Docs
Documentation and technical details will grow with the project. Check back as ChoicePilot moves beyond Week 0.
Prompt library
Extraction and generation logic used on ChoicePilot is documented here as it's added, whether it's a real model prompt or, as with the entry below, a simulated rule-based stand-in.
core-extraction-v1 — Decision core extraction
Simulated — rule-based, not AIUsed by POST /api/core/extract on /core. Takes a free-text decision description and returns the options and priorities embedded in it, via plain text parsing — no external AI/LLM call this week.
Options logic
If the text contains "between", split the phrase that follows on "and", "or", "vs", and commas. Otherwise split the whole text on "or", "vs", and commas. Each resulting segment is trimmed, and blank or overly long (60+ character) segments are dropped before de-duplicating.
Priorities logic
The text is checked for exact-word matches against a fixed keyword dictionary: price, cost, budget, time, speed, quality, comfort, safety, commute, distance, location, coverage, flexibility, convenience, reliability, support. Every match found is included once.
Example
Input: "I'm choosing between Apartment A and Apartment B, price and commute time matter most"
{
"options": ["Apartment A", "Apartment B"],
"priorities": ["Price", "Time", "Commute"]
}Routing this through an API route, even while simulated, keeps the seam open to swap in a real model call in a later week without changing the /core frontend.