Boss AI
A GitHub project imported from techadnank9/boss.
Stars
0
Forks
0
Language
Python
Repo updated
Nov 21, 2026
Stack
0 tools
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.
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
