feat: add Requesty as a config-driven provider

.env.example

@@ -41,6 +41,11 @@ AGENT_DEBUG_MODE=false
 # Note: OpenRouter does not currently support embedding or reranker models.
 OPENROUTER_API_KEY=
 
+# Get your API key from: https://app.requesty.ai/api-keys
+# Requesty is an OpenAI-compatible LLM router; uses the same provider/model id
+# convention as OpenRouter (e.g. openai/gpt-4o-mini).
+REQUESTY_API_KEY=
+
 # Get your API key from: https://ai.azure.com/
 AZURE_OPENAI_API_KEY=
 AZURE_OPENAI_ENDPOINT=

docs/CONFIGURATION_GUIDE.md

@@ -24,6 +24,7 @@ ValueCell supports multiple LLM providers. Choose at least one:
 | Provider        | Sign Up                                             |
 | --------------- | --------------------------------------------------- |
 | **OpenRouter**  | [openrouter.ai](https://openrouter.ai/)             |
+| **Requesty**    | [app.requesty.ai/api-keys](https://app.requesty.ai/api-keys) |
 | **SiliconFlow** | [siliconflow.cn](https://www.siliconflow.cn/)       |
 | **Google**      | [ai.google.dev](https://ai.google.dev/)             |
 | **OpenAI**      | [platform.openai.com](https://platform.openai.com/) |
@@ -86,6 +87,7 @@ python/
 │   ├── config.{environment}.yaml      # Environment-specific overrides
 │   ├── providers/
 │   │   ├── openrouter.yaml           # OpenRouter provider config
+│   │   ├── requesty.yaml             # Requesty provider config
 │   │   ├── siliconflow.yaml          # SiliconFlow provider config
 │   │   ├── dashscope.yaml            # DashScope (Alibaba Cloud) provider config
 │   │   └── other_provider.yaml
@@ -191,6 +193,9 @@ models:
     openrouter:
       config_file: "providers/openrouter.yaml"
       api_key_env: "OPENROUTER_API_KEY"
+    requesty:
+      config_file: "providers/requesty.yaml"
+      api_key_env: "REQUESTY_API_KEY"
     siliconflow:
       config_file: "providers/siliconflow.yaml"
       api_key_env: "SILICONFLOW_API_KEY"
@@ -327,12 +332,13 @@ ValueCell automatically selects a primary provider based on available API keys:
 The selection logic is implemented in `python/valuecell/config/manager.py`:
 
 1. OpenRouter
-2. SiliconFlow
-3. Google
-4. OpenAI
-5. OpenAI-Compatible
-6. Azure
-7. Other configured providers (including DashScope, DeepSeek, etc.)
+2. Requesty
+3. SiliconFlow
+4. Google
+5. OpenAI
+6. OpenAI-Compatible
+7. Azure
+8. Other configured providers (including DashScope, DeepSeek, etc.)
 
 Override this with an environment variable:
 

python/configs/config.yaml

@@ -23,6 +23,10 @@ models:
       config_file: "providers/openrouter.yaml"
       api_key_env: "OPENROUTER_API_KEY"
 
+    requesty:
+      config_file: "providers/requesty.yaml"
+      api_key_env: "REQUESTY_API_KEY"
+
     siliconflow:
       config_file: "providers/siliconflow.yaml"
       api_key_env: "SILICONFLOW_API_KEY"

python/configs/providers/requesty.yaml

@@ -0,0 +1,58 @@
+# ============================================
+# Requesty Provider Configuration
+# ============================================
+name: "Requesty"
+provider_type: "requesty"
+
+enabled: true # Default is true if not specified
+
+# Connection Configuration
+connection:
+  base_url: "https://router.requesty.ai/v1"
+  api_key_env: "REQUESTY_API_KEY"
+
+# Default model if none specified
+default_model: "openai/gpt-4o-mini"
+
+# Model Parameters Defaults
+defaults:
+  temperature: 0.5
+
+# Extra headers for Requesty API (used for analytics attribution)
+extra_headers:
+  HTTP-Referer: "https://valuecell.ai"
+  X-Title: "ValueCell"
+
+# Available Models (commonly used)
+models:
+  - id: "anthropic/claude-sonnet-4-5"
+    name: "Claude Sonnet 4.5"
+  - id: "openai/gpt-4o"
+    name: "GPT-4o"
+  - id: "openai/gpt-4o-mini"
+    name: "GPT-4o mini"
+  - id: "deepseek/deepseek-chat"
+    name: "DeepSeek Chat"
+  - id: "google/gemini-2.5-flash"
+    name: "Gemini 2.5 Flash"
+
+# ============================================
+# Embedding Models Configuration
+# ============================================
+# Requesty exposes OpenAI-compatible embedding models
+embedding:
+  # Default embedding model
+  default_model: "openai/text-embedding-3-small"
+
+  # Default parameters
+  defaults:
+    dimensions: 1536
+    encoding_format: "float"
+
+  # Available embedding models
+  models:
+  - id: "openai/text-embedding-3-small"
+    name: "OpenAI Text Embedding 3 Small"
+    dimensions: 1536
+    max_input: 8192
+    description: "OpenAI Text Embedding Model"

python/valuecell/adapters/models/factory.py

@@ -116,6 +116,82 @@ def create_model(self, model_id: Optional[str] = None, **kwargs):
         )
 
 
+class RequestyProvider(ModelProvider):
+    """Requesty model provider
+
+    Requesty (https://requesty.ai) is an OpenAI-compatible LLM router. It uses the
+    same ``provider/model`` id convention as OpenRouter (e.g. ``openai/gpt-4o-mini``)
+    and the same ``HTTP-Referer`` / ``X-Title`` analytics headers, with a fixed
+    OpenAI-compatible base URL (``https://router.requesty.ai/v1``). This mirrors
+    OpenRouterProvider, using agno's OpenAILike client since the endpoint is
+    OpenAI-compatible.
+    """
+
+    def create_model(self, model_id: Optional[str] = None, **kwargs):
+        """Create Requesty model via agno (OpenAI-compatible)"""
+        try:
+            from agno.models.openai import OpenAILike
+        except ImportError:
+            raise ImportError(
+                "agno package not installed. Install with: pip install agno"
+            )
+
+        # Use provided model_id or default
+        model_id = model_id or self.config.default_model
+
+        # Merge parameters: provider defaults < kwargs
+        params = {**self.config.parameters, **kwargs}
+
+        # Get extra headers from config (analytics attribution)
+        extra_headers = self.config.extra_config.get("extra_headers", {})
+
+        logger.info(f"Creating Requesty model: {model_id}")
+
+        return OpenAILike(
+            id=model_id,
+            api_key=self.config.api_key,
+            base_url=self.config.base_url,
+            default_headers=extra_headers if extra_headers else None,
+            temperature=params.get("temperature"),
+            max_tokens=params.get("max_tokens"),
+            top_p=params.get("top_p"),
+            frequency_penalty=params.get("frequency_penalty"),
+            presence_penalty=params.get("presence_penalty"),
+        )
+
+    def create_embedder(self, model_id: Optional[str] = None, **kwargs):
+        """Create embedder via Requesty (OpenAI-compatible)"""
+        try:
+            from agno.knowledge.embedder.openai import OpenAIEmbedder
+        except ImportError:
+            raise ImportError("agno package not installed")
+
+        # Use provided model_id or default embedding model
+        model_id = model_id or self.config.default_embedding_model
+
+        if not model_id:
+            raise ValueError(
+                f"No embedding model specified for provider '{self.config.name}'"
+            )
+
+        # Merge parameters: provider embedding defaults < kwargs
+        params = {**self.config.embedding_parameters, **kwargs}
+
+        logger.info(f"Creating Requesty embedder: {model_id}")
+
+        return OpenAIEmbedder(
+            id=model_id,
+            api_key=self.config.api_key,
+            base_url=self.config.base_url,
+            dimensions=int(params.get("dimensions", 1536)),
+            encoding_format=params.get("encoding_format"),
+        )
+
+    def is_available(self) -> bool:
+        """Check if provider is available (needs both API key and base URL)"""
+        return bool(self.config.api_key and self.config.base_url)
+
+
 class GoogleProvider(ModelProvider):
     """Google Gemini model provider"""
 
@@ -601,6 +677,7 @@ class ModelFactory:
     # Registry of provider classes
     _providers: Dict[str, type[ModelProvider]] = {
         "openrouter": OpenRouterProvider,
+        "requesty": RequestyProvider,
         "google": GoogleProvider,
         "azure": AzureProvider,
         "siliconflow": SiliconFlowProvider,

python/valuecell/config/manager.py

@@ -130,6 +130,7 @@ def primary_provider(self) -> str:
                 # Priority order for auto-selection
                 preferred_order = [
                     "openrouter",
+                    "requesty",
                     "siliconflow",
                     "google",
                     "openai",

python/valuecell/config/tests/test_provider_config.py

@@ -0,0 +1,65 @@
+"""Tests for provider configuration loading.
+
+Validates that config-driven providers (openrouter, requesty) load through the
+shared ConfigLoader / ConfigManager path with the expected connection settings.
+"""
+
+import os
+
+import pytest
+
+from valuecell.config.loader import ConfigLoader
+from valuecell.config.manager import ConfigManager
+
+
+@pytest.fixture()
+def manager() -> ConfigManager:
+    loader = ConfigLoader()
+    return ConfigManager(loader=loader)
+
+
+def test_requesty_provider_is_discovered():
+    """requesty.yaml is picked up by the provider glob, like openrouter.yaml."""
+    loader = ConfigLoader()
+    providers = loader.list_providers()
+    assert "openrouter" in providers
+    assert "requesty" in providers
+
+
+def test_requesty_provider_config_matches_expected(manager: ConfigManager):
+    """Requesty loads through the same code path as OpenRouter with its
+    fixed OpenAI-compatible base URL, API-key env var, and analytics headers."""
+    config = manager.get_provider_config("requesty")
+
+    assert config is not None
+    assert config.name == "requesty"
+    assert config.enabled is True
+    assert config.base_url == "https://router.requesty.ai/v1"
+    assert config.default_model == "openai/gpt-4o-mini"
+
+    # provider/model id convention, mirroring OpenRouter
+    model_ids = {m["id"] for m in config.models}
+    assert "openai/gpt-4o-mini" in model_ids
+    assert all("/" in mid for mid in model_ids)
+
+    # Analytics headers mirror OpenRouter's HTTP-Referer / X-Title
+    extra_headers = config.extra_config.get("extra_headers", {})
+    assert extra_headers.get("X-Title") == "ValueCell"
+    assert "HTTP-Referer" in extra_headers
+
+
+def test_requesty_reads_api_key_env(manager: ConfigManager, monkeypatch):
+    """REQUESTY_API_KEY is resolved into the provider config."""
+    monkeypatch.setenv("REQUESTY_API_KEY", "sk-test-requesty")
+    # Fresh manager to avoid loader cache picking up an earlier (empty) value
+    fresh = ConfigManager(loader=ConfigLoader())
+    config = fresh.get_provider_config("requesty")
+    assert config is not None
+    assert config.api_key == "sk-test-requesty"
+
+
+def test_requesty_registered_in_factory():
+    """The requesty provider name maps to a provider class in the factory."""
+    from valuecell.adapters.models.factory import ModelFactory, RequestyProvider
+
+    assert ModelFactory._providers.get("requesty") is RequestyProvider