File size: 10,760 Bytes
caa28aa | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 | """50-query validation sweep against the live HF Space.
Usage:
python3 scripts/probe_50.py [--base URL] [--concurrency N] [--timeout S]
Default base: https://lablab-ai-amd-developer-hackathon-riprap-nyc.hf.space
"""
import argparse
import asyncio
import json
import time
from pathlib import Path
from urllib.parse import quote
import aiohttp
BASE = "https://lablab-ai-amd-developer-hackathon-riprap-nyc.hf.space"
QUERIES_FILE = Path("tests/queries_50.json")
RESULTS_FILE = Path("tests/probe_50_results.json")
CONCURRENCY = 3
TIMEOUT_S = 120
STEP_STONE_MAP = {
"sandy_inundation": "sandy",
"dep_stormwater": "dep",
"nyc311": "311",
"floodnet": "floodnet",
"floodnet_forecast": "floodnet",
"noaa_tides": "noaa",
"nws_alerts": "nws",
"nws_obs": "nws",
"microtopo_lidar": "microtopo",
"ida_hwm_2021": "ida",
"ttm_forecast": "ttm",
"ttm_battery_surge": "ttm",
"ttm_311_forecast": "ttm",
"prithvi_eo_v2": "prithvi_v2",
"prithvi_eo_live": "prithvi_live",
"gliner_extract": "gliner",
"rag_granite_embedding": "rag",
"mellea_reconcile_address": "mellea",
"geocode": None,
"mta_entrance_exposure": "mta",
"terramind_synthesis": "terramind",
}
def _parse_sse(chunk: str):
events = []
event_type = "message"
data_lines = []
for line in chunk.splitlines():
if line.startswith("event:"):
event_type = line[6:].strip()
elif line.startswith("data:"):
data_lines.append(line[5:].strip())
elif line == "" and data_lines:
raw = " ".join(data_lines)
try:
payload = json.loads(raw)
except json.JSONDecodeError:
payload = {"raw": raw}
events.append((event_type, payload))
event_type = "message"
data_lines = []
return events
async def stream_query(session: aiohttp.ClientSession, query_obj: dict, base: str, timeout_s: float) -> dict:
qid = query_obj["id"]
query = query_obj["query"]
url = f"{base}/api/agent/stream?q={quote(query)}"
result = {
"id": qid,
"query": query,
"status": "ERROR",
"wall_clock_s": None,
"intent_returned": None,
"mellea_passed": None,
"mellea_rerolls": 0,
"stones_fired": [],
"stones_errored": [],
"stones_silent": [],
"citations_resolved": None,
"compare_targets": None,
"error": None,
}
t0 = time.monotonic()
buf = ""
plan_seen = False
final_seen = False
try:
async with session.get(url, timeout=aiohttp.ClientTimeout(total=timeout_s + 10)) as resp:
if resp.status != 200:
result["error"] = f"HTTP {resp.status}"
result["wall_clock_s"] = round(time.monotonic() - t0, 2)
return result
deadline = t0 + timeout_s
async for chunk in resp.content.iter_any():
if time.monotonic() > deadline:
result["status"] = "TIMEOUT"
result["wall_clock_s"] = round(time.monotonic() - t0, 2)
return result
buf += chunk.decode("utf-8", errors="replace")
# process complete SSE blocks (separated by double-newline)
while "\n\n" in buf:
block, buf = buf.split("\n\n", 1)
for evt_type, payload in _parse_sse(block + "\n\n"):
if evt_type == "plan":
plan_seen = True
result["intent_returned"] = payload.get("intent")
targets = payload.get("targets", [])
if result["intent_returned"] == "compare":
result["compare_targets"] = len(targets)
elif evt_type == "step":
step = payload.get("step", "")
ok = payload.get("ok")
if step in STEP_STONE_MAP and STEP_STONE_MAP[step]:
stone = STEP_STONE_MAP[step]
if ok is True:
if stone not in result["stones_fired"]:
result["stones_fired"].append(stone)
elif ok is False:
if stone not in result["stones_errored"]:
result["stones_errored"].append(stone)
elif evt_type == "final":
final_seen = True
mellea = payload.get("mellea") or {}
req_passed = len(mellea.get("requirements_passed") or [])
req_total = mellea.get("requirements_total") or 4
result["mellea_passed"] = f"{req_passed}/{req_total}"
result["mellea_rerolls"] = (mellea.get("rerolls") or 0)
audit = payload.get("audit") or {}
result["citations_resolved"] = audit.get("citations_resolved")
elif evt_type == "error":
result["error"] = payload.get("err", "unknown error")
elif evt_type == "done":
result["wall_clock_s"] = round(time.monotonic() - t0, 2)
if final_seen:
result["status"] = "PASS"
else:
result["status"] = "ERROR"
if not result["error"]:
result["error"] = "done without final event"
return result
except asyncio.TimeoutError:
result["status"] = "TIMEOUT"
except Exception as exc:
result["status"] = "ERROR"
result["error"] = str(exc)
result["wall_clock_s"] = round(time.monotonic() - t0, 2)
return result
async def run_all(queries: list, base: str, timeout_s: float, concurrency: int) -> list:
sem = asyncio.Semaphore(concurrency)
results = []
early_stop = False
connector = aiohttp.TCPConnector(limit=concurrency + 2)
async with aiohttp.ClientSession(connector=connector) as session:
async def bounded(qobj):
nonlocal early_stop
if early_stop:
return {**qobj, "status": "SKIPPED", "wall_clock_s": None, "error": "early stop"}
async with sem:
r = await stream_query(session, qobj, base, timeout_s)
tag = f"[{r['id']}]"
wc = f"{r['wall_clock_s']:.1f}s" if r["wall_clock_s"] else "?"
mel = r.get("mellea_passed") or "-"
rr = r.get("mellea_rerolls") or 0
print(f"{tag} {r['status']} {wc} mellea={mel} rerolls={rr}", flush=True)
return r
tasks = [asyncio.create_task(bounded(q)) for q in queries]
done_count = 0
for coro in asyncio.as_completed(tasks):
r = await coro
results.append(r)
done_count += 1
# Check early-stop: >10 failures in first 20
if done_count <= 20:
bad = sum(1 for x in results if x["status"] in ("TIMEOUT", "ERROR"))
if bad > 10:
print(f"\nEARLY STOP: {bad} failures in first {done_count} queries — Space appears degraded.", flush=True)
early_stop = True
# Sort by original query order
id_order = {q["id"]: i for i, q in enumerate(queries)}
results.sort(key=lambda r: id_order.get(r["id"], 999))
return results
def main():
ap = argparse.ArgumentParser()
ap.add_argument("--base", default=BASE)
ap.add_argument("--concurrency", type=int, default=CONCURRENCY)
ap.add_argument("--timeout", type=float, default=TIMEOUT_S)
args = ap.parse_args()
queries = json.loads(QUERIES_FILE.read_text())
print(f"Running {len(queries)} queries against {args.base} (concurrency={args.concurrency}, timeout={args.timeout}s)\n", flush=True)
results = asyncio.run(run_all(queries, args.base, args.timeout, args.concurrency))
# Write results
RESULTS_FILE.write_text(json.dumps(results, indent=2))
print(f"\nResults written to {RESULTS_FILE}")
# Update verified flags in queries file
passed_ids = {r["id"] for r in results if r["status"] == "PASS"}
for q in queries:
if q["id"] in passed_ids:
q["verified"] = True
QUERIES_FILE.write_text(json.dumps(queries, indent=2))
print(f"Updated verified flags in {QUERIES_FILE}")
# Summary
total = len(results)
passed = sum(1 for r in results if r["status"] == "PASS")
timed_out = sum(1 for r in results if r["status"] == "TIMEOUT")
errored = sum(1 for r in results if r["status"] == "ERROR")
skipped = sum(1 for r in results if r["status"] == "SKIPPED")
wall_clocks = [r["wall_clock_s"] for r in results if r["status"] == "PASS" and r["wall_clock_s"]]
avg_wall = sum(wall_clocks) / len(wall_clocks) if wall_clocks else 0
max_wall = max(wall_clocks) if wall_clocks else 0
mellea_perfect = sum(1 for r in results if r.get("mellea_passed") == "4/4")
print(f"\n{'='*60}")
print(f"Total: {total}")
print(f"PASS: {passed} ({100*passed//total if total else 0}%)")
print(f"TIMEOUT: {timed_out}")
print(f"ERROR: {errored}")
if skipped:
print(f"SKIPPED: {skipped} (early stop)")
print(f"Avg wall-clock: {avg_wall:.1f}s (passing queries)")
print(f"Max wall-clock: {max_wall:.1f}s")
print(f"Mellea 4/4: {mellea_perfect} ({100*mellea_perfect//total if total else 0}%)")
failures = [r for r in results if r["status"] != "PASS"]
if failures:
print("\n--- FAILURES ---")
for r in failures:
print(f" [{r['id']}] {r['status']} — {r['query'][:60]}")
if r.get("error"):
print(f" err: {r['error'][:80]}")
slowest = sorted([r for r in results if r.get("wall_clock_s")], key=lambda x: x["wall_clock_s"], reverse=True)[:5]
print("\n--- SLOWEST 5 ---")
for r in slowest:
print(f" [{r['id']}] {r['wall_clock_s']:.1f}s — {r['query'][:60]}")
high_rr = [r for r in results if (r.get("mellea_rerolls") or 0) > 1]
if high_rr:
print("\n--- HIGH REROLLS (>1) ---")
for r in high_rr:
print(f" [{r['id']}] rerolls={r['mellea_rerolls']} — {r['query'][:60]}")
print(f"{'='*60}")
if __name__ == "__main__":
main()
|