gwqadd 0.5.1 → 0.5.2
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 +7 -5
- package/bin/gwqadd.mjs +9 -7
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -139,13 +139,15 @@ One question, one confirmation:
|
|
|
139
139
|
**The AI** is whichever of these is on your `PATH`:
|
|
140
140
|
|
|
141
141
|
```
|
|
142
|
-
claude -p → codex exec (if Claude fails) → opencode run → gemini -p
|
|
142
|
+
claude -p → codex exec --skip-git-repo-check (if Claude fails) → opencode run → gemini -p
|
|
143
143
|
```
|
|
144
144
|
|
|
145
|
-
Automatic detection tries Claude first and falls back to
|
|
146
|
-
fails or returns no usable branch
|
|
147
|
-
|
|
148
|
-
|
|
145
|
+
Automatic detection tries Claude first and falls back to
|
|
146
|
+
`codex exec --skip-git-repo-check` when it fails or returns no usable branch
|
|
147
|
+
name, then continues through the other available CLIs. The flag is required
|
|
148
|
+
because the AI runs in an empty directory rather than the user's repository.
|
|
149
|
+
An explicit `--ai '<cmd>'` or `GWQADD_AI='<cmd>'` selects one command and keeps
|
|
150
|
+
the existing manual fallback if that command fails.
|
|
149
151
|
|
|
150
152
|
No API key to obtain, no account to create — it uses what you already have.
|
|
151
153
|
Expect 6–8 seconds, almost all of it the CLI's own start-up; an elapsed counter
|
package/bin/gwqadd.mjs
CHANGED
|
@@ -76,12 +76,12 @@ NAMING HELP
|
|
|
76
76
|
modified, so the AI picks the prefix and the wording that match the repo. A
|
|
77
77
|
rejected suggestion is passed back as an exclusion, so "n" does not return it.
|
|
78
78
|
|
|
79
|
-
Automatic detection tries claude -p first, then codex exec
|
|
80
|
-
returns no usable name, followed by
|
|
81
|
-
|
|
82
|
-
confirm. Override with --ai '<cmd>'
|
|
83
|
-
explicitly; disable with --no-ai or
|
|
84
|
-
prompt for an ASCII name.
|
|
79
|
+
Automatic detection tries claude -p first, then codex exec
|
|
80
|
+
--skip-git-repo-check if Claude fails or returns no usable name, followed by
|
|
81
|
+
opencode run and gemini -p. Nothing is sent anywhere until you answer the
|
|
82
|
+
question, and nothing is created until you confirm. Override with --ai '<cmd>'
|
|
83
|
+
or GWQADD_AI='<cmd>' to use one command explicitly; disable with --no-ai or
|
|
84
|
+
GWQADD_AI=off, which leaves a plain prompt for an ASCII name.
|
|
85
85
|
|
|
86
86
|
--no-random (or GWQADD_RANDOM=off) starts at the description prompt instead.
|
|
87
87
|
--random goes the other way and never asks; it is the only naming path that
|
|
@@ -1229,7 +1229,9 @@ function namingPrompt(ctx, description, rejected) {
|
|
|
1229
1229
|
// answers a bare prompt headlessly.
|
|
1230
1230
|
const AI_CLIS = [
|
|
1231
1231
|
{ bin: 'claude', args: ['-p'] },
|
|
1232
|
-
|
|
1232
|
+
// runAi() deliberately uses an empty, non-Git directory; Codex requires this
|
|
1233
|
+
// opt-out of its repository check before it will accept that working root.
|
|
1234
|
+
{ bin: 'codex', args: ['exec', '--skip-git-repo-check'] },
|
|
1233
1235
|
{ bin: 'opencode', args: ['run'] },
|
|
1234
1236
|
{ bin: 'gemini', args: ['-p'] },
|
|
1235
1237
|
];
|