Open-source platform for 3D AI agents. Turn text or a photo into a rigged, animated GLB avatar, give it an LLM brain, memory and a wallet, and embed it anywhere with one web component. Remote MCP server, x402 payments, Three.js, browser-native.
JavaScript 3d-generationai-agentsanimationavatardigital-humangenerative-aiglbgltfimage-to-3dllmmcpmcp-servermodel-context-protocolsolanatext-to-3dthreejsweb-componentswebglwebxrx402
-- Migration: Token Intel Pre-Snipe Gate — rugpull-risk verdicts per mint.
-- Apply: psql "$DATABASE_URL" -f api/_lib/migrations/20260627120000_token_intel_pre_snipe_gate.sql
-- Idempotent.
--
-- The autonomous x402 loop (self/023) pays /api/x402/token-intel ($0.01 USDC) for
-- the freshest pump.fun mints the sniper is about to consider and writes the
-- live due-diligence verdict here, one row per (mint, network). The verdict
-- carries a 0..100 rugpull sub-score (token-intel's risk.score — higher = riskier)
-- and a `rejected` boolean set when the risk level is high/critical.
--
-- Downstream consumer: workers/agent-sniper/oracle-gate.js reads this table on the
-- pre-snipe path (new_mint, first_claim, intel_confirmed, prelaunch-radar). A fresh
-- `rejected = true` row vetoes the snipe before any SOL is committed — the
-- "high-risk mints are auto-rejected" safety floor. The oracle feed can blend the
-- same sub-score into conviction. Fail-open by design: a missing/stale row never
-- blocks a snipe, so this layer can only ever make the sniper safer.
begin;
create table if not exists token_intel_risk (
mint text not null,
network text not null default 'mainnet' check (network in ('mainnet','devnet')),
-- 0..100 rugpull sub-score (token-intel risk.score; higher = more risk).
rugpull_score int not null default 0 check (rugpull_score between 0 and 100),
risk_level text not null default 'unknown'
check (risk_level in ('low','medium','high','critical','unknown')),
-- true when level is high/critical → the sniper auto-rejects this mint.
rejected boolean not null default false,
-- the token-intel market read behind the score (audit trail + feed blending).
signal text, -- bullish|bearish|neutral
confidence numeric(4,3), -- 0..1
symbol text,
price_usd numeric(20,10),
change_24h numeric,
market_cap_usd numeric,
liquidity_usd numeric,
volume_24h_usd numeric,
factors jsonb not null default '[]'::jsonb, -- [{label,status,detail}]
-- provenance: the x402 settlement that produced this verdict.
tx_signature text,
run_id uuid,
checked_at timestamptz not null default now(),
primary key (mint, network)
);
-- Pre-snipe lookup is by (mint, network); the gate also treats rows older than a
-- freshness window as absent, so index the recency too.
create index if not exists token_intel_risk_checked
on token_intel_risk (network, checked_at desc);
create index if not exists token_intel_risk_rejected
on token_intel_risk (network, rejected, checked_at desc) where rejected = true;
comment on table token_intel_risk is
'Per-mint rugpull-risk verdicts from the x402 Token Intel Pre-Snipe Gate (self/023). '
'Consumed by workers/agent-sniper/oracle-gate.js to auto-reject high-risk mints pre-snipe.';
commit;