@akagilnc/pi-workflow-roles 0.1.2475 → 0.1.2479
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/package.json +1 -1
- package/resources/engines/zcode.md +63 -0
package/package.json
CHANGED
|
@@ -0,0 +1,63 @@
|
|
|
1
|
+
# zcode engine method material
|
|
2
|
+
|
|
3
|
+
This file is packaged technical material for the optional `zcode` labor engine
|
|
4
|
+
(Z.AI ZCode agent runtime on the host, GLM Coding Plan models).
|
|
5
|
+
|
|
6
|
+
Before invoking the engine, read `../engine-dispatch.md`, resolving that path
|
|
7
|
+
relative to this note. This note only covers this engine's CLI technical
|
|
8
|
+
parameters.
|
|
9
|
+
|
|
10
|
+
## Invocation (host-verified 2026-08-29)
|
|
11
|
+
|
|
12
|
+
The machine entrypoint is `zcode` (npm `zcode-app-cli`, wraps the official
|
|
13
|
+
ZCode runtime; host has 3.10.1-17 / runtime 0.16.5). Non-interactive labor:
|
|
14
|
+
|
|
15
|
+
```bash
|
|
16
|
+
zcode --prompt 'YOUR_LABOR_PROMPT' --cwd /path/to/worktree
|
|
17
|
+
```
|
|
18
|
+
|
|
19
|
+
`--prompt` defaults to permission mode `yolo` (no TTY permission stalls).
|
|
20
|
+
Useful extras measured from `zcode --help`: `--attach <path>` (repeatable),
|
|
21
|
+
`--mode build|edit|plan|yolo`, `--resume <sess_...>`, `--json` where
|
|
22
|
+
supported. Prefer `zcode --help` on the host over any remembered flag set.
|
|
23
|
+
|
|
24
|
+
## Realm and model traps (host-verified 2026-08-29, all three hit in sequence)
|
|
25
|
+
|
|
26
|
+
- Missing `~/.zcode/cli/config.json` → hard error `Model config is missing`.
|
|
27
|
+
The desktop app's login is NOT shared with the CLI.
|
|
28
|
+
- `zcode login` signs into the **overseas Z.AI realm only**. An account on the
|
|
29
|
+
mainland BigModel realm then fails with
|
|
30
|
+
`[1113][Insufficient balance or no resource package]`.
|
|
31
|
+
- The scaffolded config defaults `model.main` to `zai/glm-5.2`, which is not
|
|
32
|
+
in the coding plan — same 1113 failure even with a valid plan.
|
|
33
|
+
|
|
34
|
+
Working mainland configuration (per official docs
|
|
35
|
+
`docs.bigmodel.cn/cn/coding-plan/quick-start`): in
|
|
36
|
+
`~/.zcode/cli/config.json`, provider kind `anthropic` with
|
|
37
|
+
`baseURL: https://open.bigmodel.cn/api/anthropic`, `options.apiKey` = the
|
|
38
|
+
coding-plan key (host keychain: service `glm-key`, account `akagilnc`), and
|
|
39
|
+
`model.main: zai/glm-5.3`, `model.lite: zai/glm-5.3-flash`. With that config
|
|
40
|
+
the smoke prompt returns normally.
|
|
41
|
+
|
|
42
|
+
## Quota facts (owner-provided, 2026-08-28)
|
|
43
|
+
|
|
44
|
+
Plan quotas are daily and per-model (GLM-5.3 3M/day, Flash 5M/day on the
|
|
45
|
+
current plan; a weekend event granted a larger temporary pool). Whether cached
|
|
46
|
+
tokens count toward quota is unverified.
|
|
47
|
+
|
|
48
|
+
## Lane status: PARKED (host-verified 2026-08-29)
|
|
49
|
+
|
|
50
|
+
The "working" configuration above drew from a new-user gift resource package
|
|
51
|
+
(2M general-model tokens), not from any plan. Once that package expired,
|
|
52
|
+
glm-5.3-flash returns `1113` (insufficient balance) with the same key on BOTH
|
|
53
|
+
the anthropic endpoint and the coding endpoint
|
|
54
|
+
(`open.bigmodel.cn/api/coding/paas/v4`, HTTP 429 + code 1113, probed with a
|
|
55
|
+
minimal request). The ZCode Start/Weekend plan quotas are bound to the desktop
|
|
56
|
+
app's OAuth connection ("Start Plan" connection mode); the CLI's
|
|
57
|
+
`/login bigmodel-coding-plan` OAuth variant hard-fails with
|
|
58
|
+
`BigModel OAuth appSecret is required.` (the client secret ships only inside
|
|
59
|
+
the desktop app; no env override exists — only `BIGMODEL_*_API_BASE_URL`).
|
|
60
|
+
The plan page issues exactly one API key and it is the key tested above.
|
|
61
|
+
|
|
62
|
+
Net: this engine lane has no funded model until a key with actual balance or
|
|
63
|
+
an in-plan API path exists. Do not dispatch `--engine zcode` until then.
|