Broker, data, and AI integrations¶
This page defines a consistent onboarding workflow for external services. Availability depends on the build, account permissions, region, and provider policy; the options shown by your deployed instance are authoritative.
Common onboarding workflow¶
- Create a dedicated provider account or application; do not reuse a personal master key.
- Grant only required permissions and disable withdrawals, transfers, and unnecessary trading scopes.
- Save credentials in the corresponding Settings page, never in chat, screenshots, logs, or Git.
- Run the connection check and verify clock, network, proxy, and allowlists.
- Query accounts and market data before completing a paper or otherwise risk-free test.
- Assign an owner and define expiration, rotation, and revocation.
Brokers and exchanges¶
| Asset class | Common integrations | Prerequisites | First acceptance test |
|---|---|---|---|
| US equities | Alpaca, IBKR, Futu/Moomoo OpenD, Webull OpenAPI, Tiger, depending on build | Region and account permissions; some use a local gateway | Account, cash, positions, orders, paper order |
| Hong Kong equities | Futu/Moomoo OpenD, depending on build | Running and unlocked OpenD | Quotes, positions, orders, currency |
| Crypto | Exchange adapters available in the current build | API key, secret, IP allowlist, market permissions | Clock, balances, symbol, paper or minimum-size order |
Local desktop gateways such as IBKR and Futu/Moomoo must be reachable from the backend. From a container, 127.0.0.1 is the container itself, not the host; use the platform's host address and restrict exposed ports.
Read Live-trading safety before enabling production execution.
Data and research providers¶
QuantDinger can combine public market data, macro data, universes, and professional report providers. Validate in this order:
- Configure a provider under Settings → Professional report data.
- Inspect connectivity, synchronization time, coverage, and errors under Data status & sync.
- Confirm symbols, frequency, timezone, and adjustment policy in Market.
- Generate a small AI Research report and inspect data-quality and risk fields.
- Add usage budgets, caching, and failure fallback for paid or rate-limited providers.
Public sources may be delayed, incomplete, or license restricted. Production decisions should record source, timestamp, and quality state rather than treating one public source as exchange-grade realtime data.
AI / LLM¶
Settings → AI / LLM selects the provider, model, API key, and advanced parameters. Depending on the build, the UI may offer OpenAI, Google, DeepSeek, Grok, AtlasCloud, and custom OpenAI-compatible services.
- Select the provider before entering its credential and model name.
- Use a trusted HTTPS endpoint for custom services and verify the compatible API path.
- Set bounded timeouts, retries, concurrency, and daily budget.
- Do not send unredacted account data, secrets, personal data, or restricted market data to a third-party model.
- Run a minimal research job to verify text, structured fields, error handling, and cost accounting.
Notifications and sign-in¶
Go-live checklist¶
- Credentials use minimum privileges and have a rotation record.
- Test and production credentials are isolated.
- Accounts, positions, orders, currencies, and timezones render correctly.
- Create, pause, resume, and incident handling are tested in paper mode.
- Disconnect, rate-limit, and expired-key failures produce actionable messages.
- Logs contain no secrets, tokens, complete account numbers, or personal data.
Use the Error knowledge base when collecting diagnostics.