GMGN RESOURCE BUDGET

THE CONSTRAINT

what one global token bucket paces every GMGN call on this box: MIN_INTERVAL=3.2s → 0.312 req/s ≈ 27,000 calls/day. That is the entire budget, shared by ~106 callers. why it matters the fleet is asking for 2.7× what exists (71,651 attempts over the 24.0h the call log actually covers), so the gate drops the excess instantly (a caller whose slot is more than MAX_WAIT=8s away is shed, not queued). This is NOT the circuit breaker and NOT a ban — live usage is only 7,743 calls, well under budget. the decision shedding is indiscriminate: it drops whoever asks when the queue is deep, not whoever matters least. Every row below is a claim on the same budget — reallocation means making the valuable lanes patient (raise their per-caller GMGN_MAX_WAIT) and the noisy lanes cheaper (batch, cache, or move them off GMGN entirely).

shed % per hour · last 24h · left = oldest · red >80% shed

TRADING vs RESEARCH â served% at a glance Â· last 24.0h

Paper execution needs timely GMGN data to grade signals; research fills on patience ticks. Wide gap = gate is starving the trading book. Priority lanes (#1287, awaiting DECISION(CHRIS)) would flip this.

TRADING Â· EXECUTION / PAPER

9.2%

3,134 served Â· 34,087 attempts

RESEARCH Â· ALL OTHER LANES

31.3%

11,637 served Â· 37,121 attempts

Research ahead 3.4x

⚠ BIGGEST WASTE — most budget burned for nothing

Ranked by calls SHED, normalised to a projected daily rate so a job that runs every 10 minutes cannot hide behind a small per-window number. A PAUSED badge means the cost is already stopped and the row is kept for the record.

EXECUTION / PAPER

paper_trader.py

20,864

calls shed in 24.0h · 22,318/day projected · 7% hit

EXECUTION / PAPER

paper_universal.py

8,642

calls shed in 24.0h · 10,268/day projected · 16% hit

WHALE INTEL

wallet_intel_db.py

8,038

calls shed in 24.0h · 8,067/day projected · 0% hit

WHALE INTEL

cluster_detector.py

2,449

calls shed in 24.0h · 2,901/day projected · 15% hit

BUDGET BY LANE — where the scarce resource actually goes · last 24.0h

read this first. If whale intel is the asymmetric-profit lane, its share of this budget should look deliberate. Today it is whatever the shedding happened to leave.

LANE	ATTEMPTS	SHARE OF BUDGET	SERVED	HIT
■ WHALE INTEL	26,068	36.5% of the shared budget	8,849	34%
■ EXECUTION / PAPER	34,087	47.7% of the shared budget	3,134	9%
■ SIGNAL SCANNERS	7,712	10.8% of the shared budget	2,556	33%
■ ENRICHMENT / FEEDS	3,341	4.7% of the shared budget	232	7%
■ UNCLASSIFIED	289	0.4% of the shared budget	101	35%

BY PROCESS — who is taking the budget · last 24.0h · 100% attributed

ATTEMPTS = what it asked for. SERVED = what it got (live + cache + stale). HIT = its success rate; a low hit rate on a high-value lane is the thing to fix, a low hit rate on a noisy lane is the gate working as intended.

LANE	PROCESS	ATTEMPTS	SHARE	SERVED	LIVE	HIT	SHED	PROJ/DAY	STATUS	VERDICT
EXECUTION / PAPER	paper_trader.py	22,318	31.2% of demand	1,453	225	7%	20,864	22,318		HEAVY + FAILING — batch, cache, or move off GMGN
EXECUTION / PAPER	paper_universal.py	10,268	14.4% of demand	1,621	517	16%	8,642	10,268		contended
WHALE INTEL	wallet_intel_db.py	8,067	11.3% of demand	29	8	0%	8,038	8,067		STARVED — raise its GMGN_MAX_WAIT (patient lane)
WHALE INTEL	wallet_sell_tracker.py	4,091	5.7% of demand	3,755	1,938	92%	7	4,091		healthy
WHALE INTEL	cluster_detector.py	2,901	4.1% of demand	444	189	15%	2,449	2,901		contended
WHALE INTEL	wallet_total_profit.py	2,562	3.6% of demand	1,634	1,384	64%	19	2,562	ACTIVE
WHALE INTEL	stealth_whale_convergence.py	2,304	3.2% of demand	378	79	16%	1,889	2,304	ACTIVE	contended
SIGNAL SCANNERS	gmgn_trending_multi.py	1,831	2.6% of demand	1,734	1,079	95%	83	1,831		healthy
ENRICHMENT / FEEDS	token_info_enrichment.py	1,827	2.6% of demand	30	11	2%	1,797	1,827	ACTIVE	HEAVY + FAILING — batch, cache, or move off GMGN
SIGNAL SCANNERS	alphafeed_convergence.py	1,706	2.4% of demand	82	28	5%	1,622	1,706		HEAVY + FAILING — batch, cache, or move off GMGN
SIGNAL SCANNERS	accumulation_detector.py	1,694	2.4% of demand	53	16	3%	1,641	1,694		HEAVY + FAILING — batch, cache, or move off GMGN
WHALE INTEL	sapphy_hot_alert.py	1,224	1.7% of demand	539	192	44%	685	1,224	ACTIVE	contended
EXECUTION / PAPER	outcome_tracker.py	1,156	1.6% of demand	33	12	3%	1,123	1,156		HEAVY + FAILING — batch, cache, or move off GMGN
ENRICHMENT / FEEDS	ta_enrichment.py	1,079	1.5% of demand	91	74	8%	988	1,079	ACTIVE	HEAVY + FAILING — batch, cache, or move off GMGN
WHALE INTEL	whale_registry_dna_enrich.py	854	1.2% of demand	766	472	90%	87	854	ACTIVE	healthy
WHALE INTEL	sapphy_whale_mover_reporter.py	637	0.9% of demand	6	6	1%	631	637	ACTIVE	STARVED — raise its GMGN_MAX_WAIT (patient lane)
WHALE INTEL	scout_wallet_resolver.py	560	0.8% of demand	468	468	84%	89	560		healthy
WHALE INTEL	whalemover_tracker.py	487	0.7% of demand	476	67	98%	8	487		healthy
SIGNAL SCANNERS	fomo_card.py	480	0.7% of demand	46	30	10%	433	480		HEAVY + FAILING — batch, cache, or move off GMGN
WHALE INTEL	mega_cluster_discovery.py	474	0.7% of demand	87	84	18%	387	474	ACTIVE	contended
SIGNAL SCANNERS	trenches_detector.py	432	0.6% of demand	48	13	11%	384	432		HEAVY + FAILING — batch, cache, or move off GMGN
ENRICHMENT / FEEDS	wfb_feeder.py	350	0.5% of demand	111	111	32%	234	350		contended
EXECUTION / PAPER	apex_runner.py	345	0.5% of demand	27	4	8%	318	345		HEAVY + FAILING — batch, cache, or move off GMGN
SIGNAL SCANNERS	momentum_signal_detector.py	288	0.4% of demand	88	25	31%	200	288		contended
SIGNAL SCANNERS	momentum_scalp_detector.py	288	0.4% of demand	240	116	83%	46	288		healthy
WHALE INTEL	wallet_profile_sig.py	240	0.3% of demand	5	0	2%	234	240		STARVED — raise its GMGN_MAX_WAIT (patient lane)
WHALE INTEL	wallet_supply_pct.py	238	0.3% of demand	5	5	2%	233	238		STARVED — raise its GMGN_MAX_WAIT (patient lane)
WHALE INTEL	whale_cluster_expander.py	221	0.3% of demand	13	13	6%	208	221	ACTIVE	STARVED — raise its GMGN_MAX_WAIT (patient lane)
SIGNAL SCANNERS	fomo_graduate.py	203	0.3% of demand	1	1	0%	202	203		HEAVY + FAILING — batch, cache, or move off GMGN
SIGNAL SCANNERS	base_trending_scanner.py	188	0.3% of demand	37	18	20%	151	188		contended

BY CALL TYPE — what is being asked for

A family with a huge share and a near-zero hit rate is pure waste: it never gets an answer and it crowds out everything else while failing.

FAMILY	ATTEMPTS	SHARE OF DEMAND	SERVED	HIT
token/info	29,509	41.2%	2,019	7%
token/traders	11,336	15.8%	2,415	21%
portfolio/stats	10,881	15.2%	1,668	15%
portfolio/activity	7,174	10.0%	4,946	69%
track/smartmoney	2,081	2.9%	559	27%
market/trending	1,859	2.6%	1,532	82%
token/holders	1,852	2.6%	144	8%
portfolio/holdings	1,602	2.2%	826	52%
track/follow-wallet	1,361	1.9%	0	0%
market/kline	1,185	1.7%	75	6%
track/kol	903	1.3%	100	11%
market/signal	864	1.2%	328	38%
market/trenches	680	0.9%	237	35%
token/security	263	0.4%	30	11%

PROVENANCE

source gmgn_calls.jsonl (one line per gate invocation) · requested window 24h, log actually spans 24.0h and the demand rate is computed against that OBSERVED span, not the nominal window · capacity from GMGN_MIN_INTERVAL · caller identity added to the gate 260824 by walking /proc, so rows logged before that read unattributed and the ATTRIBUTED KPI climbs toward 100% as jobs cycle. No addresses, keys, or arguments are ever logged — only the calling script's basename.