Summarize streams with real progress: adapters reassemble SSE/NDJSON deltas and report 0-99% as tokens arrive

backend/tests/test_ai_adapters.py

@@ -129,6 +129,32 @@ async def test_openai_compat_parses_and_sanitizes():
     assert res.model == "qwen"
 
 
+async def test_openai_compat_streams_with_progress():
+    """stream:true replies (SSE deltas) reassemble and tick progress."""
+    import json as _json
+
+    def sse(request: httpx.Request) -> httpx.Response:
+        payload = _json.loads(request.content.decode())
+        assert payload["stream"] is True
+        pieces = ['{"short": "Stand', 'up.", "detailed": "The team met."',
+                  ', "key_points": ["a"]', "}"]
+        lines = [f"data: {_json.dumps({'choices': [{'delta': {'content': p}}]})}"
+                 for p in pieces]
+        lines.append("data: [DONE]")
+        return httpx.Response(
+            200, content=("\n\n".join(lines) + "\n\n").encode(),
+            headers={"content-type": "text/event-stream"})
+
+    p = OpenAICompatProvider("http://llm:8000", model="qwen",
+                             http_client=mock_client(sse))
+    seen_pcts = []
+    res = await p.summarize("t", on_progress=seen_pcts.append)
+    assert res.short == "Standup."
+    assert res.key_points == ("a",)
+    assert seen_pcts and seen_pcts == sorted(seen_pcts)
+    assert all(0 <= p <= 99 for p in seen_pcts)  # never claims 100 early
+
+
 async def test_openai_compat_partial_json_gets_defaults():
     import json as _json
 
@@ -165,14 +191,21 @@ async def test_ollama_happy_path():
     async def ok(request: httpx.Request) -> httpx.Response:
         assert request.url.path == "/api/chat"
         payload = _json.loads(request.content.decode())
-        assert payload["format"] == "json" and payload["stream"] is False
-        return httpx.Response(200, json={
-            "message": {"content": _json.dumps({"short": "S.", "detailed": "D."})}
-        })
+        assert payload["format"] == "json" and payload["stream"] is True
+        body = "\n".join([
+            _json.dumps({"message": {"content": "{\"short\": \"S."}}),
+            _json.dumps({"message": {"content": "\", \"detailed\": \"D.\"}"}}),
+            _json.dumps({"done": True}),
+        ])
+        return httpx.Response(
+            200, json=None, content=body.encode(),
+            headers={"content-type": "application/x-ndjson"})
     p = OllamaProvider("http://ollama:11434", model="llama3",
                        http_client=mock_client(ok))
-    res = await p.summarize("meeting notes")
+    seen_pcts = []
+    res = await p.summarize("meeting notes", on_progress=seen_pcts.append)
     assert res.short == "S." and res.model == "llama3"
+    assert seen_pcts and seen_pcts == sorted(seen_pcts)  # monotonic progress
 
 
 async def test_ollama_404_is_config_error():

backend/tests/test_ai_pipeline.py

@@ -95,7 +95,8 @@ class FakeLlm:
         self.fail = fail
         self.seen = []
 
-    async def summarize(self, transcript, *, title=None, tone=None):
+    async def summarize(self, transcript, *, title=None, tone=None,
+                        on_progress=None):
         self.seen.append(transcript)
         if self.fail is not None:
             raise self.fail

backend/tests/test_transcript_edits.py

@@ -131,7 +131,8 @@ async def test_user_edit_survives_auto_pipeline(client, monkeypatch):
     class FakeLlm:
         name = "fake-llm"
 
-        async def summarize(self, transcript, *, title=None, tone=None):
+        async def summarize(self, transcript, *, title=None, tone=None,
+                            on_progress=None):
             return SummaryResult(short="s", detailed="d", key_points=(),
                                  decisions=(), action_items=(), questions=(),
                                  model="fake-llm-1")