pi-multikey 1.3.0 → 1.4.0

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 (4) hide show
  1. package/README.md +4 -3
  2. package/README.zh.md +4 -3
  3. package/package.json +1 -1
  4. package/presets.ts +15 -1
package/README.md CHANGED
@@ -48,11 +48,12 @@ Built-in presets decouple "model settings" from "keys". The data comes from b.ai
48
48
 
49
49
  ### OpenCode Zen (free tier)
50
50
 
51
- Endpoint `https://opencode.ai/zen/v1`; keys from [opencode.ai/auth](https://opencode.ai/auth) → workspace Keys. Context / max-output are the **Zen free-tier serving limits** (consistent across models.dev `opencode` provider + pi's built-in opencode catalog); the raw models are bigger — MiMo V2.5 = 1M ctx, Hy3 = 262K ctx. `muse-spark-1.2-contributor-free` uses the OpenAI **Responses** API endpoint; the other six use chat completions.
51
+ Endpoint `https://opencode.ai/zen/v1`; keys from [opencode.ai/auth](https://opencode.ai/auth) → workspace Keys. Context / max-output are the **Zen free-tier serving limits** (consistent across models.dev `opencode` provider + pi's built-in opencode catalog); the raw models are bigger — MiMo V2.5 = 1M ctx, Hy3 = 262K ctx. `muse-spark-1.2-contributor-free` uses the OpenAI **Responses** API endpoint; the other seven use chat completions.
52
52
 
53
53
  | Model | ctx / max-out | Modalities | Supported thinking levels |
54
54
  |---|---|---|---|
55
55
  | big-pickle | 200K / 32K | text | always-on (no thinkingLevelMap, like pi's catalog) |
56
+ | deepseek-v4-flash-free | 200K / 128K | text | off · low · high · max (mirrors b.ai's deepseek-v4-flash) |
56
57
  | mimo-v2.5-free | 200K / 32K | text+image | always-on (no thinkingLevelMap) |
57
58
  | hy3-free | 190K / 64K | text | low · medium · high (no off) |
58
59
  | ling-3.0-flash-fin-free | 256K / 32K | text | always-on (no thinkingLevelMap) |
@@ -60,7 +61,7 @@ Endpoint `https://opencode.ai/zen/v1`; keys from [opencode.ai/auth](https://open
60
61
  | nemotron-3.5-lightning-free | 256K / 256K | text | always-on (no thinkingLevelMap) |
61
62
  | muse-spark-1.2-contributor-free | 1M / 128K | text+image | minimal · low · medium · high · xhigh (no off, Responses API) |
62
63
 
63
- > All seven models are free (zero per-token cost) for a limited time while OpenCode collects feedback; data may be used to improve the models (Nemotron free endpoints are NVIDIA trials — don't send confidential data).
64
+ > All eight models are free (zero per-token cost) for a limited time while OpenCode collects feedback; data may be used to improve the models (Nemotron free endpoints are NVIDIA trials — don't send confidential data).
64
65
 
65
66
  To add a preset: append one entry to the `PRESETS` array in `presets.ts`.
66
67
 
@@ -116,7 +117,7 @@ The detected header style is stored as `"auth": "api-key"` only when the endpoin
116
117
  │ └─ Delete pool
117
118
  ├─ Add pool…
118
119
  │ ├─ Preset: B.AI all model settings preloaded; paste keys (verified by a probe) and you're done
119
- │ ├─ Preset: OpenCode Zen free-tier models preloaded (7 models); paste keys and you're done
120
+ │ ├─ Preset: OpenCode Zen free-tier models preloaded (8 models); paste keys and you're done
120
121
  │ └─ Custom… id + base URL + keys, then auto-probe, model multi-select, safe defaults
121
122
  └─ Reload config from disk
122
123
  ```
package/README.zh.md CHANGED
@@ -54,11 +54,12 @@ DeepSeek / Tencent / 小米官方文档，并对每个 thinking 档位做过实
54
54
  端点 `https://opencode.ai/zen/v1`；密钥从 [opencode.ai/auth](https://opencode.ai/auth) → workspace Keys 获取。
55
55
  上下文/最大输出为 **Zen 免费层限制**（models.dev `opencode` provider 与 pi 内置 opencode 目录一致）；
56
56
  原始模型更大——MiMo V2.5 = 1M ctx，Hy3 = 262K ctx。
57
- `muse-spark-1.2-contributor-free` 使用 OpenAI **Responses** API；其余六个使用 chat completions。
57
+ `muse-spark-1.2-contributor-free` 使用 OpenAI **Responses** API；其余七个使用 chat completions。
58
58
 
59
59
  | 模型 | ctx / max-out | 模态 | 生效 thinking 档位 |
60
60
  |---|---|---|---|
61
61
  | big-pickle | 200K / 32K | text | 始终思考（无 thinkingLevelMap，与 pi 内置目录一致） |
62
+ | deepseek-v4-flash-free | 200K / 128K | text | off · low · high · max（沿用 b.ai preset 的 deepseek-v4-flash） |
62
63
  | mimo-v2.5-free | 200K / 32K | text+image | 始终思考（无 thinkingLevelMap） |
63
64
  | hy3-free | 190K / 64K | text | low · medium · high（无 off） |
64
65
  | ling-3.0-flash-fin-free | 256K / 32K | text | 始终思考（无 thinkingLevelMap） |
@@ -66,7 +67,7 @@ DeepSeek / Tencent / 小米官方文档，并对每个 thinking 档位做过实
66
67
  | nemotron-3.5-lightning-free | 256K / 256K | text | 始终思考（无 thinkingLevelMap） |
67
68
  | muse-spark-1.2-contributor-free | 1M / 128K | text+image | minimal · low · medium · high · xhigh（无 off，Responses API） |
68
69
 
69
- > 以上七个模型在 OpenCode 收集反馈期间均免费（零 token 费用）；数据可能用于改进模型（Nemotron 免费端点为 NVIDIA 试用——请勿提交机密数据）。
70
+ > 以上八个模型在 OpenCode 收集反馈期间均免费（零 token 费用）；数据可能用于改进模型（Nemotron 免费端点为 NVIDIA 试用——请勿提交机密数据）。
70
71
 
71
72
  新增 preset：在 `presets.ts` 的 `PRESETS` 数组里加一项即可。
72
73
 
@@ -125,7 +126,7 @@ JSON 后 `Reload config from disk`。
125
126
  │ └─ Delete pool
126
127
  ├─ Add pool…
127
128
  │ ├─ Preset: B.AI 预置全部模型设定，粘贴 key（自动校验）即可用
128
- │ ├─ Preset: OpenCode Zen 免费层模型预置（7 个模型），粘贴 key 即可用
129
+ │ ├─ Preset: OpenCode Zen 免费层模型预置（8 个模型），粘贴 key 即可用
129
130
  │ └─ Custom… 只填 id + Base URL + key，随后自动探测、多选模型、安全默认值
130
131
  └─ Reload config from disk
131
132
  ```
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "pi-multikey",
3
- "version": "1.3.0",
3
+ "version": "1.4.0",
4
4
  "description": "One pi provider backed by many API keys: automatic 429 rotation, per-request key leases for concurrent subagents, and a /multikey management TUI",
5
5
  "keywords": [
6
6
  "pi-package",
package/presets.ts CHANGED
@@ -131,7 +131,7 @@ export const PRESETS: Preset[] = [
131
131
  {
132
132
  id: "opencode-zen",
133
133
  name: "OpenCode Zen",
134
- description: "opencode.ai/zen free tier — Big Pickle, MiMo V2.5, Hy3, Ling 3.0 Fin, Nemotron 3 Ultra/Lightning, Muse Spark 1.2 (7 free models)",
134
+ description: "opencode.ai/zen free tier — Big Pickle, DeepSeek V4 Flash, MiMo V2.5, Hy3, Ling 3.0 Fin, Nemotron 3 Ultra/Lightning, Muse Spark 1.2 (8 free models)",
135
135
  defaultPoolId: "zen",
136
136
  baseUrl: "https://opencode.ai/zen/v1",
137
137
  api: "openai-completions",
@@ -148,6 +148,20 @@ export const PRESETS: Preset[] = [
148
148
  maxTokens: 32_000,
149
149
  compat: ZEN_CHAT_COMPAT,
150
150
  },
151
+ {
152
+ // DeepSeek V4 Flash on the Zen FREE tier — id confirmed via GET /zen/v1/models
153
+ // (deepseek-v4-flash-free); serving limits 200K/128K from models.dev `opencode`
154
+ // provider. Thinking levels mirror the b.ai preset's deepseek-v4-flash
155
+ // (off/low/high/max; medium & xhigh alias to high server-side).
156
+ id: "deepseek-v4-flash-free",
157
+ name: "DeepSeek V4 Flash Free",
158
+ reasoning: true,
159
+ input: ["text"],
160
+ contextWindow: 200_000,
161
+ maxTokens: 128_000,
162
+ thinkingLevelMap: levels({ off: "none", low: "low", high: "high", max: "max" }),
163
+ compat: ZEN_CHAT_COMPAT,
164
+ },
151
165
  {
152
166
  // Xiaomi MiMo V2.5 omni; raw model is 1M ctx but the Zen FREE tier serves 200K/32K.
153
167
  id: "mimo-v2.5-free",