P0: Fix strategy config schema mismatch (backtesting broken) #25

New Issue

shoko · 2026-04-09T07:11:41+02:00

shoko commented

2026-04-09 07:11:41 +02:00

Problem

The LLM outputs a nested params structure, but the backtest and simulation engines expect flat fields. This is a complete pipeline break - AI-parsed strategies will never trigger any trades.

Current Flow

LLM Output:     { type: "price_drop", params: { token: "PEPE", threshold_percent: 5 } }
Validator:      Expects params.threshold_percent
Engines:        Look for flat threshold -> Returns 0, conditions never trigger
Frontend:       Uses flat structure { token, threshold }

Schema Comparison

Component	Token Field	Threshold Field
LLM Output	`params.token`	`params.threshold_percent`
Validator	`params.token`	`params.threshold_percent`
Backtest/Simulate Engines	`token` (flat)	`threshold` (flat)
Frontend Types	`token`	`threshold`

Solution

Flatten LLM output to match what engines and frontend expect. This avoids having to update engines and frontend which are harder to change safely.

Target Schema

{
  "conditions": [
    {
      "type": "price_drop",
      "token": "PEPE",
      "chain": "bsc",
      "threshold": 5,
      "timeframe": "1h"
    }
  ],
  "actions": [...],
  "risk_management": {...}
}

Acceptance Criteria

LLM connector outputs flat structure
Validator validates flat structure
Backtest engine correctly reads conditions
Simulate engine correctly reads conditions

## Problem The LLM outputs a **nested `params`** structure, but the backtest and simulation engines expect **flat fields**. This is a complete pipeline break - AI-parsed strategies will never trigger any trades. ### Current Flow ``` LLM Output: { type: "price_drop", params: { token: "PEPE", threshold_percent: 5 } } Validator: Expects params.threshold_percent Engines: Look for flat threshold -> Returns 0, conditions never trigger Frontend: Uses flat structure { token, threshold } ``` ### Schema Comparison | Component | Token Field | Threshold Field | |-----------|-------------|-----------------| | LLM Output | `params.token` | `params.threshold_percent` | | Validator | `params.token` | `params.threshold_percent` | | Backtest/Simulate Engines | `token` (flat) | `threshold` (flat) | | Frontend Types | `token` | `threshold` | ## Solution Flatten LLM output to match what engines and frontend expect. This avoids having to update engines and frontend which are harder to change safely. ### Target Schema ```json { "conditions": [ { "type": "price_drop", "token": "PEPE", "chain": "bsc", "threshold": 5, "timeframe": "1h" } ], "actions": [...], "risk_management": {...} } ``` ## Acceptance Criteria - [ ] LLM connector outputs flat structure - [ ] Validator validates flat structure - [ ] Backtest engine correctly reads conditions - [ ] Simulate engine correctly reads conditions

shoko referenced this issue from a commit

2026-04-09 09:31:36 +02:00

fix: flatten strategy config schema to match engine expectations

shoko referenced a pull request that will close this issue

2026-04-09 09:32:43 +02:00

fix: flatten strategy config schema (backtesting broken) #35

shoko closed this issue

2026-04-09 09:32:50 +02:00

Sign in to join this conversation.