@1endpoint/coding-helper 0.1.0 → 0.1.2
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 +128 -75
- package/docs/PUBLISHING.md +153 -0
- package/docs/RELEASE.md +3 -1
- package/package.json +2 -2
package/README.md
CHANGED
|
@@ -1,123 +1,176 @@
|
|
|
1
1
|
# @1endpoint/coding-helper
|
|
2
2
|
|
|
3
|
-
|
|
3
|
+
Connect **DeepSeek Harness** or **Claude Code** to 1endpoint with an interactive setup menu. Choose your language, select models, and manage your connection without editing configuration files by hand.
|
|
4
4
|
|
|
5
|
-
|
|
6
|
-
|
|
5
|
+
## Before you start
|
|
6
|
+
|
|
7
|
+
- Install **Node.js 22.19 or newer**.
|
|
8
|
+
- Install [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness) or [Claude Code](https://code.claude.com/docs/en/overview) separately. The helper configures these tools; it does not install them.
|
|
9
|
+
- Have a 1endpoint API key ready. Keep it private: do not paste it into command arguments, shared configuration files or a repository.
|
|
10
|
+
|
|
11
|
+
## 1. Open the helper
|
|
12
|
+
|
|
13
|
+
Run this in your terminal:
|
|
14
|
+
|
|
15
|
+
```bash
|
|
16
|
+
npx @1endpoint/coding-helper
|
|
7
17
|
```
|
|
8
18
|
|
|
9
|
-
|
|
19
|
+
1. Choose **English** or **Tiếng Việt**.
|
|
20
|
+
2. Choose **Setup**.
|
|
21
|
+
3. Select **DSH** or **Claude Code**.
|
|
22
|
+
4. Select models from the 1endpoint catalog.
|
|
23
|
+
5. Review the selection, then choose **Confirm**.
|
|
24
|
+
|
|
25
|
+
Use **↑ / ↓** to move and **Enter** to select. Press **Escape** to cancel. Confirmation is a selection, not a typed `y/n` answer.
|
|
26
|
+
|
|
27
|
+
### Choosing models
|
|
28
|
+
|
|
29
|
+
For DSH, choose the model you want to use by default.
|
|
30
|
+
|
|
31
|
+
For Claude Code, choose a 1endpoint model for each tier: **Haiku**, **Sonnet** and **Opus**. Each tier can use a different model. Sonnet is used as the main/default selection. Select optional **Fable** mapping only if your installed Claude Code version supports that tier; otherwise keep the standard three-tier option.
|
|
10
32
|
|
|
11
|
-
|
|
33
|
+
If the catalog cannot be loaded, choose **Retry**, explicitly choose **Manual model ID**, or **Cancel**. The helper does not silently substitute a model.
|
|
12
34
|
|
|
13
|
-
|
|
14
|
-
- Install the desired coding tool separately from its official distribution. The helper does not download tools or run installation scripts.
|
|
15
|
-
- An enabled public model ID from the [1endpoint model catalog](https://console.1endpoint.dev/models).
|
|
16
|
-
- Set `ONEENDPOINT_API_KEY` in the launching process environment using your shell or secret manager. Never put the key in command arguments, a Git repository, or a shared configuration file.
|
|
35
|
+
## 2. Provide your API key
|
|
17
36
|
|
|
18
|
-
|
|
37
|
+
The helper does **not** save your API key. Set `ONEENDPOINT_API_KEY` in the terminal where you will launch the tool, using your secret manager or a hidden prompt.
|
|
19
38
|
|
|
20
|
-
|
|
39
|
+
For **Bash**, the following prompt keeps the key out of the command itself and hides it while you type:
|
|
21
40
|
|
|
22
41
|
```bash
|
|
23
|
-
|
|
24
|
-
|
|
25
|
-
|
|
26
|
-
|
|
27
|
-
|
|
28
|
-
|
|
29
|
-
|
|
30
|
-
|
|
31
|
-
|
|
42
|
+
read -r -s -p "1endpoint API key: " ONEENDPOINT_API_KEY
|
|
43
|
+
printf '\n'
|
|
44
|
+
export ONEENDPOINT_API_KEY
|
|
45
|
+
```
|
|
46
|
+
|
|
47
|
+
For another shell, use its secure environment-variable mechanism or your secret manager. Do not type a real key into a command you intend to share. A key set in one terminal is not automatically available in another terminal or after restarting the shell.
|
|
48
|
+
|
|
49
|
+
## 3. Check the connection setup
|
|
50
|
+
|
|
51
|
+
For DSH:
|
|
52
|
+
|
|
53
|
+
```bash
|
|
54
|
+
npx @1endpoint/coding-helper doctor --tool dsh
|
|
32
55
|
```
|
|
33
56
|
|
|
34
|
-
|
|
57
|
+
For Claude Code:
|
|
35
58
|
|
|
36
59
|
```bash
|
|
37
|
-
|
|
38
|
-
yarn test
|
|
39
|
-
node src/cli.mjs setup --tool dsh --model your-model-id --dry-run
|
|
40
|
-
node src/cli.mjs setup --tool dsh --model your-model-id
|
|
60
|
+
npx @1endpoint/coding-helper doctor --tool claude
|
|
41
61
|
```
|
|
42
62
|
|
|
43
|
-
|
|
63
|
+
Doctor checks your saved connection configuration, whether the tool is installed, and whether the key environment variable is present. It does not validate the key against the server or make an inference request.
|
|
64
|
+
|
|
65
|
+
## 4. Start a new tool session
|
|
44
66
|
|
|
45
|
-
|
|
67
|
+
**After setup or reconfiguration, close the old tool session and start a new session through the helper. An already-open session does not pick up the new connection or model mapping.**
|
|
46
68
|
|
|
47
|
-
|
|
69
|
+
Start DSH:
|
|
48
70
|
|
|
49
|
-
|
|
71
|
+
```bash
|
|
72
|
+
npx @1endpoint/coding-helper run --tool dsh
|
|
73
|
+
```
|
|
50
74
|
|
|
51
|
-
|
|
75
|
+
Start Claude Code:
|
|
52
76
|
|
|
53
77
|
```bash
|
|
54
|
-
coding-helper
|
|
55
|
-
coding-helper run --tool dsh
|
|
56
|
-
coding-helper run --tool dsh --profile headless -- 'Reply exactly OK'
|
|
57
|
-
coding-helper doctor --tool claude
|
|
58
|
-
coding-helper run --tool claude
|
|
59
|
-
coding-helper run --tool claude -- -p 'Reply exactly OK'
|
|
78
|
+
npx @1endpoint/coding-helper run --tool claude
|
|
60
79
|
```
|
|
61
80
|
|
|
62
|
-
|
|
81
|
+
You can also reopen the helper and select **Run**.
|
|
63
82
|
|
|
64
|
-
|
|
83
|
+
**Use the helper to launch the configured session.** Running `dsh` or `claude` directly uses the tool's ordinary configuration, not this helper connection. Tool requests may incur charges; the tool's normal permission and workspace-trust prompts remain active.
|
|
65
84
|
|
|
66
|
-
###
|
|
85
|
+
### Important Claude Code behavior
|
|
67
86
|
|
|
68
|
-
|
|
69
|
-
- Configures the native `llm-pi-ai` provider with `api: openai-completions`, `baseURL` and `apiKeyEnv: ONEENDPOINT_API_KEY`.
|
|
70
|
-
- Configures the native `agent-default-model` selection. No secret interpolation, custom proxy or injected agent instruction is required.
|
|
71
|
-
- Supported profiles: `web`, `headless`, `acp`, `sdk`. `sdk-minimal` lacks the required entries and is deliberately rejected.
|
|
72
|
-
- The helper does not reuse current sessions, plugins, model presets or provider configurations. Existing sessions can retain their own model selection; start a new session in the isolated profile.
|
|
73
|
-
- DSH owns request assembly, conversation state, streaming, tool loops and permissions. The helper only sets the connection.
|
|
74
|
-
- Verified configuration contract: DSH `0.2.1-alpha.3`. This prerelease contract may change. A successful dump exit code alone is insufficient: the integration tests also inspect provider and selected-model rows. The shipped Web schema dump has a separate preset/tree-carrier failure; headless schema validation is the supported check.
|
|
87
|
+
The helper uses an isolated configuration. **Project/local settings, including their hooks and permissions, are not loaded in this mode.** Managed organization policy still applies, and the helper does not enable permission bypass. If your project requires those local restrictions or hooks, use your organization's approved configuration instead.
|
|
75
88
|
|
|
76
|
-
###
|
|
89
|
+
### Important DSH behavior
|
|
77
90
|
|
|
78
|
-
|
|
79
|
-
- **Project and local settings, including their hooks and permissions, are intentionally not loaded.** This is a tradeoff for deterministic routing, not transparent preservation of project behavior. Managed/organization policy still applies and can override the connection. Do not use this mode if those project restrictions are required; configure the tool manually with your administrator instead.
|
|
80
|
-
- The raw process key becomes `ANTHROPIC_AUTH_TOKEN` in the child environment only. The CLI constructs the Bearer header. No key is written to settings or included in argv.
|
|
81
|
-
- Maps main, Opus, Sonnet, Haiku, optional Fable and default subagent roles. The wizard offers distinct catalog selections per tier; scripted setup accepts `--haiku-model`, `--sonnet-model`, `--opus-model` and optional `--fable-model`. `--model` is the main/default fallback; explicit role selections take priority. Without a Fable mapping it follows the Opus selection. Explicit model IDs in agent/skill definitions or managed policy are outside that default mapping; no promise is made to rewrite every model choice.
|
|
82
|
-
- Clears inherited conflicting Anthropic/cloud authentication and routing variables in the child only. It does not log out, modify the parent's environment, bypass permission prompts or execute existing key helpers.
|
|
83
|
-
- Uses the same universal base URL above. Claude appends `/v1/messages`, producing `/api/v1/v1/messages`, a supported gateway alias. Do not change the base to a different per-client URL.
|
|
84
|
-
- Claude owns request assembly, conversation state, streaming and tool execution. Version verification and fixture boundaries are recorded in the release checklist.
|
|
91
|
+
DSH runs with a separate home, so your ordinary sessions, plugins and provider settings are not imported or changed. Supported profiles are `web` (default), `headless`, `acp` and `sdk`; `sdk-minimal` is not supported.
|
|
85
92
|
|
|
86
|
-
##
|
|
93
|
+
## Change models
|
|
94
|
+
|
|
95
|
+
Open the helper again, select **Reconfigure**, choose the tool, and select new models:
|
|
96
|
+
|
|
97
|
+
```bash
|
|
98
|
+
npx @1endpoint/coding-helper
|
|
99
|
+
```
|
|
100
|
+
|
|
101
|
+
The previous connection configuration is backed up automatically. **Restart the tool session through the helper after confirming the change.**
|
|
102
|
+
|
|
103
|
+
For scripted use, preview and apply a DSH model change:
|
|
87
104
|
|
|
88
105
|
```bash
|
|
89
|
-
coding-helper
|
|
90
|
-
coding-helper reconfigure --tool dsh --model
|
|
91
|
-
coding-helper reconfigure --tool dsh --model another-public-model
|
|
92
|
-
coding-helper remove --tool dsh --dry-run
|
|
93
|
-
coding-helper remove --tool dsh --yes
|
|
106
|
+
npx @1endpoint/coding-helper reconfigure --tool dsh --model your-model-id --dry-run
|
|
107
|
+
npx @1endpoint/coding-helper reconfigure --tool dsh --model your-model-id
|
|
94
108
|
```
|
|
95
109
|
|
|
96
|
-
`
|
|
110
|
+
Replace `your-model-id` with the exact public model ID you want to use. Replace `dsh` with `claude` for Claude Code. To set separate Claude tiers explicitly:
|
|
111
|
+
|
|
112
|
+
```bash
|
|
113
|
+
npx @1endpoint/coding-helper reconfigure --tool claude \
|
|
114
|
+
--model your-main-model-id \
|
|
115
|
+
--haiku-model your-fast-model-id \
|
|
116
|
+
--sonnet-model your-balanced-model-id \
|
|
117
|
+
--opus-model your-strong-model-id
|
|
118
|
+
```
|
|
97
119
|
|
|
98
|
-
|
|
120
|
+
An optional `--fable-model` sets Fable explicitly. Without it, Fable follows Opus. Explicit tier selections take priority over the `--model` fallback. Explicit model IDs in your own agent/skill definitions or managed policy are not rewritten by these default mappings.
|
|
99
121
|
|
|
100
|
-
|
|
122
|
+
## Remove the connection
|
|
101
123
|
|
|
102
|
-
|
|
124
|
+
Stop the tool session first. Open the helper, select **Remove**, choose the tool and confirm the removal. You do not need to find or delete configuration files yourself.
|
|
103
125
|
|
|
104
|
-
|
|
126
|
+
Alternatively, preview and confirm removal:
|
|
105
127
|
|
|
106
128
|
```bash
|
|
107
|
-
|
|
108
|
-
|
|
109
|
-
yarn test
|
|
110
|
-
yarn pack --out ../coding-helper-0.1.0.tgz
|
|
129
|
+
npx @1endpoint/coding-helper remove --tool dsh --dry-run
|
|
130
|
+
npx @1endpoint/coding-helper remove --tool dsh --yes
|
|
111
131
|
```
|
|
112
132
|
|
|
113
|
-
|
|
133
|
+
Use `--tool claude` for Claude Code. Removal deletes only the helper-generated connection configuration. **Your original tool settings, runtime conversations and configuration backups are preserved.** To return to your ordinary setup, launch `dsh` or `claude` directly.
|
|
134
|
+
|
|
135
|
+
### Recover from damaged connection configuration
|
|
136
|
+
|
|
137
|
+
If the helper reports that its generated configuration has been modified or corrupted, choose **Force removal recovery** in the Remove menu, review the warning and confirm. Then run Setup again.
|
|
138
|
+
|
|
139
|
+
For scripted recovery:
|
|
140
|
+
|
|
141
|
+
```bash
|
|
142
|
+
npx @1endpoint/coding-helper remove --tool dsh --force --yes
|
|
143
|
+
```
|
|
144
|
+
|
|
145
|
+
Force removal deletes only the fixed helper connection files, including modified contents, without backing up those modified contents. It still preserves original tool settings and runtime conversations. Use it only for recovery, not routine reconfiguration.
|
|
146
|
+
|
|
147
|
+
## Configuration and privacy
|
|
148
|
+
|
|
149
|
+
- The shared API base URL is `https://1endpoint.dev/api/v1`.
|
|
150
|
+
- Connection configuration is stored under `~/.1endpoint/coding-helper`. Set an absolute `ONEENDPOINT_HELPER_HOME` if you need a different location.
|
|
151
|
+
- Setup does not overwrite your existing DSH or Claude Code configuration.
|
|
152
|
+
- The interactive menu fetches the public text-model catalog without sending your API key. Explicit setup, reconfigure and doctor commands work locally.
|
|
153
|
+
- Your key is supplied to the launched tool through its process environment, not saved in the helper's connection files.
|
|
154
|
+
- Runtime tools can create their own session data. Keep their isolated homes private.
|
|
155
|
+
- Configuration isolation is **not an operating-system sandbox**; the tools retain their ordinary filesystem and network access.
|
|
114
156
|
|
|
115
|
-
|
|
157
|
+
This is an independent integration, not an official DeepSeek or Anthropic product.
|
|
116
158
|
|
|
117
|
-
|
|
159
|
+
## Troubleshooting
|
|
118
160
|
|
|
119
|
-
|
|
161
|
+
| Problem | What to do |
|
|
162
|
+
| --- | --- |
|
|
163
|
+
| Node version error | Install Node.js 22.19 or newer, then reopen your terminal. |
|
|
164
|
+
| CLI not found | Install the selected coding tool and make sure its executable is on your `PATH`. |
|
|
165
|
+
| API key missing | Set `ONEENDPOINT_API_KEY` in the same terminal, then run Doctor again. |
|
|
166
|
+
| Catalog unavailable | Choose Retry, or explicitly choose manual entry with a known public model ID. |
|
|
167
|
+
| New model is not taking effect | Close the old session and start a new one with the helper's Run action. |
|
|
168
|
+
| Ordinary tool launch ignores the helper | Launch through `npx @1endpoint/coding-helper run --tool dsh` or `--tool claude`. |
|
|
169
|
+
| Generated configuration is damaged | Use Remove → Force removal recovery, then Setup again. |
|
|
170
|
+
| Authentication or model request fails | Confirm your key, model access and account balance in 1endpoint. Doctor does not validate server-side access. |
|
|
120
171
|
|
|
121
|
-
##
|
|
172
|
+
## More help
|
|
122
173
|
|
|
123
|
-
|
|
174
|
+
- [Console usage guide](https://console.1endpoint.dev/docs/guides/coding-tools/coding-helper)
|
|
175
|
+
- [Report a problem](https://github.com/dothinh115/1endpoint-coding-helper/issues)
|
|
176
|
+
- Maintainers: [publishing](docs/PUBLISHING.md) and [release verification](docs/RELEASE.md).
|
|
@@ -0,0 +1,153 @@
|
|
|
1
|
+
# Publishing @1endpoint/coding-helper to npm
|
|
2
|
+
|
|
3
|
+
This guide is for maintainers publishing the package, not for end users configuring a coding tool. For tool setup, see the [README](../README.md). For implementation and verification boundaries, see the [release checklist](RELEASE.md).
|
|
4
|
+
|
|
5
|
+
**Publishing is a separate action from pushing to GitHub.** Run the publish command only when you intend to release the selected version publicly. This document does not mean that a release has already been published.
|
|
6
|
+
|
|
7
|
+
## 1. Requirements
|
|
8
|
+
|
|
9
|
+
- Node.js **22.19.0 or newer**.
|
|
10
|
+
- Yarn **4.17.0**, matching the repository's `packageManager` setting.
|
|
11
|
+
- An npm account authorized to publish under the **`@1endpoint`** scope.
|
|
12
|
+
- Access to the account's required two-factor authentication method.
|
|
13
|
+
- A reviewed, committed checkout of [the repository](https://github.com/dothinh115/1endpoint-coding-helper).
|
|
14
|
+
|
|
15
|
+
The npm scope and GitHub owner are independent. Owning the GitHub repository does not grant npm publishing rights. If `1endpoint` is an npm organization, the publishing account must be a member with package creation/publish permissions. Create or obtain access to that scope before continuing; do not change the package name merely to work around a permission failure.
|
|
16
|
+
|
|
17
|
+
Never put npm passwords, access tokens, recovery codes, or one-time codes in source files, issue comments, screenshots, or chat. Use the package manager's authentication flow. The helper's `ONEENDPOINT_API_KEY` is **not** an npm credential and is not needed for publication.
|
|
18
|
+
|
|
19
|
+
## 2. Open the checkout and authenticate
|
|
20
|
+
|
|
21
|
+
From this workspace:
|
|
22
|
+
|
|
23
|
+
```bash
|
|
24
|
+
cd /Users/thinhdo/Documents/1endpoint/coding-helper
|
|
25
|
+
node --version
|
|
26
|
+
yarn --version
|
|
27
|
+
|
|
28
|
+
yarn npm login --scope 1endpoint --publish
|
|
29
|
+
yarn npm whoami --scope 1endpoint --publish
|
|
30
|
+
```
|
|
31
|
+
|
|
32
|
+
Follow the login prompts privately. The final command should identify the intended publishing account. Authentication alone does not prove permission to create this scoped package.
|
|
33
|
+
|
|
34
|
+
## 3. Review the version and source
|
|
35
|
+
|
|
36
|
+
```bash
|
|
37
|
+
git status --short
|
|
38
|
+
git log -1 --oneline
|
|
39
|
+
node --input-type=module -e 'import fs from "node:fs"; const p = JSON.parse(fs.readFileSync("package.json", "utf8")); console.log(`${p.name}@${p.version}`)'
|
|
40
|
+
```
|
|
41
|
+
|
|
42
|
+
The worktree should be clean, and the name/version must be the release you intend to publish. The initial package version is `0.1.0`.
|
|
43
|
+
|
|
44
|
+
Optionally inspect the registry before releasing:
|
|
45
|
+
|
|
46
|
+
```bash
|
|
47
|
+
yarn npm info @1endpoint/coding-helper --fields name,version,versions
|
|
48
|
+
```
|
|
49
|
+
|
|
50
|
+
Before the first publication, a package-not-found response can be expected. It does **not** establish that your account owns the scope. If the target version already exists, do not retry publishing that same version: published versions are immutable, even after unpublishing. Prepare a new version instead.
|
|
51
|
+
|
|
52
|
+
## 4. Run the release checks
|
|
53
|
+
|
|
54
|
+
```bash
|
|
55
|
+
yarn install --immutable
|
|
56
|
+
yarn check
|
|
57
|
+
yarn test
|
|
58
|
+
yarn pack --dry-run
|
|
59
|
+
git diff --check
|
|
60
|
+
```
|
|
61
|
+
|
|
62
|
+
There is no separate compilation step: the package ships native ESM source. `prepack` also runs the syntax check and test suite when packaging/publishing.
|
|
63
|
+
|
|
64
|
+
The ordinary test suite deliberately skips the opt-in installed-Claude loopback fixture. To run that fixture on a machine with Claude Code installed, set its absolute executable path:
|
|
65
|
+
|
|
66
|
+
```bash
|
|
67
|
+
CODING_HELPER_TEST_CLAUDE="$(command -v claude)" yarn test
|
|
68
|
+
```
|
|
69
|
+
|
|
70
|
+
This fixture uses a local synthetic server and dummy credentials, not paid gateway inference. Review skips and failures; a test run does not establish support for operating systems or CLI versions that were not exercised.
|
|
71
|
+
|
|
72
|
+
Inspect the package file list. It should contain only package metadata, the license, README, documentation and the intended source modules. Do not ship environment files, credentials, helper runtime homes, test output or unrelated workspace files.
|
|
73
|
+
|
|
74
|
+
## 5. Publish the package publicly
|
|
75
|
+
|
|
76
|
+
For the initial stable release:
|
|
77
|
+
|
|
78
|
+
```bash
|
|
79
|
+
yarn npm publish --access public --tag latest
|
|
80
|
+
```
|
|
81
|
+
|
|
82
|
+
If npm requests an authenticator code, follow the prompt or provide a current code using the supported option:
|
|
83
|
+
|
|
84
|
+
```bash
|
|
85
|
+
yarn npm publish --access public --tag latest --otp YOUR_CURRENT_CODE
|
|
86
|
+
```
|
|
87
|
+
|
|
88
|
+
Replace the placeholder privately. Be aware that command arguments may be visible in shell history or process listings; prefer an interactive prompt when available. Never copy the real code into documentation or chat.
|
|
89
|
+
|
|
90
|
+
`--access public` is intentional for this scoped package. `--tag latest` makes the release the default installed version. For a prerelease, use a prerelease version and a separate tag such as `next`; do not accidentally move `latest` to an experimental build.
|
|
91
|
+
|
|
92
|
+
**Do not repeat the publish command blindly after an ambiguous network failure.** Check whether the version reached the registry first.
|
|
93
|
+
|
|
94
|
+
## 6. Verify the public release
|
|
95
|
+
|
|
96
|
+
For version `0.1.0`:
|
|
97
|
+
|
|
98
|
+
```bash
|
|
99
|
+
yarn npm info @1endpoint/coding-helper --fields name,version,bin,dist
|
|
100
|
+
npx @1endpoint/coding-helper@0.1.0 --version
|
|
101
|
+
npx @1endpoint/coding-helper@0.1.0 --help
|
|
102
|
+
npx @1endpoint/coding-helper@0.1.0
|
|
103
|
+
```
|
|
104
|
+
|
|
105
|
+
Verify:
|
|
106
|
+
|
|
107
|
+
1. Registry metadata reports the expected name and version, binary entries, and tarball integrity.
|
|
108
|
+
2. `--version` prints `0.1.0` and `--help` lists the documented commands.
|
|
109
|
+
3. The interactive entry asks for **English / Tiếng Việt first**.
|
|
110
|
+
4. Arrow keys and Enter select actions; confirmation does not require typing `y/n`.
|
|
111
|
+
5. Setup offers the public 1endpoint text-model catalog, with independent Claude tier mappings.
|
|
112
|
+
6. Setup/reconfigure emphasizes restarting the session through the helper.
|
|
113
|
+
|
|
114
|
+
Do the interactive smoke in a fresh environment or set `ONEENDPOINT_HELPER_HOME` to a new absolute test directory. Setup does not require a real API key. Do not launch inference merely to verify that npm installation works. Use the helper's `remove` action to remove its generated connection configuration afterward; runtime data is intentionally preserved.
|
|
115
|
+
|
|
116
|
+
Registry propagation and local package-manager caches can briefly lag. Check the exact pinned version before assuming a publish failed. After propagation, also verify the unpinned entry resolves to the intended `latest` release:
|
|
117
|
+
|
|
118
|
+
```bash
|
|
119
|
+
npx @1endpoint/coding-helper --version
|
|
120
|
+
```
|
|
121
|
+
|
|
122
|
+
## 7. Finish the release handoff
|
|
123
|
+
|
|
124
|
+
- Record the released version, source commit, registry integrity and verification outcome.
|
|
125
|
+
- Ensure public setup documentation matches the released CLI rather than unreleased local changes.
|
|
126
|
+
- Deploy Console documentation separately through its normal pipeline if needed. Removing a Console warning does not publish an npm package or prove registry availability.
|
|
127
|
+
- Keep live-provider inference, browser/mobile QA and operating-system coverage distinct from package installation checks.
|
|
128
|
+
|
|
129
|
+
## Publishing an update
|
|
130
|
+
|
|
131
|
+
1. Make and review the change.
|
|
132
|
+
2. Update the `version` in `package.json` to a new semantic version. Use a patch release for compatible fixes, a minor release for compatible features, and a major release for breaking changes. Pre-1.0 compatibility expectations should still be documented explicitly.
|
|
133
|
+
3. Run `yarn install` to reconcile the lockfile if necessary, then repeat the checks above.
|
|
134
|
+
4. Commit and push the reviewed release source when authorized.
|
|
135
|
+
5. Publish, then verify the exact new version and the intended distribution tag.
|
|
136
|
+
|
|
137
|
+
Never overwrite an existing version. For a bad release, prepare a fixed version; consider npm deprecation according to the project's release policy rather than treating unpublish as rollback.
|
|
138
|
+
|
|
139
|
+
## Troubleshooting
|
|
140
|
+
|
|
141
|
+
| Symptom | What to check |
|
|
142
|
+
| --- | --- |
|
|
143
|
+
| `401`, `ENEEDAUTH`, or invalid credentials | Repeat `yarn npm login --scope 1endpoint --publish`, then check `whoami`. Confirm the publishing registry and account. |
|
|
144
|
+
| `403` or scope/package permission denied | Confirm organization membership, package creation/publish rights, account security requirements and token restrictions if using token authentication. Login alone is insufficient. |
|
|
145
|
+
| Package not found before the first release | Expected for an unpublished package, but also possible with the wrong registry or inaccessible package. Verify scope ownership and registry configuration. |
|
|
146
|
+
| Version already published | Select a new version; an existing version cannot be replaced. |
|
|
147
|
+
| OTP invalid or expired | Use a fresh code from the publishing account and check the device clock. Do not weaken account security. |
|
|
148
|
+
| `prepack` fails | Fix the syntax/test failure and rerun the release checks. Do not bypass the release gate to publish. |
|
|
149
|
+
| Publish timed out or disconnected | Query the exact version in the registry before attempting another publish. |
|
|
150
|
+
| `npx` cannot find the release immediately | Check exact registry metadata, allow propagation time, and use the pinned version. Confirm you did not publish only under a prerelease tag. |
|
|
151
|
+
| CLI exits because Node is unsupported | Install a Node version satisfying `>=22.19.0` and retry. |
|
|
152
|
+
|
|
153
|
+
For CI automation, configure npm trusted publishing or appropriately restricted credentials using current npm and CI-provider documentation. Do not assume this local Yarn login procedure configures CI identity, provenance or release approvals.
|
package/docs/RELEASE.md
CHANGED
|
@@ -1,5 +1,7 @@
|
|
|
1
1
|
# Release and verification gate
|
|
2
2
|
|
|
3
|
+
For the complete maintainer procedure, follow the [npm publishing guide](PUBLISHING.md): account/scope permissions, login, release checks, publication, pinned npx verification, updates and troubleshooting.
|
|
4
|
+
|
|
3
5
|
## Scope
|
|
4
6
|
|
|
5
7
|
Initial package: `@1endpoint/coding-helper@0.1.0`, public npm access. Independent repository: `dothinh115/1endpoint-coding-helper`. First Git commit: `first push`. **Git push does not authorize npm publication.** Maintainer publishes after reviewing the handoff.
|
|
@@ -34,7 +36,7 @@ Console checks: focused coding-helper documentation regressions, scoped ESLint,
|
|
|
34
36
|
- Verify the npm account has publish rights for scope `@1endpoint` and satisfies npm 2FA/trusted-publishing requirements. Never copy npm credentials into repository or chat.
|
|
35
37
|
- Re-run the checks above from the committed source.
|
|
36
38
|
- Maintainer may publish with Yarn's npm publish command, public access, and the desired release tag. This task deliberately does not run it.
|
|
37
|
-
- After publication verify registry version, integrity, bin mapping and real `npx @1endpoint/coding-helper --version` from a fresh environment
|
|
39
|
+
- After publication verify registry version, integrity, bin mapping and real `npx @1endpoint/coding-helper --version` from a fresh environment before announcing availability. The Console unpublished warning was removed separately at the user's request; that presentation change is not proof of publication. Commit/deploy Console changes through its normal pipeline when authorized.
|
|
38
40
|
|
|
39
41
|
## Verification record
|
|
40
42
|
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "@1endpoint/coding-helper",
|
|
3
|
-
"version": "0.1.
|
|
3
|
+
"version": "0.1.2",
|
|
4
4
|
"description": "Safe isolated 1endpoint setup for DeepSeek Harness and Claude Code CLI",
|
|
5
5
|
"type": "module",
|
|
6
6
|
"bin": {
|
|
@@ -42,4 +42,4 @@
|
|
|
42
42
|
"cli",
|
|
43
43
|
"setup"
|
|
44
44
|
]
|
|
45
|
-
}
|
|
45
|
+
}
|