Documentation

Text2Strategy

Describe your trading idea in plain language. partiqon builds the graph for you using AI, ready to backtest immediately.

How it works: Text2Strategy is a chat interface. You describe what you want: entry conditions, exit conditions, stop loss, take profit. The AI translates your description into a valid block graph using partiqon's block catalog. Valid graphs are automatically saved and opened in the editor.

How to use it

Open the workspace. Make sure a symbol and timeframe are set (used as context for the generated graph).
Click AI (the sparkles icon) in the graph editor toolbar.
Type your strategy description and press Enter or click Send.
The AI replies with an explanation of what it built. If the graph is valid, it's automatically saved and opened in the graph editor.
If the generated graph does not validate, the panel shows Validation failed with the reason and nothing is saved. Rephrase or add detail to your request and send it again.
Backtest, optimise, or continue refining the strategy by chatting further.

Allowance and where your text goes

Every message you send counts as one AI request, including each answer to a follow-up question and each request to explain something. Free has 5 requests a month (the count starts again on the 1st, UTC); Pro has 20 and Premium 50 a day (from midnight UTC).

Your description is processed by an AI model of an external provider. Section 7f of the Privacy Policy describes what is sent and to whom.

What to write: example prompts

The more precise your description, the better the result. Include entry conditions, stop loss, and take profit in the same message.

"RSI drops below 30, EMA 50 is sloping up, enter long. Stop 1.5 × ATR below entry, take profit 3 × ATR above entry."

"Enter short when fast EMA (9) crosses below slow EMA (21). Session filter: London hours only (7am to 4pm London time). Exit after 15 bars or when fast EMA crosses back above slow EMA."

"Bollinger Band mean reversion. Price touches the lower band and RSI is below 35. Enter long with 20 pip stop and 40 pip target."

"EMA crossover strategy on M15 bars. EMA 9 crosses above EMA 21, ADX above 25 to confirm trend. ATR-based stop loss (2×ATR), 3× ATR take profit."

Refining in the same conversation

Text2Strategy is conversational, so you can keep iterating in the same chat. Examples:

"Add a volatility filter: block entries when ATR percentile is above 80%."
"Change the take profit to 2.5× ATR instead of 3×."
"Also add a session filter. Only trade between 8am and 6pm CET."
"The RSI threshold feels too strict. Change it from 30 to 40."

The AI remembers the last three turns of the conversation (six messages) for context, so follow-up messages are interpreted relative to what was already built.

Follow-up questions

If your description leaves out something that changes the strategy, the AI may ask a clarifying question before it builds. It asks at most three questions per conversation, one at a time, and never about something you have already said. After the third, it builds with what it has, fills any remaining gaps with sensible defaults and tells you which ones it chose.

Pro and Premium: a Follow-up questions switch at the top of the panel turns this on or off. Off means the AI builds straight from your message, every time. The line next to the switch shows how many questions are left.
Free: the AI always builds in one step, from your message alone.

The more you put into your first message (indicator, entry, exit, trend or mean reversion, how much risk), the fewer questions there are to ask. Starting a new conversation resets the count.

You can also ask how something works, such as an indicator, a block, or an exit like an ATR trailing stop. The AI then explains it instead of building a strategy. An explanation is not a follow-up question: it uses none of the three, and on Free it is answered the same way.

Writing effective prompts

Be specific about numbers. "RSI below 30" is better than "RSI is oversold". "1.5× ATR stop" is better than "tight stop".
Mention entry AND exit. A prompt with only entry conditions produces a graph that has no way to exit. Always include stop loss, take profit, or a bar-count exit.
One idea per message. "Enter long when X, filter by Y, exit when Z" in one message works well. Avoid combining multiple unrelated strategies in one prompt.
Use block names if you know them. You can say "use Gate.RegimeFilter to allow entries only in mean-reverting regime" and the AI will use that block exactly.
Mention the timeframe context. "On the H1 bar" or "M5 strategy" helps the AI pick the right aggregation blocks.

What the AI can build

Text2Strategy can combine any of the 95 blocks it is offered. Commonly generated patterns include:

Indicator crossovers (EMA, MACD signal line, Stochastic %K/%D)
Threshold conditions (RSI below X, ADX above Y, Z-Score extremes)
ATR-based stop loss and take profit sizing
Session time filters
Trend gates (EMA slope filter, Regime Filter)
Volatility gates (ATR percentile)
Multi-timeframe context (daily EMA trend filter on an intraday strategy)
Chandelier trailing stops using Source.HighSinceEntry / Source.LowSinceEntry

Chat history

Your conversation history is stored locally in your browser (up to 200 messages). It persists across page reloads and sessions on the same device. It is not synced across devices. To start fresh, use the New conversation button at the top of the Text2Strategy panel: it removes the chat from your browser's storage and leaves the panel as it was on your first visit. If the last reply produced a strategy you have not saved, it asks before discarding it.

When to use Text2Strategy vs. the graph editor

Use Text2Strategy when you have a strategy idea but aren't sure which blocks to use, or when you want a working starting point quickly.
Use the graph editor directly when you know exactly which blocks you need, or when you're making small tweaks to an existing graph.
Use both: generate with Text2Strategy, then switch to the editor to fine-tune the wiring, add tunable ranges, or replace blocks that weren't quite right.