swarnimsandeep1 day agoContributor

Boss AI

A GitHub project imported from techadnank9/boss.

Check your own project

Stars

0

Forks

0

Language

Python

Repo updated

Nov 21, 2026

Stack

0 tools

Launch note

Built

Boss AI

Proof

repo / build notes

Next

Try it, remix the project, or leave feedback.

swarnimsandeep published this project so visitors can inspect the work and leave feedback.

Product page from repo

Auto-created from the repository README and metadata so builders, buyers, and agents can scan what this project does before opening the code.

overview

What it is

A GitHub project imported from techadnank9/boss.

quickstart

Run

Without SINGLESTORE_URL the trail is kept in process memory so you can test the flow. Set it before recording the demo so the trail, scoreboard, and lessons persist.

architecture

How it works

is too vague to buy the right thing (for example no quantity), Boss asks one question instead of guessing, and no agent runs. Boss stops and asks. If the item is out of stock, Boss falls back to the best runner-up and asks you to review it. Otherwise it logs your approval and opens the real product page. Boss never pays or checks out for you. agent and the judge receive on all future runs, and the agent's trust score drops.

  • POST /shop logs the request, loads the agent track record and past lessons. If the request
  • Agents A, B, and C each run a live web search with a different strategy and submit one real bid.
  • Boss scores every bid, rejects rule breakers, and picks a winner with a one line reason each.
  • POST /approve/{run_id} first re-reads the live product page. If the price rose by $5 or more,
  • POST /feedback/{run_id} records an overrule. The reason becomes a lesson that every
Build Notes
Imported from https://github.com/techadnank9/boss. One agent that manages your buying agents. Three real buying agents search the live web and bid. Boss judges the bids against your rules, shows the full audit trail, and learns when you overrule it.
Feedback