Recipe — Token page

The token dossier is ten reads: profile, signal, smart money, exit risk, inflection, holders, valuation, narrative, deltas, price structure. token-overview runs the whole fan-out server-side — one request paints the page, and each slot is byte-for-byte the single endpoint's response. All POST bodies are { "token": "0x…", "chain": "eth" } unless noted.

One call paints the page

POST /v1/token-overview
{ "token": "0x…", "chain": "eth", "sample": 8 }   # sample = smart-money buyer sample

Slots, each named after the single endpoint it mirrors: tokenIntel, tokenSignal, smartMoneyFlow, exitRisk, tokenInflection, tokenHoldersInsight, tokenValuation, tokenNarrative, tokenChanges, tokenPriceAction. A failed component becomes an error object in its slot plus an entry in failed[] — drop the tile, keep the page.

The tokenChanges slot records a snapshot exactly as the single endpoint does — a bundle call IS a scan, so your next call diffs against it.

À la carte

A lighter page does fine with three singles — profile, signal, price structure — plus smart-money-flow:

POST /v1/token-intel             # the profile: verdict, risk{}, momentum{}, entry{}, activity
POST /v1/token-signal            # one call: BUY / ACCUMULATE / HOLD / AVOID + conviction + rationale
POST /v1/token-price-action      # price structure: trend, breakout, volatility, distance-from-high
POST /v1/smart-money-flow        # are PROVEN traders buying this token right now?
POST /v1/token-risk              # standalone rug/honeypot/concentration verdict + flags[]
POST /v1/token-holders-insight   # accumulating or bleeding? organic or bot-driven?
POST /v1/exit-risk               # distribution early-warning: exitRisk + direction
POST /v1/token-inflection        # pre-move "coiling" detector
POST /v1/token-valuation         # over/under-extended vs on-chain fundamentals
POST /v1/token-narrative         # one-paragraph AI thesis over the computed signals
POST /v1/token-changes           # what FLIPPED since your last call (deltas, not a snapshot)

Fields worth rendering

Name
token-signal → signal, conviction
Type
header
Description
The single answer, with rationale — one AI-written sentence over the computed signals (the AI never invents a number).
Name
token-intel → risk.risk, risk.score, risk.flags[]
Type
risk block
Description
Flags are plain strings written to be shown as warning chips.
Name
smart-money-flow → smartMoneyBuying, mirrorWorthyBuyers
Type
social proof
Description
"3 proven wallets buying" is the strongest single line on the page.
Name
token-changes → changes[]
Type
what's new
Description
Each change has metric, direction, severity, detail — render as delta chips. First call returns baseline: true (nothing to diff yet).
Name
token-price-action → trend, structure, breakout
Type
chart strip
Description
Categorical price structure (higher-highs, breaking-out, …) — a readable chart substitute. available: false means too new to chart; say that.

UX rules

Signal and risk travel together. Never render a BUY signal without the risk verdict beside it — the pair is the product's honesty.
Deltas beat snapshots. If you poll, use token-changes and surface transitions ("exit-risk flipped rising") — that is the alpha, and it is what justifies a refresh button.
available: false and baseline: true are answers. "Too new to chart" and "baseline recorded — check back" beat a hidden tile.