+                {
+                    "user": "а ты знаешь какие то промпты с civitai",
+                    "generating": False,
+                    "intent": {"generate": False, "modelAsked": True},
+                    "swarm_prompt": "old prompt",
+                }
+            ).encode("utf-8"),
+            headers={"Content-Type": "application/json"},
+            method="POST",
+        )
+        with urllib.request.urlopen(beacon, timeout=5) as resp:
+            stored = json.loads(resp.read().decode("utf-8"))
+        assert stored["ok"] is True
+        code, data = get("/assistent/client-events")
+        assert data["ok"] is True
+        assert data["events"][-1]["generating"] is False
 
         code, data = get("/assistent/chat-eval")
         assert "ok" in data