TradingView HTML Exporter

This exporter generates a standalone HTML chart in the analysis/charts/ folder using the local uPlot-based TradingView-style renderer.

Recommended: dexbot tv (One Step)

dexbot tv fetches the candles itself (chunked Kibana scans: pools with order-book fallback for pairs) and then renders through this exporter — no manual fetch/export steps needed:

# Bot chart with AMA overlay (bot key from profiles/bots.json, default: 3 months)
dexbot tv <bot>

# Any pool by bare or full ID
dexbot tv 133
dexbot tv 1.19.133

# Any pair (pool-first, order-book fallback)
dexbot tv TOKENA/TOKENB

# MPA price-feed history instead of market candles (opt-in; MPA/MPA pairs
# cross both feeds into one quote, e.g. HONEST.USD/HONEST.EUR)
dexbot tv BTS/HONEST.USD --feed

# Options
dexbot tv <bot> --month 6              # months of 1h candles (default: 3)
dexbot tv <bot> --chart analysis/charts/custom.html
dexbot tv BTS/HONEST.USD --book # override: order-book fills instead of the pool

Output: analysis/charts/tv_<bot|pool_<id>|<a>_<b>>_1h_<N>m.html, with a clickable file:// link printed on completion. Unknown targets fail fast with the list of known bot keys.

The sections below cover manual usage (explicit candle files, direct runner flags).

What It Produces

• Log-scale price chart
• Candle timeframe buttons: 1h, 4h, 1d, 1w
• Pair-orientation switcher for A/B and B/A
• SMA overlay
• AMA preset buttons 1–4 (one-click AMA1–4, active preset highlighted); numeric inputs kept
• Bot-grid range highlight, off by default (the bot's min/max around AMA with live asymmetric tilt; red above AMA, green below). Range and Scale are AMA-derived, so turning one on opts into what it needs: Range → AMA, Scale → Range + AMA; turning AMA off takes Range and Scale back down with it, so the toggles never sit in a dead state. One-way opt-in is re-applied on load, so a chart saved with Scale on renders its band.
• Grid-reset simulation (AMA bots, on by default): replays the market adapter's two recentering triggers over the candle history — the accepted grid center as a step line, the simulated grid range around it, and a marker per reset (init = first AMA snapshot, Δ1 = AMA-price Δ, Δs = AMA-slope Δ). Thresholds come from the live config chain; a bottom-left panel shows the values, where they came from, and the reset counts (see Grid-Reset Simulation)
• Range-scale switch: fit the price axis to the range band. Toggling Range / Scale (or dragging the grid-span slider) keeps the current view — the price axis is not refitted under the cursor; the band-fit applies on the next autofit (reload, timeframe switch, x pan) or immediately on a double-click of the price axis. Opting into Scale/Range also opts into the indicators they need (see above)
• VWMA overlay
• Order overlay for bot charts (active grid buys/sells as dashed levels, reserve line at the lowest grid buy, ceiling line at the highest grid sell, spread label; pair-aware, toggle in-chart)
• Market panel (top-right): SELL / Market / BUY rows with distance-to-market %
• Range panel (bottom-right, smaller type): visible-window candle High/Low (red/green, mirroring the SELL/BUY badge)
• Volume badge (bottom-right of the volume chart, same small type): visible-window max volume, always with the currency suffix (1.2M BTS); click the badge, the legend Vol value, or the toolbar unit button to switch base/quote units (quote ≈ base × close, persisted per chart like the AMA settings; feed charts show publish counts and disable the switch)
• Bottom volume panel with Volume toggle and per-bar hover tooltip
• Crosshair legend with current candle values

Quick Start

A chart requires candle data. Either pass an explicit candle file or let the exporter resolve a bot's candles and AMA settings:

# From an explicit candle JSON file (json source is the default)
npm run analysis:tradingview -- \
  --file market_adapter/data/market_adapter_<bot-key>_1h.json

Or with a bot key (see Bot-Key Usage below):

npm run analysis:tradingview -- \
  --source market_adapter \
  --bot-key <bot-key>

This writes:

analysis/charts/tradingview_chart.html

The bare npm run analysis:tradingview (no args) errors out — the default json source requires --file, and the market_adapter source requires --bot-key.

Bot-Key Usage (Recommended)

The easiest way to generate a chart for a specific bot. Pass the bot key from profiles/bots.json and the exporter automatically resolves the candle file and AMA settings:

npm run analysis:tradingview -- \
  --source market_adapter \
  --bot-key <bot-key>

This picks up the bot's asset pair, market profile AMA defaults, and candle data from market_adapter/data/. AMA is auto-enabled when the bot uses gridPrice: "ama" (or ama1-4).

With a custom chart path:

npm run analysis:tradingview -- \
  --source market_adapter \
  --bot-key <bot-key> \
  --chart analysis/charts/<pair>_tradingview.html

CLI direct equivalent:

node dist/analysis/tradingview/analyze_tradingview.js \
  --source market_adapter \
  --bot-key <bot-key>

How It Works

1. Reads profiles/bots.json to find the bot's assetA, assetB, and ama settings.
2. Resolves the candle file at market_adapter/data/market_adapter_<bot-key>_1h.json.
3. Looks up the matching market profile in profiles/market_profiles.json for AMA defaults.
4. AMA settings priority: bot-specific ama object > market profile > constants (AMA3, slowPeriod 83.6).

The candle file must exist — run the market adapter LP exporter first if needed (see Getting Blockchain Data).

From an Explicit Candle File

node dist/analysis/tradingview/analyze_tradingview.js \
  --file market_adapter/data/market_adapter_<bot-key>_1h.json \
  --chart analysis/charts/<pair>_tradingview.html

Using LP candle files directly:

node dist/analysis/tradingview/analyze_tradingview.js \
  --file market_adapter/data/lp/<pair-folder>/lp_pool_<id>_<interval>.json \
  --chart analysis/charts/tradingview_chart.html

Input Format

The exporter accepts candle data in either of these shapes:

• Array rows: [timestamp_ms, open, high, low, close, volume]
• Object rows: { time|timestamp|ts, open, high, low, close, volume }

If you pass a raw JSON file, the runner normalizes the candles before rendering.

Getting Blockchain Data

Use the market adapter LP exporter to pull blockchain-backed candles before generating the HTML:

# Auto mode — resolves pool + precisions from bots.json / blockchain
node dist/market_adapter/inputs/fetch_lp_data.js --bot BTS-USDT --interval 1h --lookback 4392h

# Or with an explicit date range
node dist/market_adapter/inputs/fetch_lp_data.js --bot BTS-USDT --interval 1h --start 2026-02-23 --end 2026-08-23

# Manual mode — no blockchain needed
node dist/market_adapter/inputs/fetch_lp_data.js --pool 133 --precA 4 --precB 5 --interval 1h --lookback 26280h

For date range fetching, use --start and --end (e.g. --start 2024-03-06 --end 2025-03-06).

That writes a JSON file under market_adapter/data/lp/ which you can then pass to the TradingView exporter:

market_adapter/data/lp/<pair>/lp_pool_<id>_<interval>.json

Note: the fetcher may not exit on its own after printing Saved: (the live BitShares WebSocket + node monitor keep the process alive). It is safe to Ctrl+C once the file is saved.

CLI Flags

Indicator on/off state is not a generation-time flag: SMA, VWMA, AMA, Range and Scale are in-chart toolbar toggles, persisted per chart in localStorage, and Range / Scale opt into the AMA they are derived from. The --range, --no-range, --range-scale and --no-ama flags were removed — they duplicated the toolbar and could disagree with it (--no-ama on a chart with a range band produced a dead band). Nothing in this table sets an indicator on or off; only --range-span sizes the band.

Grid-Reset Simulation

Bot charts (--bot-key <key>, i.e. bots with gridPrice: "ama") additionally replay the market adapter's grid recentering over the candle history, so you can see when the grid would have moved instead of only where it sits today. See docs/GRID_RECALCULATION.md for the runtime mechanics; the chart mirrors §3 and §4 and nothing else.

What it draws

• Grid center (violet step line, legend Grid) — the accepted gridCenterPrice. It only moves on a trigger, so the line is a staircase: flat = "the grid would still be sitting here".
• Simulated range (violet band) — the grid bounds that center would own: the bot's minPrice/maxPrice multipliers around the accepted center, tilted by the accepted slope through applyAsymmetricBounds + applyNarrowingSideGuard (the same functions the live grid build uses). It re-tilts only at resets, exactly like the real thing.
• Reset markers — one vertical line per reset, colored and tagged by reason: grey init (first accepted AMA snapshot, market_adapter_bootstrap), amber Δ1 (market_adapter_delta_threshold), cyan Δs (market_adapter_ama_slope_delta_threshold).
• Panel (bottom-left) — effective AMA Δ and Slope Δ thresholds, the layer each came from, reset counts, and the last reset.

Where the thresholds come from

Resolved through the same chain the running adapter uses, by analysis/tradingview/grid_reset_config.ts:

1. modules/constants.ts (MARKET_ADAPTER.AMA_DELTA_THRESHOLD_PERCENT, AMA_SLOPE_DELTA_THRESHOLD_PERCENT × DYNAMIC_WEIGHT_AMA_MAX_SLOPE_PCT)
2. profiles/general.settings.json → MARKET_ADAPTER.AMA_DELTA_THRESHOLD_PERCENT and MARKET_ADAPTER.AMA_SLOPE_DELTA_THRESHOLD_PERCENT (the editor's AMA-Slope Δ, applied to amaSlope.deltaThresholdPct by applyRuntimeDefaultsFromGeneralSettings)
3. profiles/market_adapter_settings.json → globals → pairs[].marketAdapterSettings → pairs[].botOverrides[<botName>]

The effective numbers come from the production resolvers themselves (applyRuntimeDefaultsFromGeneralSettings, resolveBotCfg, calculateBotThreshold, MarketAdapterService.resolveAmaSlopeDeltaThresholdPercent), so the chart cannot drift from the runtime. The panel labels the winning layer; --grid-delta-pct / --grid-slope-delta-pct short-circuit the chain and report cli as the source.

Gates the replay honors

• The Δs trigger only fires for AMA-grid bots whitelisted for asymmetricBounds (range scaling) in profiles/market_adapter_whitelist.json and carrying an explicit weightDistribution — the adapter skips the slope signal otherwise. When gated off, the panel shows the threshold with off and the slope trigger is inert, like in production. A non-positive resolved threshold disables the trigger too (it never means "fire every cycle").
• The replay starts at the AMA warmup point (getAmaWarmupBars), capped at half the dataset so a short chart still shows something; the applied value is in the panel and can be forced with --grid-warmup.
• Absolute minPrice/maxPrice pin the center like clampGridPriceToBounds — drift is measured against the clamped center, so an AMA outside the bounds does not emit a reset marker every bar; "Nx" multipliers travel with the center and never clamp.
• Not simulated (out of scope for a candle chart): RMS structural divergence, available-funds resizes, manual/legacy triggers, and the staleness / gap / no-new-candle suppression gates. One evaluation per 1h bar.

The replay runs in the page, so the AMA series (and therefore the trigger points) follows the AMA inputs and preset buttons; the gating parameters that would come from config (erPeriod, lookback, warmup, clip percentile, persistence gate) are fixed at generation time. One slope window is shared by every consumer: the resolved lookbackBars drives both the replayed Δs trigger and the Scale band tilt, so a bot-configured lookback can no longer leave the plotted band measuring a different slope than the trigger does (the shared constant remains the fallback for pool/pair charts, which carry no grid-sim data), and both average with the chart's selected slope model (see Notes). The replay uses the canonical simulateGridResetSeries() from analysis/tradingview/grid_reset_sim.ts, embedded verbatim via embedFunctionSources — not a hand copy of the adapter logic.

Notes

• Slope averaging. The AMA slope is a Huber-robust linear regression of ln(AMA) over the lookback window — computeHuberWindowSlopePct in market_adapter/core/strategies/dynamic_weight_series.ts, which is the live adapter's own definition and is embedded here verbatim, so the chart and the bot run one logic path. Its tuning (C, ITERATIONS, SCALE_FLOOR, ZERO_EPSILON) is centralized in MARKET_ADAPTER.DYNAMIC_WEIGHT_AMA_HUBER and injected into the page as const AMA_SLOPE_HUBER, so a constant change flows to both. The output is in %/bar over the same window, so the tilt scale (clamp(slope/maxSlopePct) * maxSlopeOffset) and the Δs gate (|Δslope| >= 8% of maxSlopePct) keep their existing units. Measured on the repo's 1h pools with the shipped AMA preset at the shipped 16-bar window: a single-bar AMA impulse is strongly bounded (a 1%/3%/5%/10% one-bar spike shifts the two-point endpoint reading by ≈ spike/lookbackBars %/bar ≈ 0.06/0.19/0.31/0.63 — 8x to 80x the reset gate — while the robust fit barely moves), and the fit is the smoothest robust option measured (second-difference energy ~25x lower than the old median's). Set CHART_SLOPE_ESTIMATOR = 'endpoint' in the page to render the reference two-point definition instead (kept in the core for comparison); the live adapter does not use it.
• The chart embeds the vendored uPlot runtime inline (no CDN, no sibling uplot/ dir, no DEXBot2 install needed). Each export is a single self-contained HTML file that renders anywhere, even after being copied or mailed to a machine without DEXBot2.
• The displayed indicators are computed from the 1h base candles and then sampled onto the selected timeframe.
• The current volume-weighted overlay is a rolling VWMA, not a session-reset VWAP.
• SMA is disabled by default.
• VWMA is disabled by default.
• AMA is auto-enabled when the bot has gridPrice: "ama" (or ama1-4), otherwise disabled by default.
• The AMA controls start with the bot-specific AMA, then pair-specific entry from profiles/market_profiles.json when available, falling back to AMA3 values from modules/constants.ts.
• The AMA preset buttons 1–4 select the AMA1–4 defaults (active preset highlighted); numeric inputs remain for fine-tuning.
• The pair switcher inverts the candles client-side, so you can inspect both A/B and B/A views from one export. The order overlay inverts with it (same levels in display units, side colors preserved).
• The order overlay resolves from profiles/orders/<botKey>.json (same files scripts/analyze-orders.ts reads): active/partial grid orders only. Pool/pair charts without a bot key render without it, silently. --no-orders removes the whole overlay (levels, reserve/ceiling lines, spread label); the in-chart Orders checkbox does the same when orders are present. Grid bounds span the full grid, never calculated: the reserve line sits on the lowest grid buy, the ceiling line on the highest grid sell — live (active/partial) and planned (virtual) slots alike; a missing side hides its line instead of drawing an invented level. Bounds render even when no live levels exist yet (levels, spread, and panel BUY/SELL rows need live orders).
• The update marker falls back to prevUpdateLastCandleSec / prevUpdateNewBars stamped in the candle-file meta when the flags are absent; those fields are written by external incremental-fetch tooling, not by anything in this repo.
• Ctrl+0 (or Cmd+0) resets the time-axis zoom to the full dataset.
• Mouse: drag the candles to pan time + price (price drag sets a manual range); wheel zooms time, except over the price axis where it zooms price. Shift+wheel zooms price anywhere over the price pane (cursor-anchored). Dragging the price-axis gutter scales price, dragging the time-axis gutter scales the timeframe; double-click the price axis to return to autofit. While the price range is manual, timeframe moves no longer refit it.
• Indicator, timeframe, scale, and overlay-visibility changes are persisted in browser localStorage per pool/pair chart (dexbot2-tradingview-uplot-v3:<pool>:<A>_<B>:<baseSecs|base>); cursor sync between the price/volume panes uses a separate constant key. The Resets toggle is persisted the same way, so a chart opened from a previous session keeps its simulation state.
• The price axis defaults to log base 10, with a toolbar switch for Log / Linear.
• If you regenerate the HTML and then open it later, no CDN access is needed — the uPlot library (JS + CSS) is inlined into the file itself, so it renders fully offline and is independent of where the file lives on disk.

Typical Workflow

1. Pick or generate a candle JSON file.
2. Run npm run analysis:tradingview or call the runner directly.
3. Open analysis/charts/tradingview_chart.html in a browser.
4. Use the timeframe buttons and indicator controls at the top of the page.