Built Crucible Algo: Natural Language to Executable Strategy + Institutional Backtesting & Co-Pilot (Looking for Dhan Community Feedback)

APIs, Automation, Algos & Code
1

Hey everyone,

Most algorithmic trading workflows in India are heavily fragmented: you analyze setups on one charting platform, write manual code or struggle with rigid rule builders to backtest, rely on generic screeners that miss market depth, and then rebuild everything from scratch to hook into broker APIs like Dhan.

I’ve been building Crucible Algo to unify that entire pipeline into a single, institutional-grade workbench.

Here is what the platform delivers:

  • Natural Language to Executable Strategy: Describe your setup, entry/exit criteria, and risk parameters in plain English. The engine translates your natural language directly into structured, mathematically sound, executable logic—no manual scripting required.

  • Institutional-Grade Backtesting Engine: Built specifically around the reality of Indian markets. It accounts for statutory charges (STT, exchange turnover fees, GST, stamp duty), realistic liquidity, and slippage curves so your backtested net PnL matches actual contract notes.

  • Pro Charting & Quantitative Screeners: Institutional charts integrated with multi-factor screeners that filter across price action, momentum, and volume profiles in real time.

  • AI Trading & Portfolio Co-Pilot: An interactive assistant that understands markets, technical setups, and risk management. You can query it about setups or connect your brokerage account to analyze your live positions, portfolio factor risk, and exposure in real time.

  • Seamless Broker Connectivity: Bridges directly with broker APIs like DhanHQ so you can take a verified strategy from natural-language concept $\rightarrow$ backtest $\rightarrow$ live deployment without friction.

We’re opening this up for feedback from active systematic traders, API builders, and discretionary traders looking to automate.

What I’d love to know from the Dhan community:

  1. What is the biggest headache you face when translating a discretionary setup into an automated rule?

  2. When analyzing your live portfolio or running backtests on Indian equities/derivatives, what specific telemetry or risk metrics do you feel existing platforms completely ignore?

Check it out at cruciblealgo—looking forward to your thoughts, edge cases, and honest critique!

2

Hi @Priyankit_Raj ,

  1. Biggest headache in turning a discretionary setup into an automated rule: Discretionary traders use judgment that is hard to put into rules, such as “good” price action, “strong” momentum, or “clean” support. Turning these into exact numbers (candle conditions, thresholds, timeframes) is the hardest part. Other common problems are handling partial fills, order rejections, and API failures in live trading, which never show up in a backtest. It would help if your natural language engine asked follow-up questions whenever a rule is vague, and showed the exact logic it generated before running it.

  2. Metrics that existing platforms usually ignore:

    • Slippage and fill quality: actual fills versus the assumed price, especially in illiquid stocks and options.
    • Drawdown details: maximum drawdown duration and recovery time, not just the percentage.
    • Expectancy and risk of ruin: how likely the strategy is to blow up the account, not only the win rate.
    • Margin usage over time: margin blocked per trade and how it affects position sizing.
    • Portfolio concentration: exposure by sector, correlation between open positions, and net delta and vega for option portfolios.
    • Backtest versus live comparison: a running check of how far live results drift from backtest results.