Knowledge Base & FAQ

Use this page to choose the right diagnostic path. A mismatch, missing row, or compiler message is an observation—not a confirmed cause.

Scanner Connections, unavailable results, preparation, storage, and conditional monitoring

Backtest vs Alerts Imports, timezones, unmatched events, chart data, and AI review

Alerts vs Fills Broker connection, alert sources, matching, partial lifecycles, and current-window review

AI Pine Editing Chat modes, context, staged changes, validation, saving, and credits

Algo Lab Data, runtime differences, settings, backtests, and the supervised Agent

Find saved research and continue a check

Open Research overview to search your Case Studies and inspect their saved trials and research stages. Report opens a saved trial; Continue with Algo Agent opens a study-bound draft for you to review and send. Opening it does not run a measurement. A saved backtest and a gate reading may describe different versions, so inspect their scope before comparing them.

See Research overview and the research workflow.

Where are research limits configured?

Use Algo Lab → Configuration → Research defaults for new-study defaults. Use the study’s Research settings for its own saved limits. A blank limit is undeclared; it does not remove credit, execution or evidence requirements. A connected assistant must ask before increasing a reached study limit. See research budgets.

My external AI client cannot continue

Confirm the connected account, scopes and entitlement with ata_context and inspect the returned research contract before retrying. Use durable jobs for long operations and retrieve their results; a completed job is not necessarily a passed check. Never paste credentials into a prompt to work around an access refusal. Follow MCP troubleshooting and tools and workflows.

A webhook event is missing

1. Open Webhook Management and copy the current endpoint and authentication key instead of retyping them.

2. Compare the TradingView webhook URL and JSON payload with the documented webhook standard. Confirm the body is valid JSON.

3. Check the webhook event log for the same time window. Preserve the HTTP or validation message if the request was rejected.

4. Send one controlled test event before changing the Pine strategy.

Do not post the authentication key in screenshots, support messages, or AI chat.

A CSV will not import

Start with the workflow that owns the file:

• For TradingView Strategy Tester → List of Trades, use the Backtesting Upload guide.
• For alert-history CSV, use Data Scope or Scope & Matching.

Check the actual headers, one raw timestamp, and the chosen timezone. Do not rewrite symbols or timestamps until the importer shows which row or required field failed. Keep an untouched copy of the export for comparison.

A match rate changed

Rerun the same symbol, source, date window, timezone, and tolerances. Then inspect the unmatched rows and compare the source counts before attributing the change to Pine code, a provider, or market conditions.

No universal pass threshold

A match rate, latency, slippage, or engine-difference percentage is not a universal readiness score. Interpret the distribution against the strategy’s intended event semantics and a comparable prior window.

Algo Lab and TradingView disagree

Algo Lab uses an independent Pine-compatible research runtime, not TradingView’s official engine, and it supports Pine Script v6 only. Confirm the TradingView copy declares //@version=6, then compare the same saved source, symbol, timeframe, date range, inputs, sizing, commission, slippage, and session settings. Then separate compiler differences from runtime or broker-simulation differences.

Use the Engine vs TradingView guide for the full checklist. TradingView’s Pine Editor and Strategy Tester remain the final evidence for behavior inside TradingView.

The Algo Agent keeps running

Professional

Select Stop to halt the current Agent run at its next safe boundary. Use Ask for a read-only explanation or Research for a bounded research turn that does not apply editor edits. Agent mode can perform multiple supervised steps and may pause for an approval or decision.

If a candidate is staged, review the complete diff. Accepting it changes the editor source; it does not establish profitability or readiness. Validate, save, and rerun the relevant evidence gates.

Market data will not load

• Backtest vs Alerts and Algo Lab may use platform-served IEX stock bars when that fallback is enabled for the page and scope.
• A personal Alpaca connection is required when the fallback is unavailable and for Alpaca fills, portfolio data, and account-specific feeds.
• Confirm the symbol, date window, asset class, and paper/live account before retrying.

See Platform Market Data and Account Integrations.

The chart shows an IEX notice

IEX is a supported historical-data feed, not a requirement to upgrade to SIP. Its observations cover a different source from consolidated data. Close the notice to suppress it in that browser; if storage is unavailable, it stays hidden only for the current visit. Dismissal does not change the feed. The separate Live Data Feed mode in Alerts vs Fills still requires its listed entitlement. See feed selection.

Scanner will not run or its results look different

Start with the Scanner quick start. Resolve the Alpaca connection warning and wait for the current rule’s capability check. A failed check has a retry action; a visible scheduling or preparation control does not establish that the environment enables that service. Python remains Coming soon.

Inspect Unavailable rows separately from Not matched, including their reasons, session cutoff and bar count. Daily calculations use up to 250 bars; intraday calculations use up to 500, within the explicit 100-symbol scope. A missing required bar is not a failed condition or a market exit.

Prepared scalar fields can support compatible threshold edits. Exact daily rule results preserve engine decisions and scores; changing the rule requires new preparation. While a capability check is pending, a displayed previous result has not applied the updated filters. A separate chart preview has its own feed and date and does not reconstruct the scan’s original evidence.

If storage prevents a run, open the saved screen’s Run history, export what you need and remove unneeded results. Compact results already omit original candles; retrying without history does not remove their retained result storage. Some audit metadata remains charged, and settled expired preparation is reclaimed by maintenance. See MCP and monitoring for metadata-only run discovery, protected reads and conditional schedules.

Still need help?

Open Dashboard → Support and include:

• the route and exact action;
• symbol, timeframe, date window, and data source when relevant;
• the exact error message;
• a minimal reproduction; and
• a redacted screenshot or short recording for visual problems.

Never include passwords, AI-provider keys, webhook authentication keys, authorization headers, or unredacted account data.