dsh-codex-connect 0.1.0-alpha.4.23 → 0.1.0-alpha.4.25
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/INSTALL.md +13 -5
- package/README.i18n.yaml +2 -2
- package/README.md +26 -22
- package/compatibility.json +1 -1
- package/cordis.patch.yml +1 -0
- package/docs/README.zh.md +26 -22
- package/docs/auto-review.md +26 -0
- package/docs/auto-review.zh.md +26 -0
- package/docs/design.i18n.yaml +2 -2
- package/docs/design.md +1 -1
- package/docs/design.zh.md +1 -1
- package/lib/bin.js +12 -238
- package/lib/client.js +1537 -611
- package/lib/index.d.ts +14 -4
- package/lib/index.js +1 -1
- package/lib/{src-ByKP_wn9.js → src-Di2xxpDv.js} +876 -15
- package/package.json +80 -73
package/INSTALL.md
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
# Installation Runbook for CLI Agents
|
|
2
2
|
|
|
3
|
-
Alpha 4.
|
|
3
|
+
Alpha 4.25 is verified with DSH `0.1.2-alpha.5` and its declared pi-ai range `^0.84.2`; the verified registry installation resolves pi-ai `0.84.4`.
|
|
4
4
|
|
|
5
5
|
Install `dsh-codex-connect` into one requested DeepSeek Harness profile without changing its current default model, search route, global configuration, or OAuth state.
|
|
6
6
|
|
|
@@ -23,10 +23,11 @@ Check `dsh --version` before changing the requested profile. Use `dsh --help` to
|
|
|
23
23
|
| `0.1.0-rc.7` | `0.1.0-alpha.4.14` |
|
|
24
24
|
| `0.1.1-rc.2` | `0.1.0-alpha.4.21` |
|
|
25
25
|
| `0.1.2-alpha.2` | `0.1.0-alpha.4.23` |
|
|
26
|
+
| `0.1.2-alpha.5` | `0.1.0-alpha.4.25` |
|
|
26
27
|
|
|
27
28
|
If your exact DSH version is unknown or not listed, stop and verify the combination before installing. Do not blindly install `dsh-codex-connect@alpha`: `alpha` is a moving tag, not a compatibility guarantee. Do not infer support for newer DSH versions from these rows.
|
|
28
29
|
|
|
29
|
-
Alpha 4.
|
|
30
|
+
Alpha 4.25's verified contract is DSH plugin API packages `0.1.2-alpha.5`, `@earendil-works/pi-ai` `^0.84.2` (resolved as `0.84.4` during verification), and Node.js `^22.19.0 || >=24.0.0`. Alpha 4.24 remains an earlier verified choice for the same DSH version, Alpha 4.23 remains the verified choice for DSH `0.1.2-alpha.2`, Alpha 4.21 remains the verified choice for DSH `0.1.1-rc.2`, and staying on DSH `0.1.0-rc.7` means selecting Alpha 4.14. Upgrading DSH is a separate decision: upgrade the DSH API packages and pi-ai together, then rerun `dsh-codex-connect doctor --json` and `pnpm --silent run check:compatibility` for the selected combination.
|
|
30
31
|
|
|
31
32
|
These choices reflect the repository's existing verification record, not a new installation or runtime probe. This guidance does not fix upstream DSH compatibility or resolve [Issue #64](https://github.com/franksong2702/dsh-codex-connect/issues/64).
|
|
32
33
|
|
|
@@ -51,7 +52,13 @@ These choices reflect the repository's existing verification record, not a new i
|
|
|
51
52
|
dsh plugin --profile web add dsh-codex-connect@0.1.0-alpha.4.23
|
|
52
53
|
```
|
|
53
54
|
|
|
54
|
-
|
|
55
|
+
For DSH `0.1.2-alpha.5`, use Alpha 4.25:
|
|
56
|
+
|
|
57
|
+
```sh
|
|
58
|
+
dsh plugin --profile web add dsh-codex-connect@0.1.0-alpha.4.25
|
|
59
|
+
```
|
|
60
|
+
|
|
61
|
+
If npm is unavailable after the matching GitHub prerelease is created, use `dsh plugin --profile web add 'github:franksong2702/dsh-codex-connect#v0.1.0-alpha.4.21'` only for the DSH `0.1.1-rc.2` combination, `dsh plugin --profile web add 'github:franksong2702/dsh-codex-connect#v0.1.0-alpha.4.23'` only for the DSH `0.1.2-alpha.2` combination, or `dsh plugin --profile web add 'github:franksong2702/dsh-codex-connect#v0.1.0-alpha.4.25'` only for the DSH `0.1.2-alpha.5` combination.
|
|
55
62
|
|
|
56
63
|
3. Run `dsh --profile web --dump-config` and require exactly one `llm-openai-codex` row loading `dsh-codex-connect`.
|
|
57
64
|
4. Confirm the effective `agent-default-model` and `web.searchProvider` values are unchanged from before installation.
|
|
@@ -63,7 +70,7 @@ These choices reflect the repository's existing verification record, not a new i
|
|
|
63
70
|
|
|
64
71
|
6. If the user explicitly requests login, open **Settings → Plugins → Plugin configuration → Codex Connect**, or check `status` and then use `login` or `login --device-code`. OAuth approval belongs to the user.
|
|
65
72
|
|
|
66
|
-
Alpha 4.
|
|
73
|
+
Alpha 4.25 offers the same account actions in **Settings → Models → Openai-Codex**, plus a shared **More settings** dialog for model visibility, proxy, search, image, context-budget, and Auto-review controls. The original Plugin settings entry remains available; neither entry automatically starts login or changes model/search defaults.
|
|
67
74
|
|
|
68
75
|
When signed out, select **Authorize**. When signed in, use **Sign out** or **View quota**; use **More settings** for plugin options. If authorization is abandoned, use **Reopen authorization** or **Cancel sign-in** and retry; cancellation does not delete an existing account. Pending authorization expires after 10 minutes by default (`oauthTimeoutMs` in plugin configuration, applied on load).
|
|
69
76
|
|
|
@@ -80,7 +87,7 @@ The value is a full `http://` or `https://` origin including its port, not a bar
|
|
|
80
87
|
|
|
81
88
|
## Optional configuration
|
|
82
89
|
|
|
83
|
-
Use **Settings → Plugins → Plugin configuration → Codex Connect** for live, staged Save/Discard edits. The same
|
|
90
|
+
Use **Settings → Plugins → Plugin configuration → Codex Connect** for live, staged Save/Discard edits organized under Account & quota, Models, Network, and Capabilities. Switching modules preserves the draft. The same settings control `enableSearch`, `enableImageTool`, `enableImageGeneration`, and `enableAutoReview`; all four default to `false`. Enabling Auto-review permits bounded approval context, tool arguments, working directory, and the planned action to be sent to `chatgpt.com`; failures return to human approval. Enabling image generation uses the image generation capability included with the current GPT subscription and saves results as DSH attachments. Enabling search registers a provider but does not select it; selecting `web.searchProvider: openai-codex` is a second explicit profile change. Setting `agent-default-model` to `openai-codex` is also a separate explicit change.
|
|
84
91
|
|
|
85
92
|
Apply only requested choices and preserve unrelated keys:
|
|
86
93
|
|
|
@@ -90,6 +97,7 @@ Apply only requested choices and preserve unrelated keys:
|
|
|
90
97
|
enableSearch: true
|
|
91
98
|
enableImageTool: false
|
|
92
99
|
enableImageGeneration: false
|
|
100
|
+
enableAutoReview: false
|
|
93
101
|
searchMode: live
|
|
94
102
|
|
|
95
103
|
- id: web
|
package/README.i18n.yaml
CHANGED
|
@@ -2,5 +2,5 @@
|
|
|
2
2
|
# side as of the last confirmed-consistent state. Both languages carry equal authority;
|
|
3
3
|
# after editing either side, bring the other along and re-record with:
|
|
4
4
|
# git hash-object README.md docs/README.zh.md
|
|
5
|
-
README.md:
|
|
6
|
-
docs/README.zh.md:
|
|
5
|
+
README.md: 65a8d94c57ae1bd869f5247f158cb1f8d221f4c0
|
|
6
|
+
docs/README.zh.md: 1bdd0641d14d6c27da312ce9ac954fbb2664beae
|
package/README.md
CHANGED
|
@@ -10,7 +10,7 @@ Connect your ChatGPT subscription to DeepSeek Harness with OAuth, optional GPT I
|
|
|
10
10
|
<img src="https://raw.githubusercontent.com/franksong2702/dsh-codex-connect/main/docs/assets/en/hero.jpg" alt="Codex Connect — ChatGPT OAuth for DeepSeek Harness" width="100%">
|
|
11
11
|
</p>
|
|
12
12
|
|
|
13
|
-
`dsh-codex-connect` adds the `openai-codex` model catalog and a separate ChatGPT OAuth login. Models run through Harness's normal LLM service, so streaming, tool calls, reasoning replay, compaction, filesystem controls, permission gates, and approval prompts remain Harness-owned. It does not turn a ChatGPT subscription into an OpenAI Platform API credential. When an eligible GPT Codex model is selected, the Composer also shows a conversation-scoped Fast Mode toggle and
|
|
13
|
+
`dsh-codex-connect` adds the `openai-codex` model catalog and a separate ChatGPT OAuth login. Models run through Harness's normal LLM service, so streaming, tool calls, reasoning replay, compaction, filesystem controls, permission gates, and approval prompts remain Harness-owned. It does not turn a ChatGPT subscription into an OpenAI Platform API credential. When an eligible GPT Codex model is selected, the Composer also shows a conversation-scoped Fast Mode toggle and compact server-reported quota bars.
|
|
14
14
|
|
|
15
15
|
Installation is additive. The bundle does not replace the current default model or search route. Standalone search, `view_image`, and image generation remain disabled until explicitly enabled.
|
|
16
16
|
|
|
@@ -18,29 +18,30 @@ The setup and image-result screenshots in this English guide are captured from t
|
|
|
18
18
|
|
|
19
19
|
## Quick start (about five minutes)
|
|
20
20
|
|
|
21
|
-
This quick start targets DSH `0.1.2-alpha.
|
|
21
|
+
This quick start targets DSH `0.1.2-alpha.5` with Codex Connect Alpha 4.25. Check `dsh --version` first. For DSH `0.1.2-alpha.2`, `0.1.1-rc.2`, or `0.1.0-rc.7`, select the matching plugin version in [INSTALL.md](INSTALL.md). This guide uses the `web` profile; replace `web` with the name of the Harness profile you already use. From a DeepSeek Harness source checkout, prefix the commands with `pnpm`.
|
|
22
22
|
|
|
23
23
|
### 1. Install the plugin into one profile
|
|
24
24
|
|
|
25
25
|
```sh
|
|
26
|
-
dsh plugin --profile web add dsh-codex-connect@0.1.0-alpha.4.
|
|
26
|
+
dsh plugin --profile web add dsh-codex-connect@0.1.0-alpha.4.25
|
|
27
27
|
```
|
|
28
28
|
|
|
29
29
|
Expected result: the package is added to that profile. This does not change the profile's default model or global search route.
|
|
30
30
|
|
|
31
31
|
Use the exact version above to keep the verified DSH and plugin pair reproducible. `alpha` is a moving npm tag, not a compatibility guarantee.
|
|
32
32
|
|
|
33
|
-
### What's new in Alpha 4.
|
|
33
|
+
### What's new in Alpha 4.25
|
|
34
34
|
|
|
35
|
-
-
|
|
36
|
-
-
|
|
37
|
-
-
|
|
35
|
+
- Show the exact 5-hour and weekly quota windows returned by ChatGPT for the selected Codex model. Missing windows stay hidden, Spark remains separate, and plan names do not suppress server data.
|
|
36
|
+
- Separate verified compatibility from maintainer follow-up. The settings card no longer treats missing verification as a known runtime failure or recommends a plugin downgrade.
|
|
37
|
+
- Link users to the canonical tracker for their DSH version when one exists, with a prefilled compatibility-gap report as the safe fallback.
|
|
38
|
+
- Track every newer DSH candidate through one Canary issue with explicit preliminary, compatibility-failure, or infrastructure-blocked state. Canary success does not declare support or publish anything.
|
|
38
39
|
|
|
39
40
|
### Version updates
|
|
40
41
|
|
|
41
|
-
Codex Connect checks public package metadata and this repository's `verified-compatibility.json` periodically through the DSH Web server. The same card reads the locally loaded DSH package version, shows it beside the latest DSH version recorded by this project, and evaluates the exact installed plugin and DSH version pair. Local version detection uses package metadata already available to the plugin and does not require a DSH Core change.
|
|
42
|
+
Codex Connect checks public package metadata and this repository's `verified-compatibility.json` periodically through the DSH Web server. The same card reads the locally loaded DSH package version, shows it beside the latest DSH version recorded by this project, and evaluates the exact installed plugin and DSH version pair. Local version detection uses package metadata already available to the plugin and does not require a DSH Core change. The card's optional tracker lookup sends only that public DSH version to GitHub's public search API.
|
|
42
43
|
|
|
43
|
-
The compatibility record lists exact plugin and DSH versions rather than assuming every later release remains compatible. Maintainers can add a newly verified DSH version to the repository file without publishing another plugin release.
|
|
44
|
+
The compatibility record lists exact plugin and DSH versions rather than assuming every later release remains compatible. Maintainers can add a newly verified DSH version to the repository file without publishing another plugin release. The card keeps two decisions separate: its status describes only whether the installed pair or a specific upgrade path has been verified, while its GitHub action appears only when the installed latest-or-newer DSH has no verified published plugin. Missing verification never claims that a pair is known to fail. A green result means the installed pair was verified. Yellow recommends updating the plugin when the latest plugin matches the installed DSH, or updating DSH when the latest plugin matches the latest verified DSH. Red means Codex Connect has not caught up with the installed latest-or-newer DSH; it links to the canonical compatibility tracker when one exists, or offers a prefilled compatibility-gap report when the tracker lookup is unavailable or has no match. Gray means an older DSH version is not recorded or the public record could not be checked.
|
|
44
45
|
|
|
45
46
|
When a newer plugin version is available, a frame-wide DSH notice appears even if you switch conversations. It first shows the user-facing changes between your installed version and the newest version; technical release notes remain available as a secondary detail or from the release page. To update, copy the short request shown in the notice to the Agent you use for this DSH project. The Agent can inspect the project instructions and choose the appropriate install or update method; the plugin never runs an upgrade command itself.
|
|
46
47
|
|
|
@@ -64,7 +65,7 @@ Expected result: the Harness web UI opens for the selected profile.
|
|
|
64
65
|
|
|
65
66
|
### 3. Find the Openai-Codex account card
|
|
66
67
|
|
|
67
|
-
Open **Settings → Models** and find **Openai-Codex**. This is the primary Alpha 4.
|
|
68
|
+
Open **Settings → Models** and find **Openai-Codex**. This is the primary Alpha 4.25 account entry. If the profile does not expose the Models settings section, open **Settings → Plugins → Plugin configuration → Codex Connect** instead.
|
|
68
69
|
|
|
69
70
|
The Models card carries the attribution “Powered by the Codex Connect plugin.” and provides ChatGPT authorization, reauthorization, sign-out and quota. Both settings pages share one in-memory account state and polling owner. **More settings** opens the proxy, model visibility, search, image and context-budget configuration form in a dialog; the original Plugin settings entry remains available. Both entries save to the same settings scope. Close or Escape discards unsaved dialog edits. The Models footer is optional: profiles without that settings section retain the existing Plugin entry.
|
|
70
71
|
|
|
@@ -135,11 +136,11 @@ Expected result: `status --json` reports `signed-in` and exits `0`, while `docto
|
|
|
135
136
|
The two small controls are shown only when the current conversation is using a GPT model from the `openai-codex` provider. They are session controls, not profile-wide settings:
|
|
136
137
|
|
|
137
138
|
- **Fast Mode (lightning icon)** is off by default for each conversation. Click it to request the faster `1.5×` mode; click it again to return to Standard speed. The control is bound to that conversation and does not change the selected model or other conversations. Hover or focus the icon to see the current state and its quota-consumption warning.
|
|
138
|
-
- **
|
|
139
|
-
- For the exact `gpt-5.3-codex-spark` model, the Composer reads the Spark
|
|
139
|
+
- **Quota bars** are the compact `5h` and `7d` rows beside the model selector. Each row appears only when the server returns that window for the current model bucket. Their colors move from green through yellow/orange to red as the remaining amount falls. Hover or focus the control to see each exact remaining percentage and server-provided reset time. It is hidden for non-GPT models or when no recognized usage window is available.
|
|
140
|
+
- For the exact `gpt-5.3-codex-spark` model, the Composer reads the separate Spark bucket. Other GPT Codex models read the standard Codex bucket. The plugin does not infer quota windows from the ChatGPT plan name.
|
|
140
141
|
|
|
141
142
|
<p align="center">
|
|
142
|
-
<img src="https://raw.githubusercontent.com/franksong2702/dsh-codex-connect/main/docs/assets/composer-capabilities.jpg" alt="DeepSeek Harness Composer with the per-conversation Fast Mode lightning control and
|
|
143
|
+
<img src="https://raw.githubusercontent.com/franksong2702/dsh-codex-connect/main/docs/assets/composer-capabilities.jpg" alt="DeepSeek Harness Composer with the per-conversation Fast Mode lightning control and quota bars" width="820">
|
|
143
144
|
</p>
|
|
144
145
|
|
|
145
146
|
## Optional capabilities (off by default)
|
|
@@ -153,9 +154,10 @@ The installed bundle is intentionally inert beyond model-provider registration:
|
|
|
153
154
|
enableSearch: false
|
|
154
155
|
enableImageTool: false
|
|
155
156
|
enableImageGeneration: false
|
|
157
|
+
enableAutoReview: false
|
|
156
158
|
```
|
|
157
159
|
|
|
158
|
-
Open **Settings → Plugins → Plugin configuration → Codex Connect** to manage the
|
|
160
|
+
Open **Settings → Plugins → Plugin configuration → Codex Connect** to manage the same settings through four modules: **Account & quota**, **Models**, **Network**, and **Capabilities**. Switching modules preserves staged edits; the persistent Save/Discard actions apply them together. **Save changes** affects only this plugin and never selects a default model or a global search route. The Models page's **More settings** dialog uses the same organization without repeating the account module.
|
|
159
161
|
|
|
160
162
|
### Network connection and proxy detection
|
|
161
163
|
|
|
@@ -163,7 +165,7 @@ Codex Connect uses a **Direct connection** by default. A proxy is optional and a
|
|
|
163
165
|
|
|
164
166
|
Select **Detect proxy** to test only the standard proxy environment variables and the documented loopback candidates `127.0.0.1:7890`, `127.0.0.1:7897`, and `127.0.0.1:10809`. Detection makes no model call, consumes no quota, and does not write settings. A response from the canonical Codex endpoint proves network reachability; `401/403`, proxy `407`, DNS, refused connection, timeout, TLS, and CONNECT failures remain separate diagnostics.
|
|
165
167
|
|
|
166
|
-
Choose **Use this proxy** only after reviewing a candidate, then click **Save changes**. **Configure manually**
|
|
168
|
+
Choose **Use this proxy** only after reviewing a candidate, then click **Save changes**. **Configure manually** requires the exact current credential-free HTTP(S) proxy origin to pass its test before the activation action becomes available; editing the draft invalidates the earlier result. **Disable proxy** is always available. A failed probe leaves the previous mode unchanged, and an enabled proxy failure is shown as an actionable error; Codex Connect never silently retries the request through a direct connection.
|
|
167
169
|
|
|
168
170
|
### Enable only the capability you intend to use
|
|
169
171
|
|
|
@@ -198,10 +200,10 @@ The detailed image prompt is authored by the selected GPT model. Codex Connect d
|
|
|
198
200
|
|
|
199
201
|
After sign-in, the Codex Connect settings card can show several server-reported windows. They are separate buckets, not three views of one number:
|
|
200
202
|
|
|
201
|
-
-
|
|
202
|
-
-
|
|
203
|
+
- The standard **Codex** bucket can contain a **5-hour** window, a **Weekly** window, or both.
|
|
204
|
+
- The exact Spark model uses the separate **GPT-5.3-Codex-Spark** bucket and displays whichever windows that bucket returns.
|
|
203
205
|
|
|
204
|
-
Each bar shows the remaining percentage and its local reset time. OpenAI controls the returned windows, eligibility, and reset values;
|
|
206
|
+
Each bar shows the remaining percentage and its local reset time. OpenAI controls the returned windows, eligibility, and reset values; Codex Connect does not remove a returned window based on the plan name or invent a missing one.
|
|
205
207
|
|
|
206
208
|
### Change a default model or global search route separately
|
|
207
209
|
|
|
@@ -304,9 +306,11 @@ The report labels each check `supported`, `rejected`, or `unknown`, with a reaso
|
|
|
304
306
|
|
|
305
307
|
This report covers only the standalone route, not active profile routing, search/image tools, browser compatibility, provider retry behavior, or session recovery. Automatic provider failover is `rejected` because this plugin does not implement it; select an alternative provider explicitly. WebSocket-to-SSE fallback is inactive with the finite SSE default. `contextManagement` and continuation remain `unknown`; native compaction and WebSocket reuse are `rejected` by the current integration policy. No diagnostic result enables these capabilities or changes Harness history. Exit codes cover runtime, OAuth, selected model, Responses, and SSE only: `0` means all five were supported, `1` means at least one was rejected, and `2` means unknown evidence, invalid options, or an inspection failure. Rejected optional capabilities do not change that exit code.
|
|
306
308
|
|
|
307
|
-
###
|
|
309
|
+
### Codex Auto-review and capability probe
|
|
308
310
|
|
|
309
|
-
|
|
311
|
+
Codex Auto-review is an official Codex capability integrated by Codex Connect and disabled by default under **Settings → Plugins → Codex Connect**. Its short explanation stays visible; the full disclosure is expandable, and the first enablement in each profile requires confirmation. When enabled, it reviews eligible Harness approval requests after DSH policy runs. Enabling it permits the plugin to send bounded recent approval context, tool arguments, working directory, and the planned action to `chatgpt.com`; hidden reasoning and stored credentials are excluded. Only a complete structured allow result can grant one execution. Denials inject a rationale and no-circumvention instruction, repeated denials open a turn-local circuit breaker, and `/approve <denial-id>` can authorize one exact retry when the optional command service is present. See [Auto-review](docs/auto-review.md) and [自动审查](docs/auto-review.zh.md).
|
|
312
|
+
|
|
313
|
+
The separate probe remains available for diagnosing the hidden route. It does not add `codex-auto-review` to the model selector and does not review or execute a real Harness command.
|
|
310
314
|
|
|
311
315
|
```sh
|
|
312
316
|
dsh plugin --profile web exec dsh-codex-connect auto-review-probe --json
|
|
@@ -314,11 +318,11 @@ dsh plugin --profile web exec dsh-codex-connect auto-review-probe --json
|
|
|
314
318
|
|
|
315
319
|
The command sends one fixed, synthetic no-op to the hidden reviewer through the ChatGPT OAuth Responses route. It requires an unexpired stored credential, never refreshes or writes credentials, does not follow redirects or retry, caps the response at 64 KiB, and destroys its owned connection before returning. `--proxy <http(s)-origin>` and `--timeout-ms <1..60000>` have the same explicit-network meaning as the ordinary capability probe.
|
|
316
320
|
|
|
317
|
-
`supported` means only that the route accepted the exact hidden model id and returned one complete assessment matching the reviewer JSON fields. `rejected` means the request or local prerequisite was explicitly rejected. Timeouts, cancellation, malformed output, incomplete streams, rate limits, and network failures remain `unknown`. Output omits credentials, account ids, response ids, provider messages, model text, paths, headers, and proxy origins. Exit `0` requires runtime, OAuth, and reviewer checks to be supported; exit `1` means at least one was rejected; exit `2` means evidence is unknown or input is invalid. The report is evidence only: it never
|
|
321
|
+
`supported` means only that the route accepted the exact hidden model id and returned one complete assessment matching the reviewer JSON fields. `rejected` means the request or local prerequisite was explicitly rejected. Timeouts, cancellation, malformed output, incomplete streams, rate limits, and network failures remain `unknown`. Output omits credentials, account ids, response ids, provider messages, model text, paths, headers, and proxy origins. Exit `0` requires runtime, OAuth, and reviewer checks to be supported; exit `1` means at least one was rejected; exit `2` means evidence is unknown or input is invalid. The report is evidence only: it never changes the Auto-review setting, DSH policy, or authorization state.
|
|
318
322
|
|
|
319
323
|
## Compatibility and security boundary
|
|
320
324
|
|
|
321
|
-
- Alpha 4.
|
|
325
|
+
- Alpha 4.25 is verified with DSH plugin API packages `0.1.2-alpha.5`, `@earendil-works/pi-ai` `^0.84.2` (resolved as `0.84.4` during verification), and Node.js `^22.19.0 || >=24.0.0`. Alpha 4.24 remains an earlier verified choice for the same DSH version, Alpha 4.23 remains the verified choice for DSH `0.1.2-alpha.2`, and Alpha 4.21 remains verified with DSH `0.1.1-rc.2` and pi-ai `0.82.1`. [verified-compatibility.json](verified-compatibility.json) records the exact pairs; see [INSTALL.md](INSTALL.md) for installation commands.
|
|
322
326
|
- The new DSH client splits its former runtime into Session Controller, Settings, Store, and Renderer packages. Codex Connect uses those public interfaces for settings and image actions. DSH owns normalized preview encoding and dimensions; Codex Connect retains the exact original image separately.
|
|
323
327
|
- Upgrade the DSH plugin API packages and `@earendil-works/pi-ai` as one group, then run `dsh-codex-connect doctor --json` and the compatibility check again. This contract does not make claims about future versions.
|
|
324
328
|
- When the daily upstream check finds a new `latest` or `next` DSH candidate, it installs Codex Connect into an isolated profile, boots the installed model runtime without OAuth credentials, verifies model and reasoning-effort discovery, and confirms provider disposal. Live sign-in, quota, and model requests still require manual validation in the test profile.
|
package/compatibility.json
CHANGED
package/cordis.patch.yml
CHANGED
package/docs/README.zh.md
CHANGED
|
@@ -10,7 +10,7 @@
|
|
|
10
10
|
<img src="https://raw.githubusercontent.com/franksong2702/dsh-codex-connect/main/docs/assets/zh/hero.jpg" alt="Codex Connect — 通过 ChatGPT OAuth 连接 DeepSeek Harness" width="100%">
|
|
11
11
|
</p>
|
|
12
12
|
|
|
13
|
-
`dsh-codex-connect` 提供 `openai-codex` 模型目录和独立的 ChatGPT OAuth 登录。模型仍走 Harness 标准 LLM 服务，因此流式输出、工具调用、reasoning replay、压缩、文件系统控制、权限门禁和审批提示仍由 Harness 负责。ChatGPT 订阅不会因此变成 OpenAI Platform API 凭据。选择符合条件的 GPT Codex 模型后，Composer 还会显示按对话绑定的 Fast Mode
|
|
13
|
+
`dsh-codex-connect` 提供 `openai-codex` 模型目录和独立的 ChatGPT OAuth 登录。模型仍走 Harness 标准 LLM 服务，因此流式输出、工具调用、reasoning replay、压缩、文件系统控制、权限门禁和审批提示仍由 Harness 负责。ChatGPT 订阅不会因此变成 OpenAI Platform API 凭据。选择符合条件的 GPT Codex 模型后，Composer 还会显示按对话绑定的 Fast Mode 开关和服务端返回的紧凑额度条。
|
|
14
14
|
|
|
15
15
|
安装是增量的：bundle 不会替换当前主模型或搜索路由；独立搜索、`view_image` 和图片生成也默认关闭，必须显式开启。
|
|
16
16
|
|
|
@@ -18,29 +18,30 @@
|
|
|
18
18
|
|
|
19
19
|
## 五分钟快速开始
|
|
20
20
|
|
|
21
|
-
本快速指南适用于 DSH `0.1.2-alpha.
|
|
21
|
+
本快速指南适用于 DSH `0.1.2-alpha.5` 与 Codex Connect Alpha 4.25。请先运行 `dsh --version`。如果使用 DSH `0.1.2-alpha.2`、`0.1.1-rc.2` 或 `0.1.0-rc.7`，请在 [INSTALL.md](../INSTALL.md) 中选择匹配的插件版本。本指南使用 `web` profile；请把 `web` 替换成你已经在用的 Harness profile 名称。如果在 DeepSeek Harness 源码 checkout 中运行，请在命令前加 `pnpm`。
|
|
22
22
|
|
|
23
23
|
### 1. 将插件装入一个 profile
|
|
24
24
|
|
|
25
25
|
```sh
|
|
26
|
-
dsh plugin --profile web add dsh-codex-connect@0.1.0-alpha.4.
|
|
26
|
+
dsh plugin --profile web add dsh-codex-connect@0.1.0-alpha.4.25
|
|
27
27
|
```
|
|
28
28
|
|
|
29
29
|
预期结果：包被加入该 profile。这个动作不会更改 profile 的默认模型或全局搜索路由。
|
|
30
30
|
|
|
31
31
|
请使用上面的精确版本，确保已验证的 DSH 与插件组合可以复现。`alpha` 是会移动的 npm 标签，不是兼容性保证。
|
|
32
32
|
|
|
33
|
-
### Alpha 4.
|
|
33
|
+
### Alpha 4.25 更新内容
|
|
34
34
|
|
|
35
|
-
-
|
|
36
|
-
-
|
|
37
|
-
-
|
|
35
|
+
- 显示 ChatGPT 为当前 Codex 模型实际返回的 5 小时和每周额度窗口。缺少的窗口不会凭空出现，Spark 额度保持独立，也不会因套餐名称而隐藏服务端数据。
|
|
36
|
+
- 将“当前组合是否已验证”和“是否需要提醒维护者”分开判断。缺少验证记录不再被说成已知运行失败，也不会建议降级插件。
|
|
37
|
+
- 如果已经有当前 DSH 版本的统一跟进 Issue，设置卡片会直接链接过去；否则提供预填的兼容性缺口反馈。
|
|
38
|
+
- 每个更高版本的 DSH 候选只维护一个 Canary 跟进 Issue，明确区分等待完整验证、兼容性失败和基础设施阻塞。Canary 成功不会自动声明兼容或发布版本。
|
|
38
39
|
|
|
39
40
|
### 版本更新提醒
|
|
40
41
|
|
|
41
|
-
Codex Connect 会通过 DSH Web 服务定期检查公开的 npm 包信息，以及本仓库维护的 `verified-compatibility.json`。同一张卡片会读取本机实际加载的 DSH 软件包版本，与本项目记录的最新 DSH 版本并列显示，并对当前安装的插件版本和 DSH 版本这一组精确组合进行判断。本机版本检测只使用插件已经能够取得的软件包元数据，不要求 DSH Core
|
|
42
|
+
Codex Connect 会通过 DSH Web 服务定期检查公开的 npm 包信息，以及本仓库维护的 `verified-compatibility.json`。同一张卡片会读取本机实际加载的 DSH 软件包版本，与本项目记录的最新 DSH 版本并列显示，并对当前安装的插件版本和 DSH 版本这一组精确组合进行判断。本机版本检测只使用插件已经能够取得的软件包元数据，不要求 DSH Core 作出任何修改。卡片仅在需要兼容性跟进时查询 Issue，并且只向 GitHub 公开搜索 API 发送这个公开的 DSH 版本号。卡片将两项判断分开：状态只说明当前精确组合或明确升级路径是否已验证；只有当前安装的 DSH 等于或新于记录中的最新版、且尚无任何已验证的已发布插件时，才显示 GitHub 跟进操作。缺少验证记录不代表该组合已知不可运行。
|
|
42
43
|
|
|
43
|
-
兼容性清单按插件版本和 DSH 版本精确记录，不假设后续版本天然兼容。维护者确认新的 DSH
|
|
44
|
+
兼容性清单按插件版本和 DSH 版本精确记录，不假设后续版本天然兼容。维护者确认新的 DSH 版本后，只需更新仓库文件，不必重新发布插件。绿色表示当前安装组合已经通过验证。黄色状态下，如果最新插件适配当前 DSH，就建议升级插件；如果最新插件适配最新 DSH，就建议升级 DSH。红色表示当前 DSH 已达到或超过清单记录的最新版，但最新发布的插件仍未完成验证；如果已有该版本的统一兼容性跟进 Issue，卡片会直接链接过去，如果查询失败或没有匹配项，则提供预填的兼容性缺口反馈。灰色表示较早的 DSH 版本未被记录，或暂时无法检查公开清单。
|
|
44
45
|
|
|
45
46
|
发现插件新版本后，即使你切换了对话，DSH 界面也会显示全局更新提醒。提醒会先展示从你当前版本到最新版本之间对用户有用的功能；完整技术发布说明放在次级详情或发布页面中。插件不会自动执行升级命令。
|
|
46
47
|
|
|
@@ -66,7 +67,7 @@ dsh web
|
|
|
66
67
|
|
|
67
68
|
### 3. 找到 Openai-Codex 账户卡
|
|
68
69
|
|
|
69
|
-
打开 **设置 → 模型**，找到 **Openai-Codex**。这是 Alpha 4.
|
|
70
|
+
打开 **设置 → 模型**，找到 **Openai-Codex**。这是 Alpha 4.25 的主要账户入口。如果当前 profile 没有模型设置分区，请改用 **设置 → 插件 → 插件配置 → Codex Connect**。
|
|
70
71
|
|
|
71
72
|
模型页账户卡标注“由 Codex Connect 插件提供支持。”，用于 ChatGPT 授权、重新授权、退出和查看额度。两个设置页面共用同一份内存账户状态及轮询。**更多设置** 会在弹窗中打开代理、模型显示、搜索、图片和上下文预算配置表单，原插件设置入口仍保留。两处保存到同一份配置；关闭弹窗或按 Escape 会放弃弹窗内未保存的修改。模型页底部入口是可选增强：没有该设置分区的 profile 仍保留原插件入口。
|
|
72
73
|
|
|
@@ -137,11 +138,11 @@ dsh plugin --profile web exec dsh-codex-connect doctor --json
|
|
|
137
138
|
只有当前对话选择了 `openai-codex` 提供方的 GPT 模型时，Composer 才会显示下面两个小控件。它们都是当前对话级别的控制，不是 profile 全局设置：
|
|
138
139
|
|
|
139
140
|
- **Fast Mode（闪电图标）**：每个对话默认关闭。点击后请求更快的 `1.5 倍` 模式，再点一次恢复标准速度。它只绑定当前对话，不会改变模型选择，也不会影响其他对话。鼠标悬停或键盘聚焦闪电图标，可以看到当前状态和额度消耗提示。
|
|
140
|
-
-
|
|
141
|
-
- 对于精确模型 `gpt-5.3-codex-spark`，Composer
|
|
141
|
+
- **额度进度条**：位于模型选择器旁边，以紧凑的 `5h` 和 `7d` 行显示。只有服务端为当前模型额度桶返回对应窗口时，该行才会出现。剩余额度越低，颜色会从绿色经过黄色/橙色变为红色。鼠标悬停或键盘聚焦时，会显示每个窗口的精确剩余百分比和服务端提供的重置时间。非 GPT 模型或没有可识别额度窗口时不会显示。
|
|
142
|
+
- 对于精确模型 `gpt-5.3-codex-spark`，Composer 读取独立的 Spark 额度桶；其他 GPT Codex 模型读取标准 Codex 额度桶。插件不会根据 ChatGPT 套餐名称猜测额度窗口。
|
|
142
143
|
|
|
143
144
|
<p align="center">
|
|
144
|
-
<img src="https://raw.githubusercontent.com/franksong2702/dsh-codex-connect/main/docs/assets/composer-capabilities.jpg" alt="DeepSeek Harness Composer 中按对话绑定的 Fast Mode
|
|
145
|
+
<img src="https://raw.githubusercontent.com/franksong2702/dsh-codex-connect/main/docs/assets/composer-capabilities.jpg" alt="DeepSeek Harness Composer 中按对话绑定的 Fast Mode 闪电控件和额度进度条" width="820">
|
|
145
146
|
</p>
|
|
146
147
|
|
|
147
148
|
## 可选能力（默认关闭）
|
|
@@ -155,9 +156,10 @@ dsh plugin --profile web exec dsh-codex-connect doctor --json
|
|
|
155
156
|
enableSearch: false
|
|
156
157
|
enableImageTool: false
|
|
157
158
|
enableImageGeneration: false
|
|
159
|
+
enableAutoReview: false
|
|
158
160
|
```
|
|
159
161
|
|
|
160
|
-
打开 **设置 → 插件 → 插件配置 → Codex Connect
|
|
162
|
+
打开 **设置 → 插件 → 插件配置 → Codex Connect**，即可通过**账户与额度**、**模型**、**网络**和**能力**四个模块管理同一组设置。切换模块不会丢失暂存修改，常驻的保存/放弃操作会统一处理这些修改。**保存更改**只影响本插件，绝不会选择默认模型或全局搜索路由。模型页的**更多设置**弹窗采用相同组织方式，但不重复账户模块。
|
|
161
163
|
|
|
162
164
|
### 网络连接与代理检测
|
|
163
165
|
|
|
@@ -165,7 +167,7 @@ Codex Connect 默认使用**直连**。代理是可选项，只作用于本插
|
|
|
165
167
|
|
|
166
168
|
点击 **检测代理** 时，只会测试标准代理环境变量，以及文档列出的本机候选地址：`127.0.0.1:7890`、`127.0.0.1:7897` 和 `127.0.0.1:10809`。检测不调用模型、不消耗额度，也不会写入设置。规范 Codex 端点返回任何 HTTP 响应都表示网络可达；`401/403`、代理 `407`、DNS、连接被拒绝、超时、TLS 和 CONNECT 失败会分别显示为诊断类别。
|
|
167
169
|
|
|
168
|
-
请先检查候选地址，再选择 **使用此代理**，最后点击
|
|
170
|
+
请先检查候选地址，再选择 **使用此代理**，最后点击 **保存更改**。**手动配置**要求当前这个不带凭据的 HTTP(S) proxy origin 测试成功后才允许启用；修改草稿会使之前的测试结果失效。**停用代理**始终可用。检测失败会保留原来的模式；代理已启用但请求失败时，界面会给出可操作的错误，绝不会静默改走直连。
|
|
169
171
|
|
|
170
172
|
### 只开启你准备使用的能力
|
|
171
173
|
|
|
@@ -200,10 +202,10 @@ Codex Connect 默认使用**直连**。代理是可选项，只作用于本插
|
|
|
200
202
|
|
|
201
203
|
登录后，Codex Connect 设置卡片可能显示多个服务端额度窗口。它们是不同的额度桶，不是同一个数字重复显示：
|
|
202
204
|
|
|
203
|
-
- **Codex
|
|
204
|
-
-
|
|
205
|
+
- 标准 **Codex** 额度桶可能包含 **5 小时额度**、**每周额度**，或者同时包含两者。
|
|
206
|
+
- 精确 Spark 模型使用独立的 **GPT-5.3-Codex-Spark** 额度桶，并显示该额度桶实际返回的窗口。
|
|
205
207
|
|
|
206
|
-
每条进度条都会显示剩余百分比和按本地时区格式化的重置时间。额度窗口、模型资格和重置时间由 OpenAI
|
|
208
|
+
每条进度条都会显示剩余百分比和按本地时区格式化的重置时间。额度窗口、模型资格和重置时间由 OpenAI 返回；Codex Connect 不会根据套餐名称删除已返回的窗口，也不会虚构缺失窗口。
|
|
207
209
|
|
|
208
210
|
### 单独更改默认模型或全局搜索路由
|
|
209
211
|
|
|
@@ -306,9 +308,11 @@ dsh plugin --profile web exec dsh-codex-connect capabilities --model gpt-5.6-sol
|
|
|
306
308
|
|
|
307
309
|
本报告仅涵盖独立路由，不验证活动 profile 路由、搜索/图片工具、浏览器兼容性、provider 重试行为或会话恢复。本插件没有实现自动 provider 故障切换，因此该项为 `rejected`，需要用户明确选择其他 provider。有限 SSE 默认路径不会触发 WebSocket 到 SSE 的回退。`contextManagement` 和续接仍为 `unknown`；原生 compaction 和 WebSocket reuse 在当前集成策略下为 `rejected`。诊断结果不会启用这些能力，也不会更改 Harness 历史。退出码只覆盖运行时、OAuth、所选模型、Responses 和 SSE：`0` 表示五项均可用，`1` 表示至少一项被拒绝，`2` 表示证据未知、选项无效或检查失败。被拒绝的可选能力不影响该退出码。
|
|
308
310
|
|
|
309
|
-
###
|
|
311
|
+
### Codex 自动审查与能力探针
|
|
310
312
|
|
|
311
|
-
|
|
313
|
+
Codex 自动审查是 Codex Connect 接入的 Codex 官方能力，在 **设置 → 插件 → Codex Connect** 中默认关闭。界面常驻简短说明，完整告知可展开，每个 profile 首次启用时需要确认。启用后，它会在 DSH 策略检查之后审查符合条件的 Harness 审批请求。启用即允许插件把有界的最近审批上下文、工具参数、工作目录和待执行动作发送到 `chatgpt.com`；隐藏推理和已保存凭据会被排除。只有完整的结构化允许结果才能授权一次执行。拒绝会注入理由和禁止绕行指引；连续拒绝会打开当前轮熔断器；存在可选命令服务时，`/approve <拒绝记录 ID>` 可以授权一次完全相同的重试。详见[自动审查](auto-review.zh.md)和 [Auto-review](auto-review.md)。
|
|
314
|
+
|
|
315
|
+
独立探针继续用于诊断隐藏路由。它不会把 `codex-auto-review` 加入模型选择器，也不会审查或执行真实的 Harness 命令。
|
|
312
316
|
|
|
313
317
|
```sh
|
|
314
318
|
dsh plugin --profile web exec dsh-codex-connect auto-review-probe --json
|
|
@@ -316,11 +320,11 @@ dsh plugin --profile web exec dsh-codex-connect auto-review-probe --json
|
|
|
316
320
|
|
|
317
321
|
该命令通过 ChatGPT OAuth Responses 路由向隐藏 reviewer 发送一条固定的合成空操作。它要求已存凭据尚未过期，不刷新或写入凭据，不跟随重定向、不重试，响应上限为 64 KiB，并在返回前销毁自有连接。`--proxy <http(s)-origin>` 和 `--timeout-ms <1..60000>` 与普通能力探针具有相同的显式网络含义。
|
|
318
322
|
|
|
319
|
-
`supported` 只表示路由接受了精确的隐藏模型 ID，并返回一条符合 reviewer JSON 字段的完整审查结果。`rejected` 表示请求或本地前置条件被明确拒绝。超时、取消、格式错误、不完整流、限流和网络故障均保持 `unknown`。输出会省略凭据、账号 ID、response ID、服务端消息、模型文本、路径、headers 和代理 origin。只有运行时、OAuth 和 reviewer 三项均为可用时才返回 `0`；至少一项被拒绝时返回 `1`；证据未知或输入无效时返回 `2
|
|
323
|
+
`supported` 只表示路由接受了精确的隐藏模型 ID，并返回一条符合 reviewer JSON 字段的完整审查结果。`rejected` 表示请求或本地前置条件被明确拒绝。超时、取消、格式错误、不完整流、限流和网络故障均保持 `unknown`。输出会省略凭据、账号 ID、response ID、服务端消息、模型文本、路径、headers 和代理 origin。只有运行时、OAuth 和 reviewer 三项均为可用时才返回 `0`；至少一项被拒绝时返回 `1`；证据未知或输入无效时返回 `2`。该报告只提供证据：它绝不会修改自动审查设置、DSH 策略或授权状态。
|
|
320
324
|
|
|
321
325
|
## 兼容性与安全边界
|
|
322
326
|
|
|
323
|
-
- Alpha 4.
|
|
327
|
+
- Alpha 4.25 已与 DSH 插件 API packages `0.1.2-alpha.5`、`@earendil-works/pi-ai` `^0.84.2`（验证时解析为 `0.84.4`）和 Node.js `^22.19.0 || >=24.0.0` 完成验证。Alpha 4.24 仍是同一 DSH 版本较早的已验证选择；Alpha 4.23 仍是 DSH `0.1.2-alpha.2` 的已验证选择；Alpha 4.21 仍与 DSH `0.1.1-rc.2` 和 pi-ai `0.82.1` 保持已验证状态。[verified-compatibility.json](../verified-compatibility.json) 记录精确组合；安装命令见 [INSTALL.md](../INSTALL.md)。
|
|
324
328
|
- 新版 DSH 将原来的 client runtime 拆分为 Session Controller、Settings、Store 和 Renderer 包。Codex Connect 通过这些公开接口接入设置和图片操作。规范化预览的编码与尺寸由 DSH 决定；Codex Connect 另行保留字节完全一致的原图。
|
|
325
329
|
- 升级时请将 DSH 插件 API packages 与 `@earendil-works/pi-ai` 作为一组升级，再运行 `dsh-codex-connect doctor --json` 和兼容性检查。本契约不对未来版本作判断。
|
|
326
330
|
- 每日上游检查发现新的 DSH `latest` 或 `next` 候选版本时，会把 Codex Connect 安装到隔离 Profile 中，在没有 OAuth 凭据的情况下启动已安装的模型运行时，验证模型与推理强度发现，并确认提供方可被正确卸载。真实登录、额度和模型请求仍需在测试 Profile 中人工验证。
|
|
@@ -0,0 +1,26 @@
|
|
|
1
|
+
# Auto-review
|
|
2
|
+
|
|
3
|
+
Auto-review is an official Codex capability. Codex Connect integrates its reviewer with eligible DeepSeek Harness approval requests. The feature is disabled by default and participates only when the active request provider is `openai-codex`; another provider's session always keeps its existing answerer chain. It does not weaken Harness approval policy, sandboxing, tool restrictions, or permission checks: Harness decides whether approval is needed before Auto-review runs.
|
|
4
|
+
|
|
5
|
+
Enable **Codex Auto-review** under **Settings → Plugins → Codex Connect → Optional capabilities**. The settings card keeps a short explanation visible and places the full disclosure behind **Learn what is sent and how failures are handled**. The first attempt to enable Auto-review in a profile requires confirmation; that acknowledgement is saved in the profile and is not requested again in another browser. Enabling it permits Codex Connect to send the recent approval context, tool arguments, working directory, and planned action to `chatgpt.com`. Hidden reasoning and stored credentials are excluded. Disabling it immediately delegates every request to the existing human answerer chain.
|
|
6
|
+
|
|
7
|
+
## Decision rules
|
|
8
|
+
|
|
9
|
+
- Only a complete structured `allow` result returns `allowed-once`.
|
|
10
|
+
- A structured `deny` rejects the action and adds the rationale plus a no-circumvention instruction to the next model step.
|
|
11
|
+
- Missing or ambiguous action data, missing credentials, unsupported routes, malformed responses, and transport failures return to human approval.
|
|
12
|
+
- Cancellation stays cancelled. A timeout is reported separately; one retry of the exact action is permitted before later timeouts return to human approval.
|
|
13
|
+
- Three consecutive denials, or ten denials in the last fifty reviews in one turn, stop that turn.
|
|
14
|
+
- `/approve <denial-id>` authorizes one retry only when the tool, canonical arguments, and working directory exactly match the selected denial. A mismatch consumes the one-shot authorization without allowing the action. This command requires the optional `@deepseek-ai/dsh-commands` host capability.
|
|
15
|
+
|
|
16
|
+
Approval requests and outcomes remain durable through Harness `approval/asked` and `approval/decided` events. `/approve` uses Harness command lifecycle events. Codex Connect logs only the action fingerprint and structured assessment labels; it does not duplicate raw tool arguments in a new audit event.
|
|
17
|
+
|
|
18
|
+
## Context limits
|
|
19
|
+
|
|
20
|
+
The reviewer input uses conservative UTF-8 byte limits before network transmission: 20,000 bytes for retained narrative, 10,000 bytes for tool context, 5,000 bytes per narrative entry, 1,000 bytes per tool entry, and at most forty recent non-user entries. User text is labeled trusted; assistant and plugin text is not authorization. Omission and truncation counts are included in the request.
|
|
21
|
+
|
|
22
|
+
## Service status
|
|
23
|
+
|
|
24
|
+
OpenAI documents Auto-review as a Codex feature, but does not promise the `codex-auto-review` OAuth route as a stable public API. The separate `auto-review-probe` command checks only whether the current OAuth route accepts one synthetic no-op assessment. Runtime failures always return to human approval; they never authorize execution.
|
|
25
|
+
|
|
26
|
+
See [OpenAI Auto-review](https://learn.chatgpt.com/docs/sandboxing/auto-review), [OpenAI guardrails and approvals](https://developers.openai.com/api/docs/guides/agents/guardrails-approvals), and [Issue #84](https://github.com/franksong2702/dsh-codex-connect/issues/84).
|
|
@@ -0,0 +1,26 @@
|
|
|
1
|
+
# 自动审查
|
|
2
|
+
|
|
3
|
+
Auto-review 是 Codex 官方能力，Codex Connect 把它的审核器接入符合条件的 DeepSeek Harness 审批请求。该功能默认关闭，并且只在当前请求提供方是 `openai-codex` 时参与审批；其他提供方的会话始终保留原有回答器链。它也不会削弱 Harness 的审批策略、沙箱、工具限制或权限检查：是否需要审批始终由 Harness 先决定。
|
|
4
|
+
|
|
5
|
+
在 **设置 → 插件 → Codex Connect → 可选能力** 中启用 **Codex 自动审查**。设置卡片常驻一句简短说明，完整告知放在 **了解发送内容与失败处理** 中。每个 profile 首次启用时必须确认；该确认保存在 profile 中，更换浏览器后不会再次弹出。启用即允许 Codex Connect 把最近的审批上下文、工具参数、工作目录和待执行动作发送到 `chatgpt.com`。隐藏推理和已保存凭据不会进入审核请求。关闭后立即把所有请求交还原有人工审批链。
|
|
6
|
+
|
|
7
|
+
## 决策规则
|
|
8
|
+
|
|
9
|
+
- 只有完整、结构化的 `allow` 才返回 `allowed-once`。
|
|
10
|
+
- 结构化 `deny` 会拒绝动作，并把理由及禁止绕行指引加入下一模型步骤。
|
|
11
|
+
- 动作缺失或歧义、凭据缺失、路由不支持、响应畸形和传输失败都回到人工审批。
|
|
12
|
+
- 取消仍是取消。超时会单独提示；同一精确动作只允许重试审核一次，之后的超时回到人工审批。
|
|
13
|
+
- 同一轮连续拒绝三次，或最近五十次审核中拒绝十次，会停止该轮。
|
|
14
|
+
- `/approve <拒绝记录 ID>` 只授权一次完全相同的重试：工具、规范化参数和工作目录必须与所选拒绝记录一致。不匹配会消费这次授权，但不会放行动作。该命令需要宿主提供可选的 `@deepseek-ai/dsh-commands` 能力。
|
|
15
|
+
|
|
16
|
+
审批问题和结果继续由 Harness 的 `approval/asked`、`approval/decided` 事件持久记录；`/approve` 使用 Harness 命令生命周期事件。Codex Connect 只额外记录动作指纹和结构化评估标签，不在新审计事件中重复原始工具参数。
|
|
17
|
+
|
|
18
|
+
## 上下文限制
|
|
19
|
+
|
|
20
|
+
网络请求前使用保守的 UTF-8 字节限制：会话文本 20,000 字节，工具上下文 10,000 字节，单条会话文本 5,000 字节，单条工具记录 1,000 字节，并且最多保留四十条最近的非用户记录。用户文本标记为可信授权来源；assistant 和插件文本不构成授权。请求中会明确包含省略和截断数量。
|
|
21
|
+
|
|
22
|
+
## 服务状态
|
|
23
|
+
|
|
24
|
+
OpenAI 已把 Auto-review 记录为 Codex 功能，但没有承诺 `codex-auto-review` OAuth 路由是稳定公共 API。独立的 `auto-review-probe` 命令只检查当前 OAuth 路由能否完成一次合成的空操作评估。运行时失败始终回到人工审批，绝不会放行执行。
|
|
25
|
+
|
|
26
|
+
参见 [OpenAI Auto-review](https://learn.chatgpt.com/docs/sandboxing/auto-review)、[OpenAI 审批与护栏](https://developers.openai.com/api/docs/guides/agents/guardrails-approvals) 和 [Issue #84](https://github.com/franksong2702/dsh-codex-connect/issues/84)。
|
package/docs/design.i18n.yaml
CHANGED
|
@@ -1,4 +1,4 @@
|
|
|
1
1
|
# Bilingual-pair consistency record for the design document. Re-record with:
|
|
2
2
|
# git hash-object docs/design.md docs/design.zh.md
|
|
3
|
-
design.md:
|
|
4
|
-
design.zh.md:
|
|
3
|
+
design.md: 86c383527dad511eb3458fb6c6389d5f8e9e9c16
|
|
4
|
+
design.zh.md: f6112594d2c88ebab0f35e5f8f8dd2e949e611b5
|
package/docs/design.md
CHANGED
|
@@ -26,4 +26,4 @@ Before registration the plugin checks current provider ids. An existing `openai-
|
|
|
26
26
|
|
|
27
27
|
## Compatibility boundary
|
|
28
28
|
|
|
29
|
-
Alpha 4.
|
|
29
|
+
Alpha 4.25 pins Harness `0.1.2-alpha.5` development dependencies and follows its pi-ai range `^0.84.2`; the verified release lockfile selects pi-ai `0.84.4`. Supported Node.js remains `^22.19.0 || >=24.0.0`. The keyed `settings.plugin.item` integration remains, while client types come from their Session Controller, Settings, Store, and Renderer owners. The published compatibility record lists the exact Alpha 4.25 and DSH `0.1.2-alpha.5` pair. Backend eligibility, quotas, models, service-side context capacity, and protocol details remain controlled upstream. Tests use temporary OAuth documents and mocked network responses; CI does not perform real authentication.
|
package/docs/design.zh.md
CHANGED
|
@@ -22,4 +22,4 @@ Host 将 `llm-openai-codex` 注册为插件自有的能力 settings namespace。
|
|
|
22
22
|
|
|
23
23
|
注册前检查现有 provider id；发现 `openai-codex` 已被占用时，给出旧 bundle 或手动 provider 配置的定向迁移提示。boot-free CLI doctor 只报告包/运行时版本、OAuth 路径元数据、能力默认值和安全提示。
|
|
24
24
|
|
|
25
|
-
Alpha 4.
|
|
25
|
+
Alpha 4.25 固定使用 Harness `0.1.2-alpha.5` 开发依赖，并跟随其 pi-ai 版本范围 `^0.84.2`；已验证的发布锁文件选择 pi-ai `0.84.4`。Node.js 支持范围仍为 `^22.19.0 || >=24.0.0`。keyed `settings.plugin.item` 集成保持不变，客户端类型继续从 Session Controller、Settings、Store 和 Renderer 的所属包导入。已发布兼容性记录列出 Alpha 4.25 与 DSH `0.1.2-alpha.5` 的精确组合。资格、额度、模型、服务端上下文容量和后端协议仍由上游控制。测试仅使用临时 OAuth 文档和模拟网络响应，CI 不执行真实认证。
|