feat: add OrcaRouter as a built-in model provider

.env.template

@@ -82,6 +82,15 @@ OLLAMA_ENABLED=true
 OLLAMA_API_BASE=http://localhost:11434
 OLLAMA_MODELS=qwen3:32b # models with good code generation capabilities recommended
 
+# OrcaRouter (OpenAI-compatible AI gateway)
+# Provides adaptive routing, automatic failover, zero-markup inference,
+# observability, guardrails, and agent-tool governance behind one endpoint.
+# See: https://www.orcarouter.ai
+ORCAROUTER_ENABLED=true
+ORCAROUTER_API_KEY=#your-orcarouter-api-key
+ORCAROUTER_API_BASE=https://api.orcarouter.ai/v1
+ORCAROUTER_MODELS=auto # comma separated list of models; use e.g. "auto" or "openai/gpt-4.1-mini"
+
 # Add other LiteLLM-supported providers with PROVIDER_API_KEY, PROVIDER_MODELS, etc.
 
 # -------------------------------------------------------------------

README.md

@@ -76,7 +76,7 @@ Here are milestones that lead to the current design:
 - **v0.2** ([Demos](https://github.com/microsoft/data-formulator/releases/tag/0.2)): Large data support with DuckDB integration
 - **v0.1.7** ([Demos](https://github.com/microsoft/data-formulator/releases/tag/0.1.7)): Dataset anchoring for cleaner workflows
 - **v0.1.6** ([Demo](https://github.com/microsoft/data-formulator/releases/tag/0.1.6)): Multi-table support with automatic joins
-- **Model Support**: OpenAI, Azure, Ollama, Anthropic via [LiteLLM](https://github.com/BerriAI/litellm) ([feedback](https://github.com/microsoft/data-formulator/issues/49))
+- **Model Support**: OpenAI, Azure, Ollama, Anthropic, [OrcaRouter](https://www.orcarouter.ai) via [LiteLLM](https://github.com/BerriAI/litellm) ([feedback](https://github.com/microsoft/data-formulator/issues/49))
 - **Python Package**: Easy local installation ([try it](#get-started))
 - **Visualization Challenges**: Test your skills ([challenges](https://github.com/microsoft/data-formulator/issues/53))
 - **Data Extraction**: Parse data from images and text ([demo](https://github.com/microsoft/data-formulator/pull/31#issuecomment-2403652717))

py-src/data_formulator/agents/client_utils.py

@@ -219,7 +219,7 @@ def _salvage_tool_calls_from_content(response, tools):
 class Client(object):
     """
     Returns a LiteLLM client configured for the specified endpoint and model.
-    Supports OpenAI, Azure, Ollama, and other providers via LiteLLM.
+    Supports OpenAI, Azure, Ollama, OrcaRouter, and other providers via LiteLLM.
     """
     def __init__(self, endpoint, model, api_key=None,  api_base=None, api_version=None):
 
@@ -274,6 +274,16 @@ def __init__(self, endpoint, model, api_key=None,  api_base=None, api_version=No
                 self.model = model
             else:
                 self.model = f"ollama/{model}"
+        elif self.endpoint == "orcarouter":
+            # OrcaRouter exposes an OpenAI-compatible API, so route the model
+            # through LiteLLM's openai provider against the OrcaRouter base URL.
+            # The ``orcarouter/`` prefix is preserved by LiteLLM (unlike
+            # ``openai/``, which it strips), which is how OrcaRouter's gateway
+            # addresses its model routers.
+            self.params["api_base"] = (api_base or "https://api.orcarouter.ai/v1").rstrip("/")
+            self.params["custom_llm_provider"] = "openai"
+            if "/" not in model:
+                self.model = f"orcarouter/{model}"
 
     def _strip_image_blocks(self, content):
         """Remove image_url blocks from multimodal content arrays."""

py-src/data_formulator/model_registry.py

@@ -4,15 +4,15 @@
 import os
 from typing import Optional, Dict, List
 
-BUILTIN_PROVIDERS = {'openai', 'azure', 'anthropic', 'gemini', 'ollama'}
+BUILTIN_PROVIDERS = {'openai', 'azure', 'anthropic', 'gemini', 'ollama', 'orcarouter'}
 
 
 class ModelRegistry:
     """
     Load global model configurations from environment variables.
 
     Supports both built-in providers (openai / azure / anthropic / gemini /
-    ollama) and arbitrary custom providers (e.g. DEEPSEEK, QWEN).
+    ollama / orcarouter) and arbitrary custom providers (e.g. DEEPSEEK, QWEN).
 
     For a custom provider, set:
         {PROVIDER}_ENABLED=true

src/views/ModelSelectionDialog.tsx

@@ -109,7 +109,8 @@ export const ModelSelectionButton: React.FC<ModelSelectionButtonProps> = ({ appe
         'azure': [],
         'anthropic': [],
         'gemini': [],
-        'ollama': []
+        'ollama': [],
+        'orcarouter': []
     });
     const serverConfig = useSelector((state: DataFormulatorState) => state.serverConfig);
 
@@ -223,7 +224,8 @@ export const ModelSelectionButton: React.FC<ModelSelectionButtonProps> = ({ appe
             'azure': [],
             'anthropic': [],
             'gemini': [],
-            'ollama': []
+            'ollama': [],
+            'orcarouter': []
         };
 
         globalModels.forEach((modelConfig: any) => {
@@ -436,7 +438,7 @@ export const ModelSelectionButton: React.FC<ModelSelectionButtonProps> = ({ appe
                     setNewModelDiagnostic(null);
                 }}
             >
-                {['openai', 'azure', 'ollama', 'anthropic', 'gemini'].map(provider => (
+                {['openai', 'azure', 'ollama', 'anthropic', 'gemini', 'orcarouter'].map(provider => (
                     <MenuItem key={provider} value={provider}>{provider}</MenuItem>
                 ))}
             </TextField>

tests/backend/agents/test_client_utils.py

@@ -56,6 +56,35 @@ def test_openai_model_prefixed(self):
         assert c.model == "openai/gpt-4o"
 
 
+# ---------------------------------------------------------------------------
+# OrcaRouter endpoint
+# ---------------------------------------------------------------------------
+
+class TestOrcaRouter:
+    def test_default_base_url(self):
+        c = Client("orcarouter", "auto", api_key="k")
+        assert c.params["api_base"] == "https://api.orcarouter.ai/v1"
+
+    def test_custom_base_url_strips_trailing_slash(self):
+        c = Client("orcarouter", "auto", api_key="k",
+                   api_base="https://api.orcarouter.ai/v1/")
+        assert c.params["api_base"] == "https://api.orcarouter.ai/v1"
+
+    def test_uses_openai_compatible_provider(self):
+        c = Client("orcarouter", "auto", api_key="k")
+        assert c.params["custom_llm_provider"] == "openai"
+
+    def test_model_prefixed_with_orcarouter_namespace(self):
+        """The ``orcarouter/`` prefix is preserved by LiteLLM (unlike
+        ``openai/``, which it strips) so OrcaRouter's gateway can route it."""
+        c = Client("orcarouter", "auto", api_key="k")
+        assert c.model == "orcarouter/auto"
+
+    def test_model_prefix_not_doubled(self):
+        c = Client("orcarouter", "orcarouter/auto", api_key="k")
+        assert c.model == "orcarouter/auto"
+
+
 # ---------------------------------------------------------------------------
 # Ollama api_base normalisation
 # ---------------------------------------------------------------------------

tests/backend/agents/test_model_registry.py

@@ -50,6 +50,12 @@ def _make_env(providers: dict[str, dict[str, str]]) -> dict[str, str]:
         "api_base": "https://api.deepseek.com/v1",
         "models": "deepseek-chat",
     },
+    "orcarouter": {
+        "enabled": "true",
+        "api_key": "sk-orca-secret-key",
+        "api_base": "https://api.orcarouter.ai/v1",
+        "models": "auto",
+    },
 })
 
 
@@ -68,11 +74,12 @@ def test_discovers_all_enabled_providers(self):
         assert "global-openai-gpt-5" in ids
         assert "global-ollama-qwen3:32b" in ids
         assert "global-deepseek-deepseek-chat" in ids
+        assert "global-orcarouter-auto" in ids
 
     @patch.dict(os.environ, SAMPLE_ENV, clear=True)
     def test_total_model_count(self):
         registry = ModelRegistry()
-        assert len(registry.list_public()) == 4  # 2 openai + 1 ollama + 1 deepseek
+        assert len(registry.list_public()) == 5  # 2 openai + 1 ollama + 1 deepseek + 1 orcarouter
 
     @patch.dict(os.environ, {}, clear=True)
     def test_empty_env_yields_no_models(self):
@@ -141,6 +148,15 @@ def test_builtin_provider_uses_own_name_as_endpoint(self):
         assert config is not None
         assert config["endpoint"] == "openai"
 
+    @patch.dict(os.environ, SAMPLE_ENV, clear=True)
+    def test_orcarouter_builtin_uses_own_name_as_endpoint(self):
+        """orcarouter is in BUILTIN_PROVIDERS, so it uses itself as endpoint."""
+        registry = ModelRegistry()
+        config = registry.get_config("global-orcarouter-auto")
+        assert config is not None
+        assert config["endpoint"] == "orcarouter"
+        assert config["api_base"] == "https://api.orcarouter.ai/v1"
+
     @patch.dict(os.environ, {
         "MYVENDOR_ENABLED": "true",
         "MYVENDOR_API_KEY": "key123",