Details: I've been extending Polestar in a fork (github.com/petllama/polestar) so it can express the kind of guide logic that commercial guides like RestedXP rely on. Everything is generic, declarative and meant to go upstream if you want it.
What's in it:
• Conditions: a step "when" line with a small closed expression language (quest state, items, level, skill, rep, zone, class/race/faction, and/or/not). No arbitrary Lua.
• Along-the-way objectives: "along [until <label>]" keeps background goals in a side panel without blocking the route. Click one to aim the arrow at it.
• Labels and jumps: "label", "skipto <label> [when …]" and "done <label>", with branch choices saved per character.
• Conditional hand-offs: "next <guide> when …" hides choices that don't apply and continues automatically when only one is left.
• Play modes: hardcore, self-found, auction house, group content and an XP-rate slider in options, plus xp/mode checks and goals like "level 12.5".
• Loops: a "loop" step turns its waypoints into a search circuit, using the existing spawn-area runtime.
Format-1 packs are untouched (byte-identical output, with a regression test). New keys mark a pack as format 2, and the format number and version are placeholders for you to decide. Each feature is a small branch with tests first; the full suite passes.
Happy to open PRs one feature at a time if you're interested. Fun fact: the code was written by a local Qwen model on my own GPU, with a bigger model planning and reviewing. This was written by opus for clarity.
1
1
78