71QuantLab manual

Applies to v0.2.7 · last updated 2026-10-10

What it is

71QuantLab packages three things into one locally deployed Windows app:

1. Data collection: daily bars, indices, limit pools, news and fund flows
2. 9 AI masters: each with its own input shape and knowledge base, consultable on its own
3. The roundtable: seats debate → the decision master concludes

The knowledge base ships inside the installer. The model API and data sources are yours to configure; by default it uses public sources that need no key.

Requirements

Installing

1. Get the installer from the download page.
2. Run it. Windows may show "Windows protected your PC" — the installer isn't code-signed yet, so click "More info → Run anyway".
3. Don't install into Program Files. Data is written to a data\ folder inside the install directory, and system folders aren't writable. Something like D:\71QuantLab is better.
4. Launch it from the Start menu or the desktop shortcut.

First run

On first launch, work through it in this order:

1. Environment check

Go to "Settings → Diagnostics", or run the check from the command line:

runtime\python.exe engine\cli.py doctor

Twelve checks; anything that fails explains the cause and the next step.

2. Create the database

Offline, a few seconds, schema only:

runtime\python.exe engine\cli.py db init

3. First data pull

Choose a data range under "Settings → Data", or use the command line:

runtime\python.exe engine\cli.py sync bootstrap --mode quick

This fetches 8 indices plus the CSI 300 constituents and takes about 5–10 minutes. You can interrupt it and it will resume from where it stopped.

4. Daily increments

runtime\python.exe engine\cli.py sync daily

The app also runs an incremental sync each time it starts.

Data sync

Source strategy: AKShare as the primary source, BaoStock as a supplement, with Tushare optional. The engine has fallbacks and timeout guards built in.

The Overview and Settings pages show the latest date in each table, source availability, sync history and failure reasons.

Ask a Master

The left rail holds 9 masters, each with a fixed input shape:

Codes and names support fuzzy matching. The flow:

1. Select a master on the left
2. Fill in the ticker or question in the shape that master expects
3. Wait for the answer — waiting shows a breathing state, not a spinner

Data is pre-checked before a run: anything missing is flagged first and you decide whether to continue. A running task keeps going while you switch pages; a progress bar sits at the top and can be cancelled.

What an answer looks like

Answers aren't chat bubbles; they're research reports:

• Byline: persona + model + data date + confidence scale
• Body: 40em measure for readability
• Sources: collapsible, expand when you want the basis
• Actions: follow-up, save to archive (VIP)

Confidence is shown as a scale (high ▮▮▮ / medium ▮▮▯ / low ▮▯▯), not a progress ring.

Roundtable

Enter a code or name to open a session. The stages are visible as they run:

Three sessions a day on the free tier, unlimited with VIP. Every seat uses the same model configured in Settings.

Decision archive

Indexed by month on the left, with a memo-style detail view in the middle: the original question, each seat's view, and the later review.

• Stored long term, reviewable and searchable
• Outcome reconciliation: expected range vs actual result vs deviation
• Exportable (VIP)

Configuring the model API

The AI features need a model. The app speaks the OpenAI protocol, so one base URL + key + model name covers DeepSeek / Qwen (dashscope) / Kimi (moonshot) / OpenAI / a local Ollama.

1. Open "Settings → Model API"
2. Click a vendor preset to fill in the URL and model name
3. Paste your own API key
4. Click Save, then Test connection

Keys are encrypted with Windows DPAPI and written to config\settings.json; the interface only echoes the last four characters.

Free vs VIP

Licensing is "free for everyone, plus VIP": install it and use it, with no trial period that locks you out.

VIP comes in monthly, yearly and lifetime tiers. Monthly and yearly fall back to the free tier when they expire rather than locking you out: Ask a Master, the roundtable and default screening keep working, and only the VIP items are withdrawn. The lifetime tier has no expiry date.

Full comparison and pricing are on the 71QuantLab product page.

Activation and changing machines

Licensing uses an activation code plus a one-time online check. Enter it once in Settings and it keeps working offline afterwards.

Activating

1. Open "Settings → Licensing" and copy the device code (it looks like F916-CD9A-4FE5-ED44)
2. Send the device code and the tier you want to me (see the purchase request)
3. Once payment is confirmed I send you the activation code; enter it in Settings and activate — it verifies online and takes effect immediately

Settings → Licensing also shows the activation service address, the activation code field, activation and renewal status, days remaining, the last online check, your device code and the feature matrix.

Changing machines / renewing

The device fingerprint hashes the motherboard serial, CPU ID and system install ID together, so a single hardware change doesn't invalidate it immediately; if none can be read it falls back to the install path.

Folders and data

The layout after installation:

<install dir>\
├── app\             program (Electron / Node / front-end build)
├── runtime\         embedded Python
├── assets\          persona and knowledge base (read-only)
├── data\            ← your data lives here by default
│   ├── quantcanvas.db   market database
│   ├── research.db      research database
│   └── backup\          automatic backups, last 3 kept
├── config\settings.json user settings (keys encrypted with DPAPI)
└── logs\  reports\  cache\

Troubleshooting

The most useful thing to send me is the error from the logs\ folder, plus which data source and model you were using.

Not available

Still stuck?

Go to the contact page with the version, the error message and what you did. Buying VIP and changing machines are handled there too.