@bacnh85/pi-subagent 0.11.0 → 0.12.3

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.
Files changed (9) hide show
  1. package/CHANGELOG.md +15 -0
  2. package/README.md +17 -15
  3. package/agents/general-purpose.md +2 -2
  4. package/agents/planner.md +1 -2
  5. package/agents/reviewer.md +1 -2
  6. package/agents/scout.md +1 -1
  7. package/agents/tester.md +1 -1
  8. package/agents/worker.md +2 -2
  9. package/package.json +20 -12
package/CHANGELOG.md CHANGED
@@ -1,5 +1,20 @@
1
1
  # Changelog
2
2
 
3
+ ## 0.12.3 (2026-07-30)
4
+
5
+ ### Improvements
6
+
7
+ - Patch version bump for release sync and package documentation update.
8
+
9
+ ## 0.12.0 (2026-07-24)
10
+
11
+ ### Model routing
12
+
13
+ - Chains are now **free-first** to conserve the metered opencode-go budget: paid DeepSeek moves to the **last** position so free nvidia/openrouter fallbacks are tried first. Previously DeepSeek sat at position 2 and burned quota whenever the free GLM primary rate-limited.
14
+ - **Removed two dead fallbacks** found by live-testing every free model on 2026-07-24: `nvidia/moonshotai/kimi-k2.6` (HTTP 404, NVCF not provisioned for the account) and `nvidia/z-ai/glm-5.2` (timeout). These are non-rate-limit failures, so the rate-limit retry loop did not advance past them — a subagent that reached them died instead of falling through to the working `:free` entries behind them.
15
+ - New chains: scout/tester `glm-5-turbo` → `nvidia/gpt-oss-20b` → `deepseek-v4-flash`; worker/general-purpose `glm-5.1` → `nvidia/mistral-small-4-119b-2603` → `nemotron-3-super:free` → `deepseek-v4-flash`; planner/reviewer `glm-5.2` → `nemotron-3-ultra:free` → `deepseek-v4-pro`.
16
+ - Verified-working free additions: `nvidia/openai/gpt-oss-20b` (1.2s), `nvidia/mistralai/mistral-small-4-119b-2603` (2.6s, 119B reasoning), `openrouter/nvidia/nemotron-3-nano-30b-a3b:free` (1.2s).
17
+
3
18
  ## 0.11.0 (2026-07-23)
4
19
 
5
20
  ### Model routing
package/README.md CHANGED
@@ -1,4 +1,4 @@
1
- # pi-subagent
1
+ # @bacnh85/pi-subagent
2
2
 
3
3
  Isolated in-process subagents for Pi. The `subagent` tool supports single, parallel (8 tasks, 4 concurrent), and chained execution; `/agent` opens inspectable child threads.
4
4
 
@@ -6,8 +6,6 @@ Isolated in-process subagents for Pi. The `subagent` tool supports single, paral
6
6
 
7
7
  ```bash
8
8
  pi install npm:@bacnh85/pi-subagent
9
- # local checkout
10
- pi install ./pi-subagent
11
9
  ```
12
10
 
13
11
  Requires Node.js >= 20.18.
@@ -16,14 +14,14 @@ Requires Node.js >= 20.18.
16
14
 
17
15
  | Role | Ordered model preferences | Thinking | Tools |
18
16
  | --- | --- | --- | --- |
19
- | `scout` | `zai-coding-cn/glm-5-turbo` → `opencode-go/deepseek-v4-flash` → `nvidia/openai/gpt-oss-20b` | off | read, grep, find, ls |
20
- | `tester` | `zai-coding-cn/glm-5-turbo` → `opencode-go/deepseek-v4-flash` → `nvidia/openai/gpt-oss-20b` | off | read, bash, grep, find, ls |
21
- | `worker` | `zai-coding-cn/glm-5.1` → `opencode-go/deepseek-v4-flash` → `nvidia/moonshotai/kimi-k2.6` → `openrouter/nvidia/nemotron-3-super-120b-a12b:free` | medium | read, bash, edit, write, grep, find, ls |
22
- | `general-purpose` | `zai-coding-cn/glm-5.1` → `opencode-go/deepseek-v4-flash` → `nvidia/moonshotai/kimi-k2.6` → `openrouter/nvidia/nemotron-3-super-120b-a12b:free` | medium | read, bash, edit, write, grep, find, ls |
23
- | `planner` | `zai-coding-cn/glm-5.2` → `nvidia/z-ai/glm-5.2` → `opencode-go/deepseek-v4-pro` → `openrouter/nvidia/nemotron-3-ultra-550b-a55b:free` | high | read, grep, find, ls |
24
- | `reviewer` | `zai-coding-cn/glm-5.2` → `nvidia/z-ai/glm-5.2` → `opencode-go/deepseek-v4-pro` → `openrouter/nvidia/nemotron-3-ultra-550b-a55b:free` | high | read, grep, find, ls |
17
+ | `scout` | `zai-coding-cn/glm-5-turbo` → `nvidia/openai/gpt-oss-20b` → `opencode-go/deepseek-v4-flash` | off | read, grep, find, ls |
18
+ | `tester` | `zai-coding-cn/glm-5-turbo` → `nvidia/openai/gpt-oss-20b` → `opencode-go/deepseek-v4-flash` | off | read, bash, grep, find, ls |
19
+ | `worker` | `zai-coding-cn/glm-5.1` → `nvidia/mistralai/mistral-small-4-119b-2603` → `openrouter/nvidia/nemotron-3-super-120b-a12b:free` → `opencode-go/deepseek-v4-flash` | medium | read, bash, edit, write, grep, find, ls |
20
+ | `general-purpose` | `zai-coding-cn/glm-5.1` → `nvidia/mistralai/mistral-small-4-119b-2603` → `openrouter/nvidia/nemotron-3-super-120b-a12b:free` → `opencode-go/deepseek-v4-flash` | medium | read, bash, edit, write, grep, find, ls |
21
+ | `planner` | `zai-coding-cn/glm-5.2` → `openrouter/nvidia/nemotron-3-ultra-550b-a55b:free` → `opencode-go/deepseek-v4-pro` | high | read, grep, find, ls |
22
+ | `reviewer` | `zai-coding-cn/glm-5.2` → `openrouter/nvidia/nemotron-3-ultra-550b-a55b:free` → `opencode-go/deepseek-v4-pro` | high | read, grep, find, ls |
25
23
 
26
- Each role uses the first authenticated preference available through Pi's model registry, then falls back to the authenticated parent model. Chains are provider-diverse and cost-ascending on failure: **zai-coding-cn** (free GLM, primary) → **opencode-go** (one DeepSeek per role, matched to strength: `deepseek-v4-flash` for fast/strong-coding, `deepseek-v4-pro` for deep reasoning) → **nvidia** (free NIM) → **openrouter** (`:free` models, last resort — omitted from scout/tester, whose `thinking: off` is incompatible with reasoning-mandatory `:free` models). opencode-go is metered — its GLM models cost ~$1.40/$4.40 per M versus zai-coding-cn's free GLM, so GLM stays on zai-coding-cn and opencode-go contributes only its DeepSeek models. User/project agent files remain stronger overrides and may set legacy `model`, ordered `models`, and `thinking`.
24
+ Each role uses the first authenticated preference available through Pi's model registry, then falls back to the authenticated parent model. Chains are **free-first** to conserve the metered opencode-go budget: **zai-coding-cn** (free GLM, primary) → free **nvidia** NIM and **openrouter** `:free` models → **opencode-go** (paid DeepSeek, last resort — one per role: `deepseek-v4-flash` for fast/strong-coding, `deepseek-v4-pro` for deep reasoning). opencode-go's GLM models cost ~$1.40/$4.40 per M versus zai-coding-cn's free GLM, so GLM stays on zai-coding-cn. Fallback models were live-verified on 2026-07-24; `nvidia/moonshotai/kimi-k2.6` and `nvidia/z-ai/glm-5.2` return 404/timeout on the user's account and were removed — non-rate-limit failures kill the subagent instead of advancing the chain. User/project agent files remain stronger overrides and may set legacy `model`, ordered `models`, and `thinking`.
27
25
 
28
26
  ## Agent files
29
27
 
@@ -36,8 +34,8 @@ description: Locate relevant files and symbols
36
34
  tools: read, grep, find, ls
37
35
  models:
38
36
  - zai-coding-cn/glm-5-turbo
39
- - opencode-go/deepseek-v4-flash
40
37
  - nvidia/openai/gpt-oss-20b
38
+ - opencode-go/deepseek-v4-flash
41
39
  ---
42
40
 
43
41
  Return concise evidence with file/symbol anchors.
@@ -171,12 +169,16 @@ The raw `stopReason` from the Pi SDK is preserved in the result.
171
169
 
172
170
  `pi-subagent` owns the `pi-subagent:run` event contract for one named-agent request. `pi-review` uses it for isolated review. Requests use an immediate boolean `accept()` claim and exactly one `respond()` callback; this suppresses duplicate responders while missing services and timeouts remain caller-controlled.
173
171
 
172
+ ## Changelog
173
+
174
+ See [CHANGELOG.md](CHANGELOG.md) for release history.
175
+
174
176
  ## Compatibility
175
177
 
176
- - Requires `@earendil-works/pi-coding-agent >=0.80.0 <0.81.0`
177
- - Requires `@earendil-works/pi-ai >=0.80.0 <0.81.0`
178
- - Requires `@earendil-works/pi-agent-core >=0.80.0 <0.81.0`
179
- - Requires `@earendil-works/pi-tui >=0.80.0 <0.81.0`
178
+ - Requires `@earendil-works/pi-coding-agent >=0.80.0 <0.83.0`
179
+ - Requires `@earendil-works/pi-ai >=0.80.0 <0.83.0`
180
+ - Requires `@earendil-works/pi-agent-core >=0.80.0 <0.83.0`
181
+ - Requires `@earendil-works/pi-tui >=0.80.0 <0.83.0`
180
182
  - Requires `typebox >=1.3.0 <2.0.0`
181
183
  - Requires Node.js >= 20.18
182
184
 
package/agents/general-purpose.md CHANGED
@@ -4,9 +4,9 @@ description: General-purpose sub-agent for any delegated task. Use when no speci
4
4
  tools: read, bash, edit, write, grep, find, ls
5
5
  models:
6
6
  - zai-coding-cn/glm-5.1
7
- - opencode-go/deepseek-v4-flash
8
- - nvidia/moonshotai/kimi-k2.6
7
+ - nvidia/mistralai/mistral-small-4-119b-2603
9
8
  - openrouter/nvidia/nemotron-3-super-120b-a12b:free
9
+ - opencode-go/deepseek-v4-flash
10
10
  thinking: medium
11
11
  color: yellow
12
12
  ---
package/agents/planner.md CHANGED
@@ -4,9 +4,8 @@ description: Read-only planning and architecture specialist. Use for consequenti
4
4
  tools: read, grep, find, ls
5
5
  models:
6
6
  - zai-coding-cn/glm-5.2
7
- - nvidia/z-ai/glm-5.2
8
- - opencode-go/deepseek-v4-pro
9
7
  - openrouter/nvidia/nemotron-3-ultra-550b-a55b:free
8
+ - opencode-go/deepseek-v4-pro
10
9
  thinking: high
11
10
  color: blue
12
11
  sandbox: read-only
package/agents/reviewer.md CHANGED
@@ -4,9 +4,8 @@ description: Code review specialist. Use for correctness, security, regression,
4
4
  tools: read, grep, find, ls
5
5
  models:
6
6
  - zai-coding-cn/glm-5.2
7
- - nvidia/z-ai/glm-5.2
8
- - opencode-go/deepseek-v4-pro
9
7
  - openrouter/nvidia/nemotron-3-ultra-550b-a55b:free
8
+ - opencode-go/deepseek-v4-pro
10
9
  thinking: high
11
10
  color: purple
12
11
  sandbox: read-only
package/agents/scout.md CHANGED
@@ -4,8 +4,8 @@ description: Fast codebase recon that returns compressed context for handoff. Us
4
4
  tools: read, grep, find, ls
5
5
  models:
6
6
  - zai-coding-cn/glm-5-turbo
7
- - opencode-go/deepseek-v4-flash
8
7
  - nvidia/openai/gpt-oss-20b
8
+ - opencode-go/deepseek-v4-flash
9
9
  thinking: off
10
10
  color: cyan
11
11
  sandbox: read-only
package/agents/tester.md CHANGED
@@ -4,8 +4,8 @@ description: Focused verification agent. Use for cheap routine test, typecheck,
4
4
  tools: read, bash, grep, find, ls
5
5
  models:
6
6
  - zai-coding-cn/glm-5-turbo
7
- - opencode-go/deepseek-v4-flash
8
7
  - nvidia/openai/gpt-oss-20b
8
+ - opencode-go/deepseek-v4-flash
9
9
  thinking: off
10
10
  color: orange
11
11
  ---
package/agents/worker.md CHANGED
@@ -3,9 +3,9 @@ name: worker
3
3
  description: General-purpose coding agent with full tool access. Use only when explicitly requested for isolated implementation.
4
4
  models:
5
5
  - zai-coding-cn/glm-5.1
6
- - opencode-go/deepseek-v4-flash
7
- - nvidia/moonshotai/kimi-k2.6
6
+ - nvidia/mistralai/mistral-small-4-119b-2603
8
7
  - openrouter/nvidia/nemotron-3-super-120b-a12b:free
8
+ - opencode-go/deepseek-v4-flash
9
9
  thinking: medium
10
10
  color: green
11
11
  ---
package/package.json CHANGED
@@ -1,17 +1,21 @@
1
1
  {
2
2
  "name": "@bacnh85/pi-subagent",
3
- "version": "0.11.0",
3
+ "version": "0.12.3",
4
4
  "description": "In-process subagents for Pi with isolated SDK sessions, parallel and chained delegation, and inspectable threads.",
5
5
  "type": "module",
6
6
  "license": "MIT",
7
- "publishConfig": { "access": "public" },
7
+ "publishConfig": {
8
+ "access": "public"
9
+ },
8
10
  "homepage": "https://github.com/bacnh85/pi-extensions/tree/main/pi-subagent",
9
11
  "repository": {
10
12
  "type": "git",
11
13
  "url": "git+https://github.com/bacnh85/pi-extensions.git",
12
14
  "directory": "pi-subagent"
13
15
  },
14
- "bugs": { "url": "https://github.com/bacnh85/pi-extensions/issues" },
16
+ "bugs": {
17
+ "url": "https://github.com/bacnh85/pi-extensions/issues"
18
+ },
15
19
  "keywords": [
16
20
  "pi-package",
17
21
  "pi-extension",
@@ -38,7 +42,11 @@
38
42
  "extensions/security.ts",
39
43
  "extensions/package.json"
40
44
  ],
41
- "pi": { "extensions": ["./extensions/index.ts"] },
45
+ "pi": {
46
+ "extensions": [
47
+ "./extensions/index.ts"
48
+ ]
49
+ },
42
50
  "engines": {
43
51
  "node": ">=20.18"
44
52
  },
@@ -48,17 +56,17 @@
48
56
  "check": "npm run typecheck && npm test"
49
57
  },
50
58
  "peerDependencies": {
51
- "@earendil-works/pi-coding-agent": ">=0.80.0 <0.82.0",
52
- "@earendil-works/pi-ai": ">=0.80.0 <0.82.0",
53
- "@earendil-works/pi-agent-core": ">=0.80.0 <0.82.0",
54
- "@earendil-works/pi-tui": ">=0.80.0 <0.82.0",
59
+ "@earendil-works/pi-coding-agent": ">=0.80.0 <0.83.0",
60
+ "@earendil-works/pi-ai": ">=0.80.0 <0.83.0",
61
+ "@earendil-works/pi-agent-core": ">=0.80.0 <0.83.0",
62
+ "@earendil-works/pi-tui": ">=0.80.0 <0.83.0",
55
63
  "typebox": ">=1.3.0 <2.0.0"
56
64
  },
57
65
  "devDependencies": {
58
- "@earendil-works/pi-agent-core": "^0.81.0",
59
- "@earendil-works/pi-ai": "^0.81.0",
60
- "@earendil-works/pi-coding-agent": "^0.81.0",
61
- "@earendil-works/pi-tui": "^0.81.0",
66
+ "@earendil-works/pi-agent-core": "^0.82.0",
67
+ "@earendil-works/pi-ai": "^0.82.0",
68
+ "@earendil-works/pi-coding-agent": "^0.82.0",
69
+ "@earendil-works/pi-tui": "^0.82.0",
62
70
  "@types/mocha": "^10.0.10",
63
71
  "@types/node": "^20.19.43",
64
72
  "mocha": "^10.8.2",