feat: capture streaming token counts from SSE final chunk

Router now buffers streaming response chunks to extract timings
(prompt_n, predicted_n, predicted_per_second) from the final
SSE data frame before yielding to the client. Streaming requests
get real throughput data instead of 0 tok/s.

Uses llama.cpp timings field in the last content chunk:
- completion_tokens = predicted_n
- tokens_per_sec = predicted_per_second
- inference_ms = predicted_ms (generation only)

Client sees identical stream, no perceptible delay.

router/router.py

@@ -352,11 +352,34 @@ def chat():
         
         if resp.status_code != 200: return jsonify({"error":"GPU error "+str(resp.status_code)}), 502
         if is_stream:
-            def gen():
+            # Buffer stream to capture timings from final SSE chunk
+            chunks = []
+            stream_timings = {}
             for raw in resp.iter_content(chunk_size=None, decode_unicode=True):
-                    if raw: yield clean_unicode(raw)
-            # Streaming: can't get token counts without parsing stream, store latency + estimated tokens
+                if raw:
+                    cleaned = clean_unicode(raw)
+                    chunks.append(cleaned)
+                    # Parse last content chunk (before [DONE]) for timings
+                    if not stream_timings and '"timings"' in cleaned and '"predicted_n"' in cleaned:
+                        try:
+                            json_str = cleaned.replace("data: ", "").strip()
+                            if json_str.startswith("{"):
+                                tj = json.loads(json_str).get("timings", {})
+                                if tj:
+                                    stream_timings = tj
+                        except: pass
+            # Store perf record with real token counts from stream
+            if stream_timings:
+                pt = stream_timings.get("prompt_n", 0)
+                ct = stream_timings.get("predicted_n", 0)
+                tps = stream_timings.get("predicted_per_second", 0)
+                gen_ms = stream_timings.get("predicted_ms", lat)
+                store_perf_record(model, agent, tier, reason, queue_ms, gen_ms, pt, ct, True)
+            else:
                 store_perf_record(model, agent, tier, reason, queue_ms, lat, estimate_tokens(rd.get("messages",[])), 0, True)
+            # Yield all chunks to client
+            def gen():
+                for c in chunks: yield c
             bcast()
             ctx_remaining = GPU_CONTEXT.get(model, 65536) - max(session_tokens, estimate_tokens(rd.get("messages",[])))
             ctx_pct = ctx_remaining / GPU_CONTEXT.get(model, 65536) * 100