Add option to retain cost history beyond the current 30-day window for OpenAI and other providers.

[ChristosBouronikos]

Right now CodexBar only stores cost data for the last 30 days and only for OpenAI.
For users who want to analyze long-term trends, auditing, budgeting, or usage patterns over multiple months, this limitation prevents meaningful historical insight for OpenAI and other providers.

It would be very useful to either extend the retention period and maybe add an option to export cost history (e.g., as a CSV file), so users can keep their own long-term records and perform external analysis.

[steipete]

Do we get accurate cost from other providers?

Current design doesn't store anything, just pulls data.
Storing would be quite a large design change.

[kquinsland]

It would be very useful to either extend the retention period and maybe add an option to export cost history (e.g., as a CSV file), so users can keep their own long-term records and perform external analysis.

I am currently using onwatch for this but would appreciate a basic "record samples to sqlite" function or similar built into codex bar since I always have a macbook on but the host that runs onwatch isn't always able to make API calls.

Storing would be quite a large design change.

Could it be as straight forward as just writing out a @timestamp, provider/account, {json_doc} row to a sqlite file? where json_doc is a simple key: $value object? Each provider has their own buckets (codex calls it session/weekly, gemeni does flash/flash-lite/pro ... etc) so it doesn't make sense to do any normalization and leave it as an exercise to the reader to extract the json and do useful things with it

[steipete]

Fixed on main in d715648c.

Added a configurable 1-365 day cost-history window in Settings and codexbar cost --days, wired OpenAI Admin API charts to that window, paged OpenAI Admin requests within the endpoint bucket limit, and made local Codex/Claude cost caches rescan when the requested window expands, including legacy cache files.

Proof: CLICostTests, CostUsageFetcherTests, OpenAI Admin API tests, focused provider/cost shard, make check, full swift test --disable-sandbox via Codex review, and clean Codex review.