1.Backtest a rule in plain words
Open this step →DoThe first example rule is already sent. Then type make it 50 days.
CheckA full result in a few seconds with Dhan's delivery charges, a run hash, and a second run that skips the model entirely.
2.Verify the listed algos
Open this step →DoNothing to press: the check runs as soon as the page opens.
Check246 backtests in a few seconds. Each strategy gets a verdict (Robust, Mixed or Fragile) and a plain-words reason.
3.Backtesting as a white-label service
Open this step →DoRead the API call and the service levels, then flip the brand switch.
CheckThe same engine and models ship under Dhan's name: Dvar runs them, Dhan owns the product.
4.Ask Dhan Assist about your portfolio
Open this step →DoShow my portfolio is already asked. Tap a row in the holdings card, then tap a suggested follow-up.
CheckAnswers from the account, with cards and a price chart, suggested next questions, and five runtime checks passed on the small tuned model.
5.An F&O explainer in Hinglish
Open this step →DoRead the answer and move along the payoff chart. Then switch the prompt language to हिन्दी and try one.
CheckA reply in Hinglish, a payoff chart computed from the strike and premium, and SEBI's risk disclosure, with no trade suggested.
6.Watch it refuse advice and confirm actions
Open this step →DoAn order request is already sent. Then try Should I buy Tata Motors now?
CheckIt drafts the order as a review card and waits for your confirmation, and declines to give a buy or sell call while still showing the stock's data.
7.See the release gate
Open this step →DoRead the three scorecards and the release gate.
CheckThe tuned small model is close to the frontier model at about 3% of the cost, and the gate still blocks it on two critical misses.
8.Compare models in the Eval Console
Open this step →DoRead the Compare step, then open the Traffic and Calibration steps.
Check180 chats, three models. The prompted small model is blocked on critical failures; the fine-tuned one passes all four gates at a few percent of the frontier model's cost.
9.Post-train Artham
Open this step →DoPress Run the rule-writer eval.
CheckA real measurement in your browser: how many of 40 plain-word strategies the current rule reader gets right, and the misses that become training data.
10.Scan a document
Open this step →DoLook at the specimen cancelled cheque, press Run the harness, then hover IFSC in the results.
CheckEvery field outlined on the document, the smudged IFSC re-read by the large model, and the format checks passing. Try the masked Aadhaar too.
11.How Dvar helps Dhan
Open this step →DoRead the five options on one page.
CheckFor each option: the value to Dhan, what Dvar delivers, and the first result.
12.Share it
Open this step →DoRead the one-page brief, then press Print or save as PDF.
CheckThe five workstreams, what each demo proves, and what we need from Dhan to start.