Skip to content

Quick start

This flow uses the current pages rather than stale fixed API examples, and follows the research-workspace path.

1. Sign in and open market data

After starting both services, open http://localhost:3000 and sign in. Go to Data → Market Data, then choose an asset class and instrument.

  • Initial loads and instrument changes read the local MySQL market-data warehouse first.
  • Query is the explicit action that fetches latest data from AkShare. If it fails, the page keeps usable local data and shows a readable warning.
  • History is shown newest first. The last instrument is remembered separately for each asset class.

2. Optionally build a knowledge base

In AI → Knowledge Base, create a knowledge base, upload or add documents, then index them. Open AI → Chat, ask a focused question, and inspect the citations.

not_indexed means that a document must be indexed or re-indexed. no_context_found means the index exists but lacks sufficiently relevant context; it is not a model-outage signal.

3. Create or select a strategy

Open Research → Strategies and choose a template, an existing strategy, or an AI-generated draft. AI research can use the default research objective or, after confirmation, ask the configured model to improve that objective.

For custom Backtrader strategies, keep price series in a custom property such as self.dataclose; preserve self.close() as the close-position method and never overwrite it with an attribute.

4. Run research and a backtest

Add the strategy to a research workspace, set the instrument, timeframe, date range, capital, and commission, then run it. The page streams research output while it runs. On completion, inspect:

  • normalized return, annual return, Sharpe, and maximum drawdown;
  • trade count, win rate, profit/loss ratio, and holding period;
  • equity curve, drawdown, configuration, and data range;
  • robustness checks or parameter optimization when needed.

5. Move to a trading workspace only after review

Do not treat a single backtest as execution approval. Complete out-of-sample, data-quality, and risk checks, then put a human-reviewed plan into a trading workspace. Use Portfolio to inspect accounts, positions, trades, cumulative P&L, drawdown, and allocation.

Next, read Strategies and AI research, Backtests and validation, and Trading workspaces.