Verify the installed module
Follow the install steps above, then run the export check from the dashboard root. Expected result: strategy export present. This checks loading only; it does not submit an order.
Sentiment Proxy Research Agent
Explore a price/volume-derived research model without confusing proxies with external data.
What this guide covers
This is downloadable source software. Hosting, trading funds and external services are separate. Build and signal checks are not certification for unattended real-money execution. Complete venue and recovery validation for your installation. Some packaged agent INSTALL/manifest files retain the older word “candidate”. This guide is tied to the registered September 28 archive by checksum; publication does not certify venue performance.
Some manifest indicator names are aspirational legacy labels, not subscribed feeds. This is not a macroeconomic news service or a coordinator for other agents. Keep it in research until your modified model and runtime are independently validated.
Before you start
- A working Core/Trader dashboard in the manifest range >=2.1.0 <2.2.0. The browser-only Core 2.0 template is incompatible.
- The module manifest declares Node >=20; the complete shipped dashboard requires Node 22.3 or newer.
- Completed, time-ordered OHLCV candles with millisecond timestamps, finite positive prices and valid volume, supplied by your runtime.
- A backup and an isolated paper/research environment. The strategy computes signals; the dashboard supplies market data, order handling and risk controls.
Purchase and download
- Sign in to the account used at checkout. Open Account and locate the confirmed order. If payment is processing, wait for fulfillment rather than buying again.
- Generate a fresh product Download link. Save the ZIP, receipt, release version and checksum. Expired links can be regenerated from the account.
- Compare the archive SHA-256 with the release recorded below. Extract into a new directory; retain the original archive for rollback.
- Open README.md, RUNBOOK.md or INSTALL.md, manifest.json where included, and LICENSE.md before changing an installation.
Windows: Get-FileHash .\your-download.zip -Algorithm SHA256. Linux: sha256sum your-download.zip. macOS: shasum -a 256 your-download.zip.
Install Sentiment Proxy Research Agent
- Check whether you already own the strategy: all six are included in Trader; Darvas is also included in Core.
- Back up the database, settings and plugins. Pause new entries and reconcile positions before changing the installation.
- Extract the module so manifest.json sits directly at plugins/macro-sentiment-agent/manifest.json. Check plugins/macro-sentiment-agent/dist/strategy.js exists; do not nest the product directory twice.
- Restart the dashboard and inspect loader errors. Select strategy ID macro-sentiment, a supported interval and the intended market. Start paused with no funds allocated.
- Use the manifest defaults for the first comparison. Inspect decisions and then independently validate the runtime trade lifecycle before any unattended execution.
Commands from the dashboard root
node -e "const p=require('./plugins/macro-sentiment-agent/dist/strategy.js'); if(!p.strategy) throw Error('strategy export missing'); console.log('strategy export present')"How the strategy works
This experimental module computes sentiment-style estimates from OHLCV candles. Funding, open interest, liquidation risk and flow names describe proxies or simulations in this release. It does not ingest news, social posts, actual funding/open-interest feeds or wallet flows.
Exact shipped defaults
These defaults are a starting point for comparison, not recommended trading parameters or a loss limit. Candle counts depend on the interval. Refer to the shipped source for validation rules before changing them.
| Setting | Default | Meaning |
|---|---|---|
| fundingThreshold | 0.01 | Extreme funding rate % |
| oiChangePeriod | 24 | Periods to measure OI change |
| oiChangeThreshold | 20 | Significant OI change % |
| sentimentSmoothing | 7 | Periods for sentiment MA |
| extremeFear | 20 | Fear & Greed extreme fear threshold |
| extremeGreed | 80 | Fear & Greed extreme greed threshold |
| volumeSpikeMultiplier | 3 | Volume spike threshold |
| priceChangeThreshold | 5 | Significant price move % |
| correlationPeriod | 30 | Reserved legacy field; not read by the shipped signal calculation. Changing it does not alter decisions in this release. |
| dominanceChangeThreshold | 2 | Reserved legacy field; not read by the shipped signal calculation. Changing it does not alter decisions in this release. |
Step-by-step practice tutorial
Work in an isolated paper or research environment. The exercise teaches installation and signal interpretation; it does not send an order.
Prepare comparable inputs
Select 1h, 4h, 1d as supported by the manifest. Supply one completed candle series for one symbol, retaining timestamps, OHLCV, source and network. Multiple listed intervals do not mean automatic multi-timeframe fetching.
Run the strategy-specific exercise
Use completed 1h candles and defaults. Record the derived fear/greed score and its relationship to extremeFear=20 and extremeGreed=80. Compare a quieter sample with a large price/volume move. Label every saved result as proxy research.
Interpret the result
Inspect reasoning and the complete indicator object. Do not cite a fundingRate or openInterest field as an exchange-reported observation. If replacing a proxy with a real feed, define units, timestamps, failure behavior and tests before using it.
Validate the runtime separately
Confirm ownership, sizing, fees, entry fill, stop/target, modifications, exit and realized P&L on testnet. Include restart, partial fill, rejected cancellation and lost-response reconciliation. A signal test is not an execution test.
Supporting video lessons
Narrated, captioned lessons showing the broader demo dashboard. Some screens and hosted workflows differ from the purchased source edition. Use this product’s written installation and practice tutorial for the exact release.
Create and inspect an agent
Monitor, pause and recover
Record your acceptance checks
□ Archive checksum and version match the guide.
□ Install completes without suppressing errors; the module loads from its declared path.
□ Correct owner, market, network and candle interval are recorded.
□ Decision timestamp, direction, reasoning and parameters are captured; neutral is accepted.
□ Settings and ownership persist after restart; backup restore is tested in isolation.
□ Before live execution: entry, fill, protection, modification, exit, fees and P&L reconcile; partial fills, cancellation failure and lost responses are tested.
Inspect reasoning and the complete indicator object. Do not cite a fundingRate or openInterest field as an exchange-reported observation. If replacing a proxy with a real feed, define units, timestamps, failure behavior and tests before using it.
Maintain, update and customize
- Keep a copy of the immutable ZIP, lockfile, private configuration and a database backup before an update. Store encrypted backups separately and prove a restore in an isolated compatible instance.
- Pause new entries and reconcile existing venue orders and positions before maintenance. Pausing entries is not the same as closing positions. Removing a plugin does not close exchange positions.
- Monitor the last successful cycle, data age, rejected orders, protection status, reconciliation gaps, database backups and disk space. Missing telemetry is an unknown state.
- For rollback, stop the new runtime, reconcile the venue, restore a compatible application/schema pair in isolation and verify ownership and protection before resuming. A database restore cannot reverse a trade or transfer.
- Customize readable source in a separate branch. Preserve the original defaults, rebuild any changed compiled module, and test the same inputs before and after. Follow LICENSE.md; source ownership does not imply unrestricted redistribution.
Troubleshooting
Payment confirmed but download missing
Check order status in Account and contact support with the order reference if fulfillment is delayed. Do not make another purchase to repair access.
Download link expired
Generate a fresh link from the purchasing account. The documentation downloads are public; the paid software archive remains entitlement-protected.
Strategy absent from the selector
Confirm manifest.json is directly inside the manifest installPath, the entryPoint exists, compatibility matches and the loader has no errors. Restart after installation; do not add another nesting level.
No trade despite available capital
Inspect the latest decision, candle history and freshness, symbol, interval, network, existing positions and risk rejection. Neutral is a valid result.
Build or database initialization fails
Keep the exact failed command and redacted diagnostic. Check Node, psql, Supabase settings and owner verification. Do not bypass types, authentication or schema integrity checks.
Balances disagree or an order outcome is unknown
Compare the same account and network at the venue, including fills, fees, margin and open orders. Reconcile before retrying; a lost response does not mean the order failed.
Related product guides
Get installation help
Include your order reference, product/version, operating system, Node version, failed step and a redacted error. Never send environment files, passwords, private keys or seed phrases.