feat(intern-decision): stack, DNS and GPU 3 acceptance for the SemIf replacement

stacks/intern-decision: compose (GPU 1, :8033, hard VRAM cap as the single .env knob,
healthcheck, Homepage group 'AI - Eval & Retrieval'), .env.example and README.
dns: intern-decision.fv.internal -> fv-ml1 (synced to ana/esh/nh3).
acceptance on fv-ml1 GPU 3, 3 fresh processes: bit-identical to the Jev bench's native rows
(pooled 240/259, Wyrd 79/84, 0/560 flips, Δp 0), negative control 10/122/14, 0 flips across
restarts; largest accepted request 200 at a 10,134 MiB card peak under a 9.25 GiB cap; 503 and
recovery proven at a tight cap. GPU 1 deploy held: nvidia-smi Free on GPU 1 is 15,442 MiB.

services/intern-decision-serve/acceptance/gpu3-2026-09-30/cap9.0/checks.json

@@ -0,0 +1,142 @@
+{
+ "url": "http://127.0.0.1:18033",
+ "started_utc": "2026-09-30T16:36:04Z",
+ "health": {
+  "status": "ok",
+  "model": {
+   "name": "Intern-Decision-4B",
+   "source": "internlm/Intern-Decision-4B",
+   "revision": "0e5e6aa7d6d750e2b1504ba11a8136cb58aeb3cd",
+   "checkpoint": "/hf/hub/models--internlm--Intern-Decision-4B/snapshots/0e5e6aa7d6d750e2b1504ba11a8136cb58aeb3cd",
+   "inference_py_sha256": "c904e2c67ca0775621a22375ee373d2ba30b52117cda870c6c9ef74143b29863",
+   "temperature": 1.99241824,
+   "dtype": "bfloat16",
+   "attn_implementation": "sdpa",
+   "device": "cuda",
+   "max_length": 8192,
+   "torch_version": "2.10.0+cu128",
+   "transformers_version": "5.17.0",
+   "vision_tower": "removed",
+   "device_name": "NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition",
+   "allocated_gib": 7.937,
+   "reserved_gib": 7.969,
+   "max_reserved_gib": 8.861
+  },
+  "vram_cap_gib": 9.0,
+  "max_tokens": 8192,
+  "max_decisions": 64,
+  "max_questions_per_call": 16,
+  "chunking": "/decide/shared questions are packed greedily, in request order, into calls of at most 16 (1-16, 17-32, ...); each call is one prompt, so the questions in a call are asked together. With orderings, ordering k of every decision forms wave k, packed the same way.",
+  "workloads": []
+ },
+ "fit": {
+  "largest_ok_state_chars": 4793,
+  "largest_ok_tokens": 7168,
+  "probes": [
+   [
+    31950,
+    422,
+    null
+   ],
+   [
+    16475,
+    422,
+    null
+   ],
+   [
+    8737,
+    503,
+    null
+   ],
+   [
+    4868,
+    503,
+    null
+   ],
+   [
+    2934,
+    200,
+    6772
+   ],
+   [
+    3901,
+    200,
+    6964
+   ],
+   [
+    4384,
+    200,
+    7056
+   ],
+   [
+    4626,
+    200,
+    7133
+   ],
+   [
+    4747,
+    200,
+    7156
+   ],
+   [
+    4807,
+    503,
+    null
+   ],
+   [
+    4777,
+    200,
+    7163
+   ],
+   [
+    4792,
+    200,
+    7168
+   ],
+   [
+    4799,
+    503,
+    null
+   ],
+   [
+    4795,
+    503,
+    null
+   ],
+   [
+    4793,
+    200,
+    7168
+   ],
+   [
+    4794,
+    503,
+    null
+   ]
+  ],
+  "t_start": 1790786164.5822356,
+  "t_end": 1790786175.4811816
+ },
+ "maxreq": {
+  "state_chars": 9508,
+  "tokens_per_call": null,
+  "decisions": 64,
+  "options_each": 16,
+  "body_bytes": 101019,
+  "runs": [
+   {
+    "status": 503,
+    "e2e_ms": 209.5,
+    "code": "out_of_memory",
+    "calls": null,
+    "input_tokens": null,
+    "reserved_gib_after": 7.969,
+    "max_reserved_gib": 9.0,
+    "t_end": 1790786178.99772
+   }
+  ],
+  "t_start": 1790786175.481284,
+  "t_end": 1790786178.9977298
+ },
+ "finished_utc": "2026-09-30T16:36:18Z"
+}

services/intern-decision-serve/acceptance/gpu3-2026-09-30/cap9.0/shape.json

@@ -0,0 +1 @@
+{"label": "cap9.0", "backend": "semif", "runs": 1, "per": 5, "max_questions": 10000, "events": [["start", 1790786179.0584767], ["short1:measure", 1790786179.1493392], ["crit21:measure", 1790786179.5161796], ["long1:measure", 1790786186.5989966], ["long16:measure", 1790786194.1815286], ["shapes:done", 1790786195.251973], ["end", 1790786195.251976]], "shapes": {"short1": {"rows": 1, "runs": [{"e2e_ms": [37.941707065328956, 38.23350998573005, 37.37540193833411, 38.53617515414953, 38.11007016338408], "server_ms": [35.29082587920129, 35.39266809821129, 34.68621987849474, 35.76885210350156, 35.205855034291744], "statuses": [200, 200, 200, 200, 200], "tokens": 181, "median_e2e": 38.11007016338408, "median_server": 35.29082587920129}], "tokens": 181, "median_e2e_ms": 38.1, "run_medians_e2e_ms": [38.1], "median_server_ms": 35.3, "non_200": 0}, "crit21": {"rows": 21, "runs": [{"e2e_ms": [86.08735189773142, 85.76549892313778, 85.83429013378918, 85.84657986648381, 85.68961801938713], "server_ms": [79.76265903562307, 79.48806718923151, 79.50853602960706, 79.57211695611477, 79.62968898937106], "statuses": [200, 200, 200, 200, 200], "tokens": null, "median_e2e": 85.83429013378918, "median_server": 79.57211695611477}], "tokens": null, "median_e2e_ms": 85.8, "run_medians_e2e_ms": [85.8], "median_server_ms": 79.6, "non_200": 0}, "long1": {"rows": 1, "runs": [{"e2e_ms": [184.56048821099102, 191.5509181562811, 187.36126995645463, 186.49141117930412, 187.67126300372183], "server_ms": [181.11044797115028, 187.99379700794816, 183.76171891577542, 183.00863006152213, 184.1394929215312], "statuses": [200, 200, 200, 200, 200], "tokens": 4019, "median_e2e": 187.36126995645463, "median_server": 183.76171891577542}], "tokens": 4019, "median_e2e_ms": 187.4, "run_medians_e2e_ms": [187.4], "median_server_ms": 183.8, "non_200": 0}, "long16": {"rows": 16, "runs": [{"e2e_ms": [211.36997686699033, 217.9243720602244, 213.36712990887463, 213.6664930731058, 211.49204880930483], "server_ms": [205.53427003324032, 212.03652513213456, 207.65956514514983, 207.81763619743288, 205.71005111560225], "statuses": [200, 200, 200, 200, 200], "tokens": null, "median_e2e": 213.36712990887463, "median_server": 207.65956514514983}], "tokens": null, "median_e2e_ms": 213.4, "run_medians_e2e_ms": [213.4], "median_server_ms": 207.7, "non_200": 0}}}

services/intern-decision-serve/acceptance/gpu3-2026-09-30/cap9.0/vram.csv

@@ -0,0 +1,294 @@
+# pid=1899253 gpu=GPU-186dacf4-f447-ef03-ffc3-12ab28e1b8ff
+1790786164.716320980,8812,8822
+1790786164.823793359,8812,8822
+1790786164.929302716,8812,8822
+1790786165.035433570,8812,9408
+1790786165.148395222,9398,9664
+1790786165.259687015,9654,9664
+1790786165.372175832,9654,9664
+1790786165.484222484,9654,9664
+1790786165.590480950,9654,9664
+1790786165.709071847,9654,9664
+1790786165.813055136,9654,9664
+1790786165.915808311,9654,9664
+1790786166.018890279,9654,9664
+1790786166.129017919,9654,9664
+1790786166.229606519,9654,9664
+1790786166.335587621,9654,9664
+1790786166.443510165,9654,9664
+1790786166.547975000,9654,9664
+1790786166.655635241,9654,9664
+1790786166.761522382,9654,9664
+1790786166.867852718,9654,9664
+1790786166.977569563,9654,9664
+1790786167.085987964,9654,9664
+1790786167.196082253,9654,9664
+1790786167.305719237,9654,9664
+1790786167.416458704,9654,9664
+1790786167.523194095,9654,9664
+1790786167.632373654,9654,9664
+1790786167.731704019,9654,9664
+1790786167.840186040,9654,9664
+1790786167.944049348,9654,9664
+1790786168.044211153,9654,9664
+1790786168.145394139,9654,9664
+1790786168.253804889,9654,9664
+1790786168.356489443,9654,9664
+1790786168.471620971,9654,9664
+1790786168.569846013,9654,9664
+1790786168.683211431,9654,9664
+1790786168.785017705,9654,9664
+1790786168.896854124,9654,9664
+1790786169.000832373,9654,9664
+1790786169.117628050,9654,9664
+1790786169.217906256,9654,9664
+1790786169.322926357,9654,9664
+1790786169.431459258,9654,9664
+1790786169.537479991,9654,9664
+1790786169.645614758,9654,9664
+1790786169.755306653,9654,9664
+1790786169.861792561,9654,9664
+1790786169.981926806,9654,9664
+1790786170.081930029,9654,9664
+1790786170.197530461,9654,9664
+1790786170.301479580,9654,9664
+1790786170.417932963,9654,9664
+1790786170.519371663,9654,9664
+1790786170.630502444,9654,9664
+1790786170.737537438,9654,9664
+1790786170.845572454,9654,9664
+1790786170.948239538,9654,9664
+1790786171.060100188,9654,9664
+1790786171.168635179,9654,9664
+1790786171.272443606,9654,9664
+1790786171.379895525,9654,9664
+1790786171.483702352,9846,9842
+1790786171.585724958,9832,9842
+1790786171.688623185,9832,8824
+1790786171.792108018,9828,9838
+1790786171.893702910,9828,8824
+1790786172.000577552,9848,9858
+1790786172.107827978,9848,9858
+1790786172.210922127,9848,8824
+1790786172.320767474,9852,9866
+1790786172.427548455,9856,9866
+1790786172.535807333,9856,9866
+1790786172.641415401,9584,9876
+1790786172.749858882,9866,9876
+1790786172.857379212,9866,9876
+1790786172.961890846,8814,9876
+1790786173.068642918,9866,9876
+1790786173.177382221,9866,9876
+1790786173.283539615,9866,9370
+1790786173.384837343,9866,9876
+1790786173.489906035,9866,9876
+1790786173.596368152,9866,8826
+1790786173.696510917,9828,9838
+1790786173.800309834,9828,9444
+1790786173.908233539,9586,9876
+1790786174.017859623,9866,9876
+1790786174.124053877,9866,9876
+1790786174.229679985,8814,9876
+1790786174.333861606,9838,9848
+1790786174.437805694,9838,9848
+1790786174.543411892,9838,9702
+1790786174.642070150,9828,9838
+1790786174.740548365,9828,8824
+1790786174.839833831,9828,9838
+1790786174.937788440,9828,8824
+1790786175.042638529,9866,9848
+1790786175.147763531,9838,9848
+1790786175.251046852,9838,9848
+1790786175.357220336,9272,9838
+1790786175.462097365,9828,9838
+1790786175.563701577,8814,8824
+1790786175.658918195,8814,8824
+1790786175.756543930,8814,8824
+1790786175.857112750,8814,8824
+1790786175.955723127,8814,8824
+1790786176.057100796,8814,8824
+1790786176.164637056,9016,9856
+1790786176.269457695,9846,9856
+1790786176.369247326,8814,8824
+1790786176.468768483,8814,8824
+1790786176.572184595,8814,8824
+1790786176.680519425,8814,8824
+1790786176.787843012,8814,8824
+1790786176.895493653,8814,9864
+1790786177.002038592,9854,9864
+1790786177.107059683,9110,9480
+1790786177.206975745,9854,9864
+1790786177.308243823,9854,8824
+1790786177.407296535,8814,8824
+1790786177.511630138,8814,8824
+1790786177.613017667,8814,8824
+1790786177.717066116,9274,9864
+1790786177.815656494,9854,9864
+1790786177.915900859,8814,9864
+1790786178.018981288,9854,9864
+1790786178.117579505,8814,8824
+1790786178.217936202,8814,8824
+1790786178.319862747,8814,8824
+1790786178.419795330,8814,8824
+1790786178.525969424,9854,9864
+1790786178.628061531,9854,8824
+1790786178.728005853,8814,8824
+1790786178.833653081,8814,9736
+1790786178.937844083,9854,9864
+1790786179.037766255,9854,8824
+1790786179.140034804,8820,8830
+1790786179.245360839,8820,8830
+1790786179.349544090,8820,8830
+1790786179.454476600,8940,8950
+1790786179.557948563,8940,8950
+1790786179.660475705,8940,8950
+1790786179.760389278,8940,8950
+1790786179.866003435,8940,8950
+1790786179.966888758,8940,8950
+1790786180.078586336,9148,9414
+1790786180.193379950,9404,9414
+1790786180.301062822,9404,9414
+1790786180.409891267,9404,9414
+1790786180.511185555,9404,9414
+1790786180.615584188,9404,9414
+1790786180.722617943,9404,9414
+1790786180.825872873,9404,9414
+1790786180.927496455,9404,9414
+1790786181.036240609,9404,9414
+1790786181.139894494,9404,9414
+1790786181.239871237,9404,9414
+1790786181.341871773,9404,9414
+1790786181.444553357,9404,9414
+1790786181.545995006,9404,9414
+1790786181.645816877,9404,9414
+1790786181.747032645,9404,9414
+1790786181.858655222,9404,9414
+1790786181.954165593,9404,9414
+1790786182.068152478,9404,9414
+1790786182.179624703,9404,9414
+1790786182.293777879,9404,9414
+1790786182.394289098,9404,9414
+1790786182.505242597,9404,9414
+1790786182.608015352,9404,9414
+1790786182.711196992,9404,9414
+1790786182.819210728,9404,9414
+1790786182.922884623,9404,9414
+1790786183.022149047,9404,9414
+1790786183.127445742,9404,9414
+1790786183.231498132,9404,9414
+1790786183.352851121,9404,9414
+1790786183.455674986,9404,9414
+1790786183.569989714,9404,9414
+1790786183.673774791,9404,9414
+1790786183.788232161,9404,9414
+1790786183.893344203,9404,9414
+1790786184.000703941,9404,9414
+1790786184.102348013,9404,9414
+1790786184.203479409,9404,9414
+1790786184.312577357,9404,9414
+1790786184.423768159,9404,9414
+1790786184.525157428,9404,9414
+1790786184.638123101,9404,9414
+1790786184.736714128,9404,9414
+1790786184.849963804,9404,9414
+1790786184.956039516,9404,9414
+1790786185.061942268,9404,9414
+1790786185.160747317,9404,9414
+1790786185.273083292,9404,9414
+1790786185.375250500,9404,9414
+1790786185.480599585,9404,9414
+1790786185.585149950,9404,9414
+1790786185.691940402,9404,9414
+1790786185.800535864,9404,9414
+1790786185.901016463,9404,9414
+1790786186.017956101,9404,9414
+1790786186.119091647,9404,9414
+1790786186.230213799,9404,9414
+1790786186.335299000,9404,9414
+1790786186.440112419,9404,9624
+1790786186.542811883,9614,9624
+1790786186.644845759,9370,9624
+1790786186.752999377,9614,9624
+1790786186.861355316,8878,9624
+1790786186.966925093,9614,9624
+1790786187.073369551,9614,9624
+1790786187.178632085,9614,8824
+1790786187.281926276,9614,9624
+1790786187.394704916,9614,9624
+1790786187.504238909,9614,9624
+1790786187.613575700,8840,9160
+1790786187.727728016,9406,9416
+1790786187.832223631,9406,9416
+1790786187.941769684,9406,9416
+1790786188.047064828,9406,9416
+1790786188.144852405,9406,9416
+1790786188.260695991,9406,9416
+1790786188.363770900,9406,9416
+1790786188.468639289,9406,9416
+1790786188.572878751,9406,9416
+1790786188.681629185,9406,9416
+1790786188.785445422,9406,9416
+1790786188.885867330,9406,9416
+1790786188.990390325,9406,9416
+1790786189.091682043,9406,9416
+1790786189.190010677,9406,9416
+1790786189.292761142,9406,9416
+1790786189.399216419,9406,9416
+1790786189.506859511,9406,9416
+1790786189.617929631,9406,9416
+1790786189.720011158,9406,9416
+1790786189.830994678,9406,9416
+1790786189.932984344,9406,9416
+1790786190.042418256,9406,9416
+1790786190.142131905,9406,9416
+1790786190.243875889,9406,9416
+1790786190.357939294,9406,9416
+1790786190.459039669,9406,9416
+1790786190.559216775,9406,9416
+1790786190.664521479,9406,9416
+1790786190.773597677,9406,9416
+1790786190.884632127,9406,9416
+1790786190.991845523,9406,9416
+1790786191.108368816,9406,9416
+1790786191.210100890,9406,9416
+1790786191.320561794,9406,9416
+1790786191.422654041,9406,9416
+1790786191.537225172,9406,9416
+1790786191.641148860,9406,9416
+1790786191.745807307,9406,9416
+1790786191.861669633,9406,9416
+1790786191.968884069,9406,9416
+1790786192.071584373,9406,9416
+1790786192.185177483,9406,9416
+1790786192.287305381,9406,9416
+1790786192.397678813,9406,9416
+1790786192.497060839,9406,9416
+1790786192.604761661,9406,9416
+1790786192.704385470,9406,9416
+1790786192.817246931,9406,9416
+1790786192.918724221,9406,9416
+1790786193.024778954,9406,9416
+1790786193.128297328,9406,9416
+1790786193.236105151,9406,9416
+1790786193.340450404,9406,9416
+1790786193.440592759,9406,9416
+1790786193.547516322,9406,9416
+1790786193.648185103,9406,9416
+1790786193.764486983,9406,9416
+1790786193.864659308,9480,9490
+1790786193.968057501,9480,9490
+1790786194.070461872,9722,9732
+1790786194.176537044,9722,9732
+1790786194.280394772,9722,9732
+1790786194.387266254,9722,9732
+1790786194.493886684,9722,9732
+1790786194.597803502,9722,9732
+1790786194.698275231,9722,9732
+1790786194.807813474,9722,9732
+1790786194.913704905,9722,9732
+1790786195.017218788,9722,9732
+1790786195.129467792,9722,9732
+1790786195.239846675,9722,9732
+1790786195.357501522,8814,8824
+1790786195.466641560,8814,8824
+1790786195.629528039,0,2