refactor(providers): remove the Requesty built-in provider

Requesty (added in #474) is an OpenAI-compatible gateway whose capabilities were a copy of OpenRouter's: same provider/model naming, same extra_body.reasoning opt-in, no adapter code of its own. Built-in slots are kept for providers with their own behaviour; a gateway that speaks an existing wire format is configured through that provider, and #1532 (API Route) is being closed on the same grounds. Migration, identical request shape: LANGCHAIN_PROVIDER=openrouter, OPENROUTER_BASE_URL=https://router.requesty.ai/v1, OPENROUTER_API_KEY=<the Requesty key>. Past news and contributor credits in the READMEs are left as history; the supported-provider lists drop it.
2026-09-21 20:35:37 +08:00
1 parent 61972397b4
commit 051aeb5509
@@ -854,7 +854,7 @@ vibe-trading-mcp               # start MCP server (stdio)
 - OpenAI Codex can also be used with ChatGPT OAuth: set `LANGCHAIN_PROVIDER=openai-codex`, then run `vibe-trading provider login openai-codex`. This does not use `OPENAI_API_KEY`.
 - GitHub Copilot can be used with an active Copilot subscription instead of a separately billed LLM API key. See [GitHub Copilot SDK provider](#github-copilot-sdk-provider).

-> **Supported LLM providers:** OpenRouter, Requesty, OpenAI, Anthropic (native Messages API), DeepSeek, OpenCode (Go / Zen), Gemini, Groq, DashScope/Qwen, Zhipu, Moonshot/Kimi, MiniMax, SiliconFlow (CN + Global), Xiaomi MIMO, Novita AI, iFlytek Spark, Z.ai, NVIDIA NIM, ModelScope, GitHub Copilot, Ollama (local). When no `*_BASE_URL` is set, each provider falls back to its canonical endpoint, so just a key is enough. See `.env.example` for config.
+> **Supported LLM providers:** OpenRouter, OpenAI, Anthropic (native Messages API), DeepSeek, OpenCode (Go / Zen), Gemini, Groq, DashScope/Qwen, Zhipu, Moonshot/Kimi, MiniMax, SiliconFlow (CN + Global), Xiaomi MIMO, Novita AI, iFlytek Spark, Z.ai, NVIDIA NIM, ModelScope, GitHub Copilot, Ollama (local). When no `*_BASE_URL` is set, each provider falls back to its canonical endpoint, so just a key is enough. See `.env.example` for config.

 > **Tip:** All markets work without any API keys thanks to automatic fallback. yfinance/Yahoo (HK/US/Canada/UK), OKX (crypto), mootdx (A-shares, TCP-direct, no IP throttle), and AKShare (A-shares, US, HK, futures, forex) are all free. LSE `.L` quotes must declare GBP or GBp so pence can be normalized before GBP accounting. Tushare token is optional — mootdx is the preferred no-token A-share fallback, with AKShare as a broader backup.

@@ -803,7 +803,7 @@ vibe-trading-mcp               # start MCP server (stdio)
 - **Docker** للمسار A
 - يمكن استخدام OpenAI Codex أيضاً عبر ChatGPT OAuth: اضبط `LANGCHAIN_PROVIDER=openai-codex`، ثم شغل `vibe-trading provider login openai-codex`. هذا لا يستخدم `OPENAI_API_KEY`.

-> **مزودو LLM المدعومون:** OpenRouter, Requesty, OpenAI, Anthropic (Messages API الأصلي), DeepSeek, Gemini, Groq, DashScope/Qwen, Zhipu, Moonshot/Kimi, MiniMax, SiliconFlow (CN + Global), Xiaomi MIMO, Novita AI, iFlytek Spark, Z.ai, NVIDIA NIM, ModelScope, GitHub Copilot, Ollama (local). عند عدم ضبط أي `*_BASE_URL`، يتراجع كل مزوّد إلى نقطة نهايته القانونية، فيكفي مفتاح واحد. راجع `.env.example` للإعداد.
+> **مزودو LLM المدعومون:** OpenRouter, OpenAI, Anthropic (Messages API الأصلي), DeepSeek, Gemini, Groq, DashScope/Qwen, Zhipu, Moonshot/Kimi, MiniMax, SiliconFlow (CN + Global), Xiaomi MIMO, Novita AI, iFlytek Spark, Z.ai, NVIDIA NIM, ModelScope, GitHub Copilot, Ollama (local). عند عدم ضبط أي `*_BASE_URL`، يتراجع كل مزوّد إلى نقطة نهايته القانونية، فيكفي مفتاح واحد. راجع `.env.example` للإعداد.

 > **نصيحة:** تعمل كل الأسواق دون مفاتيح API بفضل fallback التلقائي. yfinance/Yahoo (HK/US/كندا/المملكة المتحدة)، وOKX (crypto)، وmootdx (أسهم A، اتصال TCP مباشر بدون قيود IP)، وAKShare (A-shares, US, HK, futures, forex) كلها مجانية. يجب أن تعلن أسعار LSE `.L` عملة GBP أو GBp لتطبيع البنس قبل المحاسبة بالجنيه. رمز Tushare اختياري — mootdx هو الـ fallback الموصى به لأسهم A بدون رمز، بينما يوفر AKShare احتياطياً أوسع تغطية.

@@ -823,7 +823,7 @@ vibe-trading-mcp               # inicia el servidor MCP (stdio)
 - **Docker** para el Path A
 - OpenAI Codex también se puede usar con ChatGPT OAuth: configura `LANGCHAIN_PROVIDER=openai-codex` y luego ejecuta `vibe-trading provider login openai-codex`. Esto no usa `OPENAI_API_KEY`.

-> **Proveedores de LLM compatibles:** OpenRouter, Requesty, OpenAI, Anthropic (API de Messages nativa), DeepSeek, Gemini, Groq, DashScope/Qwen, Zhipu, Moonshot/Kimi, MiniMax, SiliconFlow (CN + Global), Xiaomi MIMO, Novita AI, iFlytek Spark, Z.ai, NVIDIA NIM, ModelScope, GitHub Copilot, Ollama (local). Cuando no se configura ningún `*_BASE_URL`, cada proveedor recurre a su endpoint canónico, así que basta con una clave. Consulta `.env.example` para la configuración.
+> **Proveedores de LLM compatibles:** OpenRouter, OpenAI, Anthropic (API de Messages nativa), DeepSeek, Gemini, Groq, DashScope/Qwen, Zhipu, Moonshot/Kimi, MiniMax, SiliconFlow (CN + Global), Xiaomi MIMO, Novita AI, iFlytek Spark, Z.ai, NVIDIA NIM, ModelScope, GitHub Copilot, Ollama (local). Cuando no se configura ningún `*_BASE_URL`, cada proveedor recurre a su endpoint canónico, así que basta con una clave. Consulta `.env.example` para la configuración.

 > **Consejo:** Todos los mercados funcionan sin ninguna clave de API gracias al fallback automático. yfinance/Yahoo (HK/US/Canadá/Reino Unido), OKX (cripto), mootdx (acciones A, conexión TCP directa, sin limitación de IP) y AKShare (acciones A, EE. UU., HK, futuros, forex) son gratuitos. Las cotizaciones LSE `.L` deben declarar GBP o GBp para normalizar los peniques antes de la contabilidad en GBP. El token de Tushare es opcional — mootdx es el fallback preferido sin token para acciones A, con AKShare como respaldo más amplio.

@@ -977,7 +977,7 @@ vibe-trading-mcp               # start MCP server (stdio)
 - OpenAI Codex juga dapat digunakan dengan ChatGPT OAuth: atur `LANGCHAIN_PROVIDER=openai-codex`, lalu jalankan `vibe-trading provider login openai-codex`. Ini tidak menggunakan `OPENAI_API_KEY`.
 - GitHub Copilot dapat digunakan dengan subscription Copilot aktif tanpa LLM API key yang ditagihkan terpisah. Lihat [provider GitHub Copilot SDK](#github-copilot-sdk-provider).

-> **Provider LLM yang didukung:** OpenRouter, Requesty, OpenAI, Anthropic (native Messages API), DeepSeek, OpenCode (Go / Zen), Gemini, Groq, DashScope/Qwen, Zhipu, Moonshot/Kimi, MiniMax, SiliconFlow (CN + Global), Xiaomi MIMO, Novita AI, iFlytek Spark, Z.ai, NVIDIA NIM, ModelScope, GitHub Copilot, Ollama (lokal). Jika `*_BASE_URL` tidak diatur, setiap provider fallback ke endpoint canonical-nya, jadi cukup key saja. Lihat `.env.example` untuk config.
+> **Provider LLM yang didukung:** OpenRouter, OpenAI, Anthropic (native Messages API), DeepSeek, OpenCode (Go / Zen), Gemini, Groq, DashScope/Qwen, Zhipu, Moonshot/Kimi, MiniMax, SiliconFlow (CN + Global), Xiaomi MIMO, Novita AI, iFlytek Spark, Z.ai, NVIDIA NIM, ModelScope, GitHub Copilot, Ollama (lokal). Jika `*_BASE_URL` tidak diatur, setiap provider fallback ke endpoint canonical-nya, jadi cukup key saja. Lihat `.env.example` untuk config.

 > **Tip:** Semua market dapat bekerja tanpa API key berkat fallback otomatis. yfinance/Yahoo (HK/AS/Kanada/UK), OKX (crypto), mootdx (A-share, TCP langsung, tanpa throttle IP), dan AKShare (A-share, AS, HK, futures, forex) semuanya gratis. Quote LSE `.L` harus mendeklarasikan GBP atau GBp agar pence dapat dinormalisasi sebelum accounting GBP. Token Tushare opsional — mootdx adalah fallback A-share tanpa token yang disarankan, dengan AKShare sebagai backup yang lebih luas.

@@ -808,7 +808,7 @@ vibe-trading-mcp               # start MCP server (stdio)
 - Path A では **Docker**
 - OpenAI Codex は ChatGPT OAuth でも利用できます。`LANGCHAIN_PROVIDER=openai-codex` を設定し、`vibe-trading provider login openai-codex` を実行してください。`OPENAI_API_KEY` は使いません。

-> **Supported LLM providers:** OpenRouter、Requesty、OpenAI、Anthropic（ネイティブ Messages API）、DeepSeek、Gemini、Groq、DashScope/Qwen、Zhipu、Moonshot/Kimi、MiniMax、SiliconFlow（CN + Global）、Xiaomi MIMO、Novita AI、iFlytek Spark、Z.ai、NVIDIA NIM、ModelScope、GitHub Copilot、Ollama（local）。`*_BASE_URL` が未設定の場合、各プロバイダーは canonical なエンドポイントにフォールバックするため、key だけで十分です。設定は `.env.example` を参照してください。
+> **Supported LLM providers:** OpenRouter、OpenAI、Anthropic（ネイティブ Messages API）、DeepSeek、Gemini、Groq、DashScope/Qwen、Zhipu、Moonshot/Kimi、MiniMax、SiliconFlow（CN + Global）、Xiaomi MIMO、Novita AI、iFlytek Spark、Z.ai、NVIDIA NIM、ModelScope、GitHub Copilot、Ollama（local）。`*_BASE_URL` が未設定の場合、各プロバイダーは canonical なエンドポイントにフォールバックするため、key だけで十分です。設定は `.env.example` を参照してください。

 > **Tip:** 自動フォールバックにより、すべての市場は API key なしで利用できます。yfinance/Yahoo（HK/US/カナダ/英国）、OKX（crypto）、mootdx（A 株、TCP 直結で IP 制限なし）、AKShare（A-shares、US、HK、futures、forex）はすべて無料です。LSE `.L` クォートは GBP または GBp の明示が必要で、GBP 会計の前にペンスを正規化します。Tushare token は任意で、A 株は mootdx が推奨の no-token fallback、AKShare がより広いカバレッジのバックアップになります。

@@ -806,7 +806,7 @@ vibe-trading-mcp               # start MCP server (stdio)
 - 경로 A용 **Docker**
 - OpenAI Codex도 ChatGPT OAuth로 사용할 수 있습니다. `LANGCHAIN_PROVIDER=openai-codex`를 설정한 뒤 `vibe-trading provider login openai-codex`를 실행하세요. 이 방식은 `OPENAI_API_KEY`를 사용하지 않습니다.

-> **지원 LLM provider:** OpenRouter, Requesty, OpenAI, Anthropic (native Messages API), DeepSeek, Gemini, Groq, DashScope/Qwen, Zhipu, Moonshot/Kimi, MiniMax, SiliconFlow (CN + Global), Xiaomi MIMO, Novita AI, iFlytek Spark, Z.ai, NVIDIA NIM, ModelScope, GitHub Copilot, Ollama(local). `*_BASE_URL`이 설정되지 않으면 각 provider는 canonical endpoint로 폴백하므로 key만 있으면 충분합니다. 설정은 `.env.example`을 참고하세요.
+> **지원 LLM provider:** OpenRouter, OpenAI, Anthropic (native Messages API), DeepSeek, Gemini, Groq, DashScope/Qwen, Zhipu, Moonshot/Kimi, MiniMax, SiliconFlow (CN + Global), Xiaomi MIMO, Novita AI, iFlytek Spark, Z.ai, NVIDIA NIM, ModelScope, GitHub Copilot, Ollama(local). `*_BASE_URL`이 설정되지 않으면 각 provider는 canonical endpoint로 폴백하므로 key만 있으면 충분합니다. 설정은 `.env.example`을 참고하세요.

 > **팁:** 자동 fallback 덕분에 모든 시장은 API key 없이도 작동합니다. yfinance/Yahoo(HK/US/캐나다/영국), OKX(crypto), mootdx(A주, TCP 직결, IP 제한 없음), AKShare(A주, US, HK, futures, forex)는 모두 무료입니다. LSE `.L` 호가는 GBP 또는 GBp로 명시되어야 하며, GBP 회계 전에 펜스를 정규화합니다. Tushare token은 선택 사항이며, mootdx가 권장 no-token A주 fallback이고 AKShare는 더 넓은 커버리지의 백업입니다.

@@ -813,7 +813,7 @@ vibe-trading-mcp               # start MCP server (stdio)
 - 路径 A 需要 **Docker**
 - OpenAI Codex 也可通过 ChatGPT OAuth 使用：设置 `LANGCHAIN_PROVIDER=openai-codex`，然后运行 `vibe-trading provider login openai-codex`。它不使用 `OPENAI_API_KEY`。

-> **支持的 LLM providers：** OpenRouter、Requesty、OpenAI、Anthropic（原生 Messages API）、DeepSeek、Gemini、Groq、DashScope/Qwen、Zhipu、Moonshot/Kimi、MiniMax、SiliconFlow（CN + Global）、Xiaomi MIMO、Novita AI、iFlytek 星火、Z.ai、NVIDIA NIM、ModelScope、GitHub Copilot、Ollama（本地）。未设置 `*_BASE_URL` 时，每个 provider 会回退到其规范端点，因此只需一个 key 即可。配置见 `.env.example`。
+> **支持的 LLM providers：** OpenRouter、OpenAI、Anthropic（原生 Messages API）、DeepSeek、Gemini、Groq、DashScope/Qwen、Zhipu、Moonshot/Kimi、MiniMax、SiliconFlow（CN + Global）、Xiaomi MIMO、Novita AI、iFlytek 星火、Z.ai、NVIDIA NIM、ModelScope、GitHub Copilot、Ollama（本地）。未设置 `*_BASE_URL` 时，每个 provider 会回退到其规范端点，因此只需一个 key 即可。配置见 `.env.example`。

 > **提示：** 由于自动 fallback，所有市场都可以在没有任何 API key 的情况下工作。yfinance/Yahoo（港股/美股/加拿大/英国）、OKX（加密）、mootdx（A 股，TCP 直连不封 IP）和 AKShare（A 股、美股、港股、期货、外汇）都是免费的。LSE `.L` 报价必须明示为 GBP 或 GBp，以便在英镑记账前将便士归一化。Tushare token 是可选项 —— mootdx 是首选的免 token A 股 fallback，AKShare 作为覆盖更广的兜底。

@@ -9,14 +9,6 @@ LANGCHAIN_MODEL_NAME=deepseek/deepseek-v4-pro
 OPENROUTER_API_KEY=sk-or-...here
 OPENROUTER_BASE_URL=https://openrouter.ai/api/v1

-# --- Requesty (OpenAI-compatible multi-model gateway) ---
-# Same provider/model naming as OpenRouter. Get a key at https://app.requesty.ai/api-keys
-# EU endpoint: https://router.eu.requesty.ai/v1
-# LANGCHAIN_PROVIDER=requesty
-# LANGCHAIN_MODEL_NAME=openai/gpt-4o-mini
-# REQUESTY_API_KEY=xxx
-# REQUESTY_BASE_URL=https://router.requesty.ai/v1
-
 # --- OpenAI ---
 # LANGCHAIN_PROVIDER=openai
 # LANGCHAIN_MODEL_NAME=gpt-5.5-instant
@@ -435,7 +435,6 @@ def _provider_key_env(provider: str | None) -> str | None:
    """Return the credential environment variable for a provider."""
    return {
        "openrouter": "OPENROUTER_API_KEY",
-        "requesty": "REQUESTY_API_KEY",
        "openai": "OPENAI_API_KEY",
        "deepseek": "DEEPSEEK_API_KEY",
        "nvidia": "NVIDIA_API_KEY",
@@ -460,7 +459,6 @@ def _provider_base_env(provider: str | None) -> str | None:
    """Return the base URL environment variable for a provider."""
    return {
        "openrouter": "OPENROUTER_BASE_URL",
-        "requesty": "REQUESTY_BASE_URL",
        "openai": "OPENAI_BASE_URL",
        "openai-codex": "OPENAI_CODEX_BASE_URL",
        "deepseek": "DEEPSEEK_BASE_URL",
@@ -5715,16 +5713,6 @@ _PROVIDER_CHOICES: list[dict[str, str | None]] = [
        "key_prefix": "sk-or-",
        "key_placeholder": "sk-or-v1-...",
    },
-    {
-        "label": "Requesty (OpenAI-compatible gateway - multiple models)",
-        "provider": "requesty",
-        "key_env": "REQUESTY_API_KEY",
-        "base_env": "REQUESTY_BASE_URL",
-        "base_url": "https://router.requesty.ai/v1",
-        "model": "openai/gpt-4o-mini",
-        "key_prefix": None,
-        "key_placeholder": "api-key...",
-    },
    {
        "label": "DeepSeek",
        "provider": "deepseek",
@@ -5932,8 +5920,6 @@ def _render_env_content(config: dict[str, str]) -> str:
        "LANGCHAIN_PROVIDER",
        "OPENROUTER_API_KEY",
        "OPENROUTER_BASE_URL",
-        "REQUESTY_API_KEY",
-        "REQUESTY_BASE_URL",
        "DEEPSEEK_API_KEY",
        "DEEPSEEK_BASE_URL",
        "NVIDIA_API_KEY",
@@ -57,12 +57,6 @@ PROVIDERS: Final[tuple[Provider, ...]] = (
             "https://openrouter.ai/api/v1", "sk-or-",
             ("deepseek/deepseek-v4-pro", "deepseek/deepseek-v4-flash",
              "openai/gpt-5.5-pro", "google/gemini-3.5-flash")),
-    Provider("requesty", "Requesty", "OpenAI-compatible gateway — 600+ models, one key",
-             "openai/gpt-4o-mini",
-             "REQUESTY_API_KEY", "REQUESTY_BASE_URL",
-             "https://router.requesty.ai/v1", None,
-             ("openai/gpt-4o-mini", "openai/gpt-4o",
-              "anthropic/claude-sonnet-4-5", "deepseek/deepseek-chat")),
    Provider("openai", "OpenAI", "GPT-5.5 direct",
             "gpt-5.5", "OPENAI_API_KEY", "OPENAI_BASE_URL",
             "https://api.openai.com/v1", "sk-",
@@ -183,16 +183,6 @@ _PROVIDERS: dict[str, ProviderCapabilities] = {
        capture_reasoning=True,
        openrouter_reasoning_body=True,
    ),
-    # Requesty is an OpenAI-compatible LLM gateway using the same
-    # ``provider/model`` naming and the same opt-in ``extra_body.reasoning``
-    # request option as OpenRouter, so it shares OpenRouter's capability shape.
-    "requesty": ProviderCapabilities(
-        "requesty",
-        "REQUESTY_API_KEY",
-        "REQUESTY_BASE_URL",
-        capture_reasoning=True,
-        openrouter_reasoning_body=True,
-    ),
    # Verified live against DeepSeek Flash in #1025: a chat-completions request
    # carrying top-level ``reasoning_effort`` came back with reasoning content
    # and non-zero reasoning usage.
@@ -294,7 +284,7 @@ def get_provider_capabilities(
    Notes:
        Model-name inference (``_infer_from_model``) activates for the default
        ``"openai"`` provider and empty/None providers. Explicit non-OpenAI
-        providers (OpenRouter, Requesty, DeepSeek, etc.) are never inferred —
+        providers (OpenRouter, DeepSeek, etc.) are never inferred —
        the explicit provider choice always wins.
    """
    normalized = (provider or "").strip().lower().replace("_", "-")
@@ -1420,8 +1420,8 @@ def _supports_top_level_reasoning_effort(caps: ProviderCapabilities) -> bool:
    strictly rejects the unknown key outright — so the default is off and the
    consequence of that default is a no-op, not a failed call.

-    Relays that take the effort inside ``extra_body.reasoning`` (OpenRouter,
-    Requesty) use that path instead and are excluded here.
+    Relays that take the effort inside ``extra_body.reasoning`` (OpenRouter)
+    use that path instead and are excluded here.

    Args:
        caps: Canonical capabilities resolved for the provider and model. Note
@@ -8,15 +8,6 @@
    "default_base_url": "https://openrouter.ai/api/v1",
    "api_key_required": true
  },
-  {
-    "name": "requesty",
-    "label": "Requesty",
-    "api_key_env": "REQUESTY_API_KEY",
-    "base_url_env": "REQUESTY_BASE_URL",
-    "default_model": "openai/gpt-4o-mini",
-    "default_base_url": "https://router.requesty.ai/v1",
-    "api_key_required": true
-  },
  {
    "name": "openai",
    "label": "OpenAI",
@@ -43,7 +43,6 @@ _PUBLIC_PROVIDERS = frozenset(
        "opencode-zen",
        "openrouter",
        "qwen",
-        "requesty",
        "siliconflow-cn",
        "siliconflow-global",
        "spark",
@@ -94,7 +94,6 @@ class TestProviderCapabilityAliases:
        [
            # Gateway providers — explicit choice must never be overridden.
            ("openrouter", "deepseek/deepseek-v4-pro", "openrouter"),
-            ("requesty", "deepseek/deepseek-v4-pro", "requesty"),
            ("openrouter", "gemini-3.5-flash", "openrouter"),
            ("openrouter", "glm-4.6", "openrouter"),
        ],
@@ -102,7 +101,7 @@ class TestProviderCapabilityAliases:
    def test_gateway_provider_not_inferred_from_model(
        self, provider: str, model: str, expected: str
    ) -> None:
-        """Gateway providers (OpenRouter/Requesty) must never be overridden. (#549)
+        """Gateway providers (OpenRouter) must never be overridden. (#549)

        Their model names contain direct-provider prefixes like ``deepseek/``
        that would trigger inference, but the explicit gateway choice must win.
@@ -15,7 +15,6 @@ from src.providers.capabilities import (

 EXPECTED_PROVIDER_DEFAULTS = {
    "openrouter": "deepseek/deepseek-v4-pro",
-    "requesty": "openai/gpt-4o-mini",
    "openai": "gpt-5.5",
    "anthropic": "claude-sonnet-4-6",
    "openai-codex": "openai-codex/gpt-5.4",
@@ -2,7 +2,7 @@

 The setting has two mutually exclusive delivery paths:

-* Relays that require an opt-in (OpenRouter, Requesty) receive
+* Relays that require an opt-in (OpenRouter) receive
  ``extra_body={"reasoning": {"effort": ...}}``.
 * ChatOpenAI-compatible providers receive a top-level ``reasoning_effort``. Its ``gpt-5.6-*``
  models reject function tools on ``/v1/chat/completions`` without one::
@@ -478,7 +478,7 @@ class TestAnthropicNativeProvider:


 class TestRelayOptIn:
-    """OpenRouter/Requesty keep the pre-existing extra_body opt-in."""
+    """OpenRouter keeps the pre-existing extra_body opt-in."""

    def test_openrouter_keeps_extra_body_and_no_top_level_field(self) -> None:
        kwargs = _capture_kwargs(
@@ -495,21 +495,6 @@ class TestRelayOptIn:
        assert kwargs["extra_body"] == {"reasoning": {"effort": "high"}}
        assert kwargs["reasoning_effort"] is None

-    def test_requesty_keeps_extra_body_and_no_top_level_field(self) -> None:
-        kwargs = _capture_kwargs(
-            {
-                "LANGCHAIN_PROVIDER": "requesty",
-                "REQUESTY_API_KEY": "rq-test",
-                "REQUESTY_BASE_URL": "https://router.requesty.ai/v1",
-                "LANGCHAIN_MODEL_NAME": "openai/gpt-4o-mini",
-                "LANGCHAIN_REASONING_EFFORT": "medium",
-                "LANGCHAIN_USE_RESPONSES_API": "false",
-            }
-        )
-
-        assert kwargs["extra_body"] == {"reasoning": {"effort": "medium"}}
-        assert kwargs["reasoning_effort"] is None
-

 class TestRequestPayload:
    """The kwarg has to survive as a top-level chat-completions request field.
@@ -66,18 +66,6 @@ def test_provider_capabilities_are_provider_specific() -> None:
    assert openrouter.send_reasoning_content is False


-def test_requesty_capabilities_mirror_openrouter() -> None:
-    """Requesty is an OpenAI-compatible gateway wired like OpenRouter."""
-    requesty = get_provider_capabilities("requesty", "openai/gpt-4o-mini")
-
-    assert requesty.name == "requesty"
-    assert requesty.api_key_env == "REQUESTY_API_KEY"
-    assert requesty.base_url_env == "REQUESTY_BASE_URL"
-    assert requesty.capture_reasoning is True
-    assert requesty.openrouter_reasoning_body is True
-    assert requesty.send_reasoning_content is False
-
-
 def test_spark_capabilities_use_generic_openai_path() -> None:
    """iFlytek Spark rides the plain OpenAI-compatible path; iflytek is an alias."""
    spark = get_provider_capabilities("spark", "4.0Ultra")
@@ -11,7 +11,6 @@ type CredentialFile = {

 const ENV_CREDENTIALS = new Set([
  "OPENROUTER_API_KEY",
-  "REQUESTY_API_KEY",
  "OPENAI_API_KEY",
  "ANTHROPIC_API_KEY",
  "DEEPSEEK_API_KEY",