@mostok/codexes 0.2.0 → 0.3.2

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -32,7 +32,7 @@ codexes --help
32
32
  - Shared `CODEX_HOME` for stable `config.toml`, `mcp.json`, and `trust/`
33
33
  - Per-account auth activation for `auth.json` and `sessions/`
34
34
  - Runtime locking to avoid concurrent mutation of the shared Codex home
35
- - Experimental account selection by remaining daily and weekly limits
35
+ - Experimental account selection by remaining primary and secondary window percentages
36
36
  - Packaged npm distribution with smoke coverage for real installs
37
37
 
38
38
  ## Example
@@ -41,10 +41,16 @@ codexes --help
41
41
  codexes account add work
42
42
  codexes account add personal
43
43
  codexes account use work
44
- CODEXES_ACCOUNT_SELECTION_STRATEGY=remaining-limit-experimental codexes chat --model gpt-5
44
+ codexes chat --model gpt-5
45
45
  ```
46
46
 
47
- The experimental selector probes `https://chatgpt.com/backend-api/wham/usage`, compares remaining quota windows, and picks the best usable account before launching the real `codex` binary.
47
+ By default, `codexes` uses the `remaining-limit` selector. It probes `https://chatgpt.com/backend-api/wham/usage`, reads `rate_limit.primary_window` and `rate_limit.secondary_window`, converts `used_percent` into remaining percent, and picks the best usable account before launching the real `codex` binary.
48
+
49
+ If you need compatibility overrides, set `CODEXES_ACCOUNT_SELECTION_STRATEGY` to `manual-default`, `single-account`, or `remaining-limit`. The legacy value `remaining-limit-experimental` is still accepted for backward compatibility.
50
+
51
+ Before launch, `codexes` prints an English account selection summary that includes every configured account, its current status, primary and secondary remaining percent, whether the data came from `fresh` probing or `cache`, and which account was selected. In interactive terminals the account lines are colorized; in plain-text contexts the same output stays ANSI-free. If probing is incomplete or unreliable, the CLI prints an explicit fallback message instead of silently behaving like `manual-default`.
52
+
53
+ `codexes account list` uses the same summary formatter, but stays diagnostic and read-only. When fallback analysis cannot produce a reliable execution winner and no valid default account is available, `account list` still renders the account table plus an execution note instead of failing. Wrapped execution commands remain strict and still require a concrete selected account before launch.
48
54
 
49
55
  ## Documentation
50
56