fix(ai): request final responses from Qwen3.5

docs/infrastructure/2026-10-20-qwen35-compatibility.md

@@ -0,0 +1,9 @@
+# Qwen3.5 compatibility preparation
+
+The owner requested `qwen3.5:9b-q4_K_M` after live deployment. The model is downloaded, but the active sidecar still uses the previous model. This change prepares the existing adapter; it does not activate a model or qualify the withheld P5-13 capability.
+
+The existing adapter consumes Ollama's final `response`. A synthetic local probe found an empty response with the Qwen3.5 thinking default; a20October probe with `think:false` returned the requested synthetic JSON and no thinking text. The adapter now sets that request field only when the configured base family is exactly `qwen3.5`. Existing temperature, JSON mode, timeouts, response parsing, other model payloads and external-provider consent/fallback remain unchanged. Custom namespace aliases are outside this exact family match.
+
+Seven new offline cases verify both text/JSON for the requested tag and bare family, and unchanged payloads for qwen2.5, qwen3 and a lookalike family. Four Qwen3.5 cases failed ordinarily before the two-line change; three compatibility cases passed. The full sidecar suite then passed **56/56**, zero failures/errors/skips,8.45seconds,20October13:32UTC. The prior49-case baseline grows by seven; no model calls are part of these tests. Raw RED/GREEN logs and JUnit are retained under `.artifacts/release-20261020/model-compat/`.
+
+Live activation, deployed adapter identity, GPU residency and final-response smoke remain pending behind the release. P5-13's deployed-identity/366-call/independent-human-review/UI requirements remain unmet and its review feature stays disabled. Production readiness remains NO-GO independently of the owner's release authorization.

tools/summarizer/app.py

@@ -646,6 +646,8 @@ def _ollama_generate(prompt: str, *, json_mode: bool, temperature: float, timeou
         "stream": False,
         "options": {"temperature": temperature},
     }
+    if OLLAMA_MODEL.split(":", 1)[0] == "qwen3.5":
+        payload["think"] = False
     if json_mode:
         payload["format"] = "json"
     body = _http_post_json(f"{OLLAMA_BASE_URL}/api/generate", payload, {}, timeout)

tools/summarizer/tests/test_app.py

@@ -8,6 +8,7 @@ import textwrap
 import zipfile
 from pathlib import Path
 
+import pytest
 from docx import Document
 from fastapi.testclient import TestClient
 from PIL import Image
@@ -253,6 +254,55 @@ def test_provider_defaults_to_ollama_and_is_unchanged(monkeypatch):
     assert captured["body"]["options"]["temperature"] == 0.1
 
 
+@pytest.mark.parametrize("ollama_model", ["qwen3.5:9b-q4_K_M", "qwen3.5"])
+@pytest.mark.parametrize("json_mode", [False, True])
+def test_qwen35_requests_nonthinking_output_without_consuming_thinking(monkeypatch, ollama_model, json_mode):
+    module = load_app_module(monkeypatch, ollama_model=ollama_model)
+    captured = {}
+    output = '{"score": 7}' if json_mode else "Rewritten CV text"
+
+    def fake_urlopen(req, timeout=None):
+        captured["body"] = json.loads(req.data.decode("utf-8"))
+        captured["timeout"] = timeout
+        captured["url"] = req.full_url
+        return _FakeResponse({
+            "response": output if captured["body"].get("think") is False else "",
+            "thinking": "Private reasoning must never become candidate text.",
+        })
+
+    monkeypatch.setattr(module.urllib_request, "urlopen", fake_urlopen)
+    result = module._ollama_generate_json("synthetic prompt") if json_mode else module._ollama_generate_text("synthetic prompt")
+
+    assert result == ({"score": 7} if json_mode else output)
+    expected = {
+        "model": ollama_model,
+        "prompt": "synthetic prompt",
+        "stream": False,
+        "options": {"temperature": 0.1 if json_mode else 0.2},
+        "think": False,
+    }
+    if json_mode:
+        expected["format"] = "json"
+    assert captured["body"] == expected
+    assert captured["url"] == f"{module.OLLAMA_BASE_URL}/api/generate"
+    assert captured["timeout"] == (120 if json_mode else 180)
+
+
+@pytest.mark.parametrize("ollama_model", ["qwen2.5:7b", "qwen3:8b", "qwen3.5-plus:9b"])
+def test_other_ollama_model_payloads_remain_unchanged(monkeypatch, ollama_model):
+    module = load_app_module(monkeypatch, ollama_model=ollama_model)
+    captured = {}
+    _install_fake_urlopen(monkeypatch, module, {"response": "  Candidate text  ", "thinking": "Reasoning"}, captured)
+
+    assert module._ollama_generate_text("synthetic prompt") == "Candidate text"
+    assert captured["body"] == {
+        "model": ollama_model,
+        "prompt": "synthetic prompt",
+        "stream": False,
+        "options": {"temperature": 0.2},
+    }
+
+
 def test_local_success_wins_even_when_external_fallback_is_permitted(monkeypatch):
     monkeypatch.setenv("AI_PROVIDER", "gemini")
     monkeypatch.setenv("GEMINI_API_KEY", "test-key")