pi-multikey 1.3.0 → 1.5.0

package/README.md

@@ -48,11 +48,14 @@ Built-in presets decouple "model settings" from "keys". The data comes from b.ai
 
 ### OpenCode Zen (free tier)
 
-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.
+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.
+
+Requests to this endpoint automatically carry the `User-Agent` header that opencode.ai expects (`opencode/1.15.0 ai-sdk/provider-utils/4.0.23 runtime/bun/1.3.13`) — appended whenever a pool's baseUrl is `https://opencode.ai/zen/v1`, on both probes and live requests.
 
 | Model | ctx / max-out | Modalities | Supported thinking levels |
 |---|---|---|---|
 | big-pickle | 200K / 32K | text | always-on (no thinkingLevelMap, like pi's catalog) |
+| deepseek-v4-flash-free | 200K / 128K | text | off · low · high · max (mirrors b.ai's deepseek-v4-flash) |
 | mimo-v2.5-free | 200K / 32K | text+image | always-on (no thinkingLevelMap) |
 | hy3-free | 190K / 64K | text | low · medium · high (no off) |
 | ling-3.0-flash-fin-free | 256K / 32K | text | always-on (no thinkingLevelMap) |
@@ -60,7 +63,7 @@ Endpoint `https://opencode.ai/zen/v1`; keys from [opencode.ai/auth](https://open
 | nemotron-3.5-lightning-free | 256K / 256K | text | always-on (no thinkingLevelMap) |
 | muse-spark-1.2-contributor-free | 1M / 128K | text+image | minimal · low · medium · high · xhigh (no off, Responses API) |
 
-> 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).
+> 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).
 
 To add a preset: append one entry to the `PRESETS` array in `presets.ts`.
 
@@ -116,7 +119,7 @@ The detected header style is stored as `"auth": "api-key"` only when the endpoin
 │  └─ Delete pool
 ├─ Add pool…
 │  ├─ Preset: B.AI           all model settings preloaded; paste keys (verified by a probe) and you're done
-│  ├─ Preset: OpenCode Zen   free-tier models preloaded (7 models); paste keys and you're done
+│  ├─ Preset: OpenCode Zen   free-tier models preloaded (8 models); paste keys and you're done
 │  └─ Custom…                id + base URL + keys, then auto-probe, model multi-select, safe defaults
 └─ Reload config from disk
 ```

package/README.zh.md

@@ -54,11 +54,14 @@ DeepSeek / Tencent / 小米官方文档，并对每个 thinking 档位做过实
 端点 `https://opencode.ai/zen/v1`；密钥从 [opencode.ai/auth](https://opencode.ai/auth) → workspace Keys 获取。
 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）；
 原始模型更大——MiMo V2.5 = 1M ctx，Hy3 = 262K ctx。
-`muse-spark-1.2-contributor-free` 使用 OpenAI **Responses** API；其余六个使用 chat completions。
+`muse-spark-1.2-contributor-free` 使用 OpenAI **Responses** API；其余七个使用 chat completions。
+
+对该端点的请求会自动携带 opencode.ai 期望的 `User-Agent` 头（`opencode/1.15.0 ai-sdk/provider-utils/4.0.23 runtime/bun/1.3.13`）——只要 pool 的 baseUrl 是 `https://opencode.ai/zen/v1`，探测和真实请求都会附加。
 
 | 模型 | ctx / max-out | 模态 | 生效 thinking 档位 |
 |---|---|---|---|
 | big-pickle | 200K / 32K | text | 始终思考（无 thinkingLevelMap，与 pi 内置目录一致） |
+| deepseek-v4-flash-free | 200K / 128K | text | off · low · high · max（沿用 b.ai preset 的 deepseek-v4-flash） |
 | mimo-v2.5-free | 200K / 32K | text+image | 始终思考（无 thinkingLevelMap） |
 | hy3-free | 190K / 64K | text | low · medium · high（无 off） |
 | ling-3.0-flash-fin-free | 256K / 32K | text | 始终思考（无 thinkingLevelMap） |
@@ -66,7 +69,7 @@ DeepSeek / Tencent / 小米官方文档，并对每个 thinking 档位做过实
 | nemotron-3.5-lightning-free | 256K / 256K | text | 始终思考（无 thinkingLevelMap） |
 | muse-spark-1.2-contributor-free | 1M / 128K | text+image | minimal · low · medium · high · xhigh（无 off，Responses API） |
 
-> 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）。
+> 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）。
 
 新增 preset：在 `presets.ts` 的 `PRESETS` 数组里加一项即可。
 
@@ -125,7 +128,7 @@ JSON 后 `Reload config from disk`。
 │  └─ Delete pool
 ├─ Add pool…
 │  ├─ Preset: B.AI           预置全部模型设定，粘贴 key（自动校验）即可用
-│  ├─ Preset: OpenCode Zen   免费层模型预置（7 个模型），粘贴 key 即可用
+│  ├─ Preset: OpenCode Zen   免费层模型预置（8 个模型），粘贴 key 即可用
 │  └─ Custom…                只填 id + Base URL + key，随后自动探测、多选模型、安全默认值
 └─ Reload config from disk
 ```

package/config.ts

@@ -339,3 +339,24 @@ export function maskKey(key: string): string {
 	if (key.length <= 10) return "…";
 	return `${key.slice(0, 6)}…${key.slice(-4)}`;
 }
+
+// ── Endpoint-required headers ────────────────────────────────────────────────
+
+/** OpenCode Zen free tier endpoint that requires a specific User-Agent header. */
+const OPENCODE_ZEN_BASE_URL = "https://opencode.ai/zen/v1";
+const OPENCODE_ZEN_USER_AGENT =
+	"opencode/1.15.0 ai-sdk/provider-utils/4.0.23 runtime/bun/1.3.13";
+
+/**
+ * Headers that must be sent to known endpoints.
+ *
+ * Currently: OpenCode Zen's free tier endpoint requires this exact User-Agent;
+ * anything else returns an empty object.
+ *
+ * The comparison is case-insensitive and tolerates a trailing slash.
+ */
+export function endpointHeaders(baseUrl: string): Record<string, string> {
+	const normalized = baseUrl.replace(/\/+$/, "").toLowerCase();
+	if (normalized === OPENCODE_ZEN_BASE_URL) return { "User-Agent": OPENCODE_ZEN_USER_AGENT };
+	return {};
+}

package/index.ts

@@ -11,7 +11,7 @@
 
 import type { ExtensionAPI, ExtensionContext } from "@earendil-works/pi-coding-agent";
 import { getApiProvider, type Api } from "@earendil-works/pi-ai";
-import { configPath, loadConfig, saveConfig, toProviderModels, type KeypoolConfig, type PoolConfig } from "./config.ts";
+import { configPath, endpointHeaders, loadConfig, saveConfig, toProviderModels, type KeypoolConfig, type PoolConfig } from "./config.ts";
 import { KeyPool } from "./pool.ts";
 import { createRotatingStreamSimple } from "./stream.ts";
 import { runManager, type ManagerHooks } from "./manage.ts";
@@ -69,7 +69,7 @@ export default function multikey(pi: ExtensionAPI) {
 			// Real keys are injected per-request by the rotating stream function.
 			apiKey: "multikey-managed",
 			api,
-			headers: pool.headers,
+			headers: { ...pool.headers, ...endpointHeaders(pool.baseUrl) },
 			models: toProviderModels(pool),
 			streamSimple: createRotatingStreamSimple(keyPool, api, notify),
 		});

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "pi-multikey",
-	"version": "1.3.0",
+	"version": "1.5.0",
 	"description": "One pi provider backed by many API keys: automatic 429 rotation, per-request key leases for concurrent subagents, and a /multikey management TUI",
 	"keywords": [
 		"pi-package",

package/presets.ts

@@ -131,7 +131,7 @@ export const PRESETS: Preset[] = [
 	{
 		id: "opencode-zen",
 		name: "OpenCode Zen",
-		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)",
+		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)",
 		defaultPoolId: "zen",
 		baseUrl: "https://opencode.ai/zen/v1",
 		api: "openai-completions",
@@ -148,6 +148,20 @@ export const PRESETS: Preset[] = [
 				maxTokens: 32_000,
 				compat: ZEN_CHAT_COMPAT,
 			},
+			{
+				// DeepSeek V4 Flash on the Zen FREE tier — id confirmed via GET /zen/v1/models
+				// (deepseek-v4-flash-free); serving limits 200K/128K from models.dev `opencode`
+				// provider. Thinking levels mirror the b.ai preset's deepseek-v4-flash
+				// (off/low/high/max; medium & xhigh alias to high server-side).
+				id: "deepseek-v4-flash-free",
+				name: "DeepSeek V4 Flash Free",
+				reasoning: true,
+				input: ["text"],
+				contextWindow: 200_000,
+				maxTokens: 128_000,
+				thinkingLevelMap: levels({ off: "none", low: "low", high: "high", max: "max" }),
+				compat: ZEN_CHAT_COMPAT,
+			},
 			{
 				// Xiaomi MiMo V2.5 omni; raw model is 1M ctx but the Zen FREE tier serves 200K/32K.
 				id: "mimo-v2.5-free",

package/probe.ts

@@ -1,3 +1,5 @@
+import { endpointHeaders } from "./config.ts";
+
 /**
  * Endpoint probing: auto-detect the auth header style and fetch the model list.
  *
@@ -57,11 +59,11 @@ function authHeaders(style: AuthStyle, key: string): Record<string, string> {
 	return style === "bearer" ? { Authorization: `Bearer ${key}` } : { "x-api-key": key };
 }
 
-async function fetchJson(url: string, headers: Record<string, string>, timeoutMs: number): Promise<{ status: number; body?: unknown }> {
+async function fetchJson(url: string, headers: Record<string, string>, timeoutMs: number, baseUrl?: string): Promise<{ status: number; body?: unknown }> {
 	try {
 		const response = await fetch(url, {
 			method: "GET",
-			headers: { Accept: "application/json", ...headers },
+			headers: { Accept: "application/json", ...endpointHeaders(baseUrl ?? url), ...headers },
 			signal: AbortSignal.timeout(timeoutMs),
 		});
 		let body: unknown;
@@ -147,7 +149,7 @@ async function chatProbe(baseUrl: string, style: AuthStyle, key: string, modelId
 	try {
 		const response = await fetch(`${trimSlash(baseUrl)}/chat/completions`, {
 			method: "POST",
-			headers: { "Content-Type": "application/json", ...authHeaders(style, key) },
+			headers: { "Content-Type": "application/json", ...endpointHeaders(baseUrl), ...authHeaders(style, key) },
 			body: JSON.stringify({ model: modelId, max_tokens: 4, messages: [{ role: "user", content: "ping" }] }),
 			signal: AbortSignal.timeout(CHAT_TIMEOUT_MS),
 		});
@@ -186,7 +188,7 @@ export async function probeEndpoint(
 			let result: { status: number; body?: unknown };
 			try {
 				emit(`GET ${url} (${style === "bearer" ? "Authorization: Bearer" : "x-api-key"})…`);
-				result = await fetchJson(url, authHeaders(style, key), MODELS_TIMEOUT_MS);
+				result = await fetchJson(url, authHeaders(style, key), MODELS_TIMEOUT_MS, baseUrl);
 			} catch (error) {
 				emit(`   network error: ${error instanceof Error ? error.message : String(error)}`);
 				continue;