ConnectOnionConnectOnion

1.9.0a8 — what 1.9.0a7's acceptance run found

An opt-in preview of 1.9.0. Stable is 1.8.9. Everything here was found by running 1.9.0a7 on a copy of the owner's notebook.

  • No more paying for the same refusal. 1.9.0a7 stopped the loop for project pages; people and organisation pages near the 20k limit still went round it, refused and retried on the same mail (643k and 933k tokens). Now every one-page turn is told the page's size and the limit before it starts, and a refused investigation waits until new mail arrives. A timeout or crash is still retried (#2041).
  • The budget meter moves. "0 of 10 points" stayed at 0 after millions of tokens, because Codex's weekly meter reports whole percents and each run used less than one. A run now counts the meter's move when there is one, and otherwise its fresh tokens: a point is a million (input minus cached, plus output). Investigations you start by hand count too (#1990).
  • Investigations read what init already saved. An unfinished mail archive was ignored until it completed; now its saved bodies are read, coverage says how many of how many are saved, and resuming the archive prints its progress (#2042).
  • co rem sync ends on a summary: one line per page as it finishes, then what was new, refused and left, the tokens, and where the full record is, instead of a field dump of about 200 lines (#2044).
  • Every run's tokens land in a usage stage, so the stages add up to the total (#2043); the queue counts a page written by sync as read, and an organisation counts only people worth investigating (#2046); the status line keeps short source labels (#2045, in part).

Install

python -m pip install --upgrade 'connectonion==1.9.0a8'
co rem status

View Markdown source

Star us on GitHub

If ConnectOnion saves you time, a ⭐ goes a long way — and earns you a coffee chat with our founder.