create_sweep
Create the search, once the person has agreed to its size. The engine builds every candidate strategy itself out of the pool you give it -- the whole signal library unless narrowed by class or by naming signals -- and draws the parameters too. USE size_sweep FIRST and say the run count it reports: this call creates a draft and validates it, and a person who has not heard a total has not agreed to one. Still free; only launch_sweep spends. If valid comes back false, errors says why in Oracle's words, nothing was kept, and the fix is another attempt. Nothing runs until launch_sweep.
Tool call template
Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.
{
"id": 1,
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"arguments": {
"sweep_name": "<sweep_name>"
},
"name": "create_sweep"
}
}Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| assets | No | Asset symbols, e.g. ["BTC", "ETH"]. | |
| regimes | No | Market conditions to keep: current, bear, neutral, bull. current means data within the last 30 days, which is recency and not direction. Omitted, every condition. | |
| windows | No | Specific window files to include whatever else is chosen. | |
| categories | No | Draw only from these signal classes, e.g. ["momentum"]. The classes are averaging, flow, momentum, oscillator, pattern and volatility. Omitted, the whole library. | |
| sweep_name | Yes | What to call it. | |
| candle_sizes | No | Candle sizes, e.g. ["1h"]. Omitted, every size Oracle holds for those assets, which multiplies the windows. | |
| exit_filters | No | Draw exit filters only from these. | |
| entry_filters | No | Draw entry filters only from these. Omitted, every filter. | |
| exit_triggers | No | Draw an exit trigger only from these. Omitted, no exit signal is drawn at all and runs exit on the risk settings. | |
| param_samples | No | How many values each signal parameter contributes. The space is their product, so this is the dial that decides whether the search explores parameters at all; at 1 every candidate runs at a single setting. | |
| signal_groups | No | Keep kinds of signal apart. A candidate draws ALL of its signals from ONE group, so two signals in different groups never appear on the same strategy. Runs are split across the groups. | |
| entry_triggers | No | Draw the entry trigger only from these signals, named as the knowledge graph registers them. Omitted, every trigger. | |
| execution_base | No | Risk settings every run shares, from the execution config schema. Omitted, platform canon. | |
| include_gapped | No | Include windows missing bars. Off by default, because a metric over a half-covered window is not comparable to one over a whole window. | |
| max_total_runs | No | A ceiling on the executed runs, applied by Oracle. | |
| candidate_shape | No | How many parts a drawn strategy gets, any of n_entry_triggers, min_entry_filters, max_entry_filters, min_exit_triggers, max_exit_triggers, min_exit_filters, max_exit_filters. Omitted, the platform default shape. | |
| execution_sweeps | No | Execution parameters to search too, each {param: {min, max, n}}. Every one of these multiplies the run count. | |
| draws_per_dataset | No | How many candidate strategies to draw PER WINDOW. The runs that execute are this multiplied by the number of windows. | |
| pre_filter_threshold | No | Skip a drawn candidate the SIEVE model predicts will not trade at all, above this confidence. A will-it-trade gate and nothing about whether it will make money. Omitted, no skipping. |