feat: add OrcaRouter as a named LLM provider (OpenAI-compatible gateway)

AGENTS.md

@@ -57,7 +57,7 @@ Jinja2 conditionals: `{%- if cookiecutter.enable_rag %}...{%- endif %}`
 ## Key Features
 
 - **5 AI Frameworks**: PydanticAI, PydanticDeep, LangChain, LangGraph, DeepAgents
-- **4 LLM Providers**: OpenAI, Anthropic, Google Gemini, OpenRouter
+- **5 LLM Providers**: OpenAI, Anthropic, Google Gemini, OpenRouter, OrcaRouter
 - **RAG**: 4 vector stores (Milvus, Qdrant, ChromaDB, pgvector), 4 embedding providers, reranking, image description
 - **Document Sources**: Local files (CLI), API upload, Google Drive (service account), S3/MinIO
 - **Sync Sources**: Configurable connectors (Google Drive, S3) with scheduled sync

CLAUDE.md

@@ -92,7 +92,7 @@ template/
 ## Key Design Decisions
 
 - 5 AI frameworks: PydanticAI, PydanticDeep, LangChain, LangGraph, DeepAgents
-- 4 LLM providers: OpenAI, Anthropic, Google Gemini, OpenRouter
+- 5 LLM providers: OpenAI, Anthropic, Google Gemini, OpenRouter, OrcaRouter
 - 4 vector store backends: Milvus, Qdrant, ChromaDB, pgvector
 - 4 embedding providers: OpenAI, Voyage, Gemini (multimodal), SentenceTransformers
 - RAG document sources: local files (CLI), Google Drive, S3/MinIO

README.md

@@ -714,7 +714,7 @@ They also ship a ready-to-use **`.claude/` toolkit** that adapts to the options
 ### 🤖 AI/LLM First
 
 - **5 AI Frameworks** - [PydanticAI](https://ai.pydantic.dev), [PydanticDeep](https://github.com/vstorm-co/pydantic-deep), [LangChain](https://python.langchain.com), [LangGraph](https://langchain-ai.github.io/langgraph/), [DeepAgents](https://github.com/vstorm-co/pydantic-deepagents)
-- **4 LLM Providers** - OpenAI, Anthropic, Google Gemini, OpenRouter
+- **5 LLM Providers** - OpenAI, Anthropic, Google Gemini, OpenRouter, OrcaRouter
 - **RAG** - Document ingestion, vector search, reranking (Milvus, Qdrant, ChromaDB, pgvector)
 - **WebSocket Streaming** - Real-time responses with full event access
 - **Rich Chat UI** - Specialized tool-call cards (web search, knowledge base, Python, charts, skills), live subagent feed, citation sources panel, plan/task checklist, reasoning view, and in-chat file previews
@@ -754,7 +754,7 @@ They also ship a ready-to-use **`.claude/` toolkit** that adapts to the options
 | Category | Integrations |
 |----------|-------------|
 | **AI Frameworks** | PydanticAI, PydanticDeep, LangChain, LangGraph, DeepAgents |
-| **LLM Providers** | OpenAI, Anthropic, Google Gemini, OpenRouter |
+| **LLM Providers** | OpenAI, Anthropic, Google Gemini, OpenRouter, OrcaRouter |
 | **RAG / Vector Stores** | Milvus, Qdrant, ChromaDB, pgvector |
 | **RAG Sources** | Local files, API upload, Google Drive, S3/MinIO, Sync Sources (per-org UI, scheduled) |
 | **Embeddings** | OpenAI, Voyage, Gemini (multimodal), SentenceTransformers |
@@ -786,7 +786,7 @@ They also ship a ready-to-use **`.claude/` toolkit** that adapts to the options
 │  │  PydanticAI · LangChain · LangGraph · DeepAgents                │     │
 │  │  ────────────────────────────────────────────────────────────   │     │
 │  │  Tools: datetime · web_search (Tavily) · search_knowledge_base  │     │
-│  │  Providers: OpenAI · Anthropic · Gemini · OpenRouter            │     │
+│  │  Providers: OpenAI · Anthropic · Gemini · OpenRouter · OrcaRouter │     │
 │  └─────────────────────────────────────────────────────────────────┘     │
 │                                                                          │
 │  ┌─────────────────────────────────────────────────────────────────┐     │
@@ -890,7 +890,7 @@ See [Architecture Documentation](https://github.com/vstorm-co/full-stack-ai-agen
 
 ## 🤖 AI Agent
 
-Choose from **5 AI frameworks** and **4 LLM providers** when generating your project:
+Choose from **5 AI frameworks** and **5 LLM providers** when generating your project:
 
 ```bash
 # PydanticAI with OpenAI (default)

docs/guides/configuration.md

@@ -10,7 +10,7 @@ All available options when generating a project.
 | `--orm` | `sqlalchemy`, `sqlmodel` | ORM choice (SQLModel for simplified syntax) |
 | `--oauth-google` | flag | Enable Google OAuth2 login |
 | `--ai-framework` | `pydantic_ai`, `pydantic_deep`, `langchain`, `langgraph`, `deepagents` | AI agent framework |
-| `--llm-provider` | `openai`, `anthropic`, `google`, `openrouter` | LLM provider |
+| `--llm-provider` | `openai`, `anthropic`, `google`, `openrouter`, `orcarouter` | LLM provider |
 | `--task-queue` | `none`, `celery`, `taskiq`, `arq` | Background task queue (Redis-backed). **Prefect** is available via the interactive wizard. |
 | `--frontend` | `none`, `nextjs` | Frontend framework |
 
@@ -31,8 +31,8 @@ fastapi-fullstack create my_app --minimal
 
 | Framework | Providers | Description |
 |-----------|-----------|-------------|
-| `pydantic_ai` | OpenAI, Anthropic, Google, OpenRouter | Type-safe agents with WebSearch/WebFetch built-in |
-| `pydantic_deep` | OpenAI, Anthropic, Google | Deep coding assistant (filesystem tools, Docker/Daytona sandbox) |
+| `pydantic_ai` | OpenAI, Anthropic, Google, OpenRouter, OrcaRouter | Type-safe agents with WebSearch/WebFetch built-in |
+| `pydantic_deep` | OpenAI, Anthropic, Google, OpenRouter, OrcaRouter | Deep coding assistant (filesystem tools, Docker/Daytona sandbox) |
 | `langchain` | OpenAI, Anthropic, Google | Comprehensive chain-based agents |
 | `langgraph` | OpenAI, Anthropic, Google | Graph-based ReAct agents |
 | `deepagents` | OpenAI, Anthropic, Google | Agentic framework with subagent delegation |

docs/index.md

@@ -48,7 +48,7 @@ See the [Quick Start guide](guides/quick-start.md) for details and the [Installa
 
 | Framework | Streaming | Observability | Providers |
 |-----------|:---------:|:-------------:|:---------:|
-| **PydanticAI** | WebSocket | Logfire | OpenAI, Anthropic, OpenRouter |
+| **PydanticAI** | WebSocket | Logfire | OpenAI, Anthropic, OpenRouter, OrcaRouter |
 | **LangChain** | WebSocket | LangSmith | OpenAI, Anthropic |
 | **LangGraph** | WebSocket | LangSmith | OpenAI, Anthropic |
 

docs/rag.md

@@ -97,7 +97,7 @@ CROSS_ENCODER_MODEL=...    # Model name (default: cross-encoder/ms-marco-MiniLM-
 | Option | Values | Description |
 |--------|--------|-------------|
 | `enable_rag` | bool | Enable RAG functionality |
-| `embedding_provider` | auto-derived | Embedding model provider (auto-derived from LLM provider: OpenAI→openai, Anthropic→voyage, OpenRouter→sentence_transformers) |
+| `embedding_provider` | auto-derived | Embedding model provider (auto-derived from LLM provider: OpenAI→openai, Anthropic→voyage, Google→gemini, OpenRouter→openai, OrcaRouter→sentence_transformers) |
 | `pdf_parser` | `pymupdf`, `llamaparse` | PDF parsing method (set via `--pdf-parser` CLI flag) |
 | `enable_reranker` | bool | Enable reranking (set via `--reranker` CLI flag: none/cohere/cross_encoder) |
 

fastapi_gen/cli.py

@@ -395,11 +395,11 @@ def new(output: Path | None, no_input: bool, name: str | None, minimal: bool) ->
 )
 @click.option(
     "--llm-provider",
-    type=click.Choice(["openai", "anthropic", "google", "openrouter", "all"]),
+    type=click.Choice(["openai", "anthropic", "google", "openrouter", "orcarouter", "all"]),
     default="openai",
     help=(
         "LLM provider (default: openai). 'all' installs every SDK and lets users "
-        "pick the model at runtime. openrouter requires pydantic_ai."
+        "pick the model at runtime. openrouter/orcarouter requires pydantic_ai."
     ),
 )
 @click.option("--redis", is_flag=True, help="Enable Redis")
@@ -1362,6 +1362,7 @@ def templates() -> None:
     console.print("  --llm-provider anthropic        Anthropic (claude-opus-4-7)")
     console.print("  --llm-provider google           Google Gemini (gemini-2.5-flash)")
     console.print("  --llm-provider openrouter       OpenRouter (pydantic_ai only)")
+    console.print("  --llm-provider orcarouter       OrcaRouter (pydantic_ai only)")
     console.print(
         "  --websockets                    Enable WebSocket support (real-time chat streaming)"
     )

fastapi_gen/config.py

@@ -96,6 +96,7 @@ class LLMProviderType(StrEnum):
     ANTHROPIC = "anthropic"
     GOOGLE = "google"
     OPENROUTER = "openrouter"
+    ORCAROUTER = "orcarouter"
     ALL = "all"
 
 
@@ -450,15 +451,15 @@ def validate_option_combinations(self) -> "ProjectConfig":
             raise ValueError("Caching requires Redis to be enabled")
         if (
             self.ai_framework != AIFrameworkType.NONE
-            and self.llm_provider == LLMProviderType.OPENROUTER
+            and self.llm_provider in (LLMProviderType.OPENROUTER, LLMProviderType.ORCAROUTER)
             and self.ai_framework
             not in (
                 AIFrameworkType.PYDANTIC_AI,
                 AIFrameworkType.PYDANTIC_DEEP,
             )
         ):
             raise ValueError(
-                f"OpenRouter is only supported with PydanticAI or PydanticDeep, "
+                f"OpenRouter/OrcaRouter is only supported with PydanticAI or PydanticDeep, "
                 f"not {self.ai_framework.value}"
             )
         if (
@@ -809,6 +810,8 @@ def to_cookiecutter_context(self) -> dict[str, Any]:
             "use_google": self.llm_provider in (LLMProviderType.GOOGLE, LLMProviderType.ALL),
             "use_openrouter": self.llm_provider
             in (LLMProviderType.OPENROUTER, LLMProviderType.ALL),
+            "use_orcarouter": self.llm_provider
+            in (LLMProviderType.ORCAROUTER, LLMProviderType.ALL),
             "use_all_providers": self.llm_provider == LLMProviderType.ALL,
             # Legacy fixed values (always enabled, not user-configurable)
             # AI
@@ -900,19 +903,23 @@ def to_cookiecutter_context(self) -> dict[str, Any]:
                 if self.llm_provider == LLMProviderType.ANTHROPIC
                 else EmbeddingProviderType.GEMINI.value
                 if self.llm_provider == LLMProviderType.GOOGLE
+                else EmbeddingProviderType.SENTENCE_TRANSFORMERS.value
+                if self.llm_provider == LLMProviderType.ORCAROUTER
                 else EmbeddingProviderType.OPENAI.value
             ),
             "use_openai_embeddings": self.rag_features.enable_rag
             and self.llm_provider
             not in (
                 LLMProviderType.ANTHROPIC,
                 LLMProviderType.GOOGLE,
+                LLMProviderType.ORCAROUTER,
             ),
             "use_voyage_embeddings": self.rag_features.enable_rag
             and self.llm_provider == LLMProviderType.ANTHROPIC,
             "use_gemini_embeddings": self.rag_features.enable_rag
             and self.llm_provider == LLMProviderType.GOOGLE,
-            "use_sentence_transformers": False,
+            "use_sentence_transformers": self.rag_features.enable_rag
+            and self.llm_provider == LLMProviderType.ORCAROUTER,
             "enable_reranker": self.rag_features.enable_rag
             and self.rag_features.reranker_type != RerankerType.NONE,
             "use_cohere_reranker": self.rag_features.enable_rag

fastapi_gen/prompts.py

@@ -786,11 +786,14 @@ def prompt_llm_provider(ai_framework: AIFrameworkType) -> LLMProviderType:
         questionary.Choice("Google Gemini (gemini-2.5-flash)", value=LLMProviderType.GOOGLE),
     ]
 
-    # OpenRouter available for PydanticAI and PydanticDeep (both use pydantic-ai)
+    # OpenRouter and OrcaRouter available for PydanticAI and PydanticDeep (both use pydantic-ai)
     if ai_framework in (AIFrameworkType.PYDANTIC_AI, AIFrameworkType.PYDANTIC_DEEP):
         choices.append(
             questionary.Choice("OpenRouter (multi-provider)", value=LLMProviderType.OPENROUTER)
         )
+        choices.append(
+            questionary.Choice("OrcaRouter (model routing)", value=LLMProviderType.ORCAROUTER)
+        )
 
     return cast(
         LLMProviderType,

template/VARIABLES.md

@@ -217,7 +217,7 @@ These variables are set automatically by the generator.
 | `use_qdrant` | bool | `false` | Qdrant vector database is selected | Computed from `vector_store` |
 | `use_chromadb` | bool | `false` | ChromaDB vector database is selected (embedded mode) | Computed from `vector_store` |
 | `use_pgvector` | bool | `false` | pgvector (PostgreSQL extension) is selected | Computed from `vector_store`, requires PostgreSQL |
-| `embedding_provider` | enum | auto-derived | Embedding model provider. Auto-derived from LLM provider: OpenAI→openai, Anthropic→voyage, OpenRouter→sentence_transformers | Auto-derived from `llm_provider` |
+| `embedding_provider` | enum | auto-derived | Embedding model provider. Auto-derived from LLM provider: OpenAI→openai, Anthropic→voyage, Google→gemini, OpenRouter→openai, OrcaRouter→sentence_transformers | Auto-derived from `llm_provider` |
 | `use_openai_embeddings` | bool | `false` | OpenAI embeddings are selected | Computed from `llm_provider` |
 | `use_voyage_embeddings` | bool | `false` | Voyage AI embeddings are selected | Computed from `llm_provider` |
 | `use_gemini_embeddings` | bool | `false` | Google Gemini multimodal embeddings are selected | Computed from `llm_provider` |
@@ -238,7 +238,7 @@ These variables are set automatically by the generator.
 **Notes:**
 
 - RAG requires a vector database (Milvus, Qdrant, ChromaDB, or pgvector)
-- Embedding provider is auto-derived from LLM provider (OpenAI→openai, Anthropic→voyage, Google→gemini, OpenRouter→sentence_transformers)
+- Embedding provider is auto-derived from LLM provider (OpenAI→openai, Anthropic→voyage, Google→gemini, OpenRouter→openai, OrcaRouter→sentence_transformers)
 - Reranker is enabled via `--reranker` CLI flag (cohere, cross_encoder)
 - Cohere and Cross-Encoder rerankers improve search result relevance
 - LlamaParse requires an API key; PyMuPDF is free and local (with tables, OCR fallback)
@@ -275,11 +275,12 @@ These variables are set automatically by the generator.
 | `use_deepagents` | bool | `false` | DeepAgents (agentic coding, LangChain) is selected | Computed from `ai_framework` |
 | `use_pydantic_deep` | bool | `false` | PydanticDeep (deep agentic coding, Docker sandbox) is selected | Computed from `ai_framework` |
 | `sandbox_backend` | enum | `"state"` | Agent sandbox environment for DeepAgents/PydanticDeep. Values: `state`, `daytona` | Only used when `use_deepagents` or `use_pydantic_deep` is true |
-| `llm_provider` | enum | `"openai"` | LLM provider. Values: `openai`, `anthropic`, `google`, `openrouter` | - |
+| `llm_provider` | enum | `"openai"` | LLM provider. Values: `openai`, `anthropic`, `google`, `openrouter`, `orcarouter` | - |
 | `use_openai` | bool | `true` | OpenAI is selected | Computed from `llm_provider` |
 | `use_anthropic` | bool | `false` | Anthropic is selected | Computed from `llm_provider` |
 | `use_google` | bool | `false` | Google Gemini is selected | Computed from `llm_provider` |
 | `use_openrouter` | bool | `false` | OpenRouter is selected | Computed from `llm_provider` |
+| `use_orcarouter` | bool | `false` | OrcaRouter is selected | Computed from `llm_provider` |
 | `enable_langsmith` | bool | `false` | Enable LangSmith observability (tracing, prompt management) | Requires LangChain, LangGraph, or DeepAgents |
 | `enable_web_search` | bool | `false` | Web search. PydanticAI/PydanticDeep use the model-native WebSearch capability; LangChain/LangGraph/DeepAgents use a Tavily-backed tool (needs `TAVILY_API_KEY`) | Requires an AI framework |
 | `enable_web_fetch` | bool | `false` | Web fetch. PydanticAI/PydanticDeep use the model-native WebFetch capability; LangChain/LangGraph/DeepAgents use the portable `fetch_url` tool | Requires an AI framework |
@@ -298,7 +299,7 @@ These variables are set automatically by the generator.
 - PydanticAI uses `iter()` for full event streaming over WebSocket
 - LangGraph implements a ReAct (Reasoning + Acting) agent pattern with graph-based architecture
 - DeepAgents provides an agentic coding assistant with built-in filesystem tools (ls, read_file, write_file, edit_file, glob, grep) and task management
-- OpenRouter with LangChain, LangGraph, or DeepAgents is not supported
+- OpenRouter and OrcaRouter with LangChain, LangGraph, or DeepAgents is not supported
 
 ---
 

template/cookiecutter.json

@@ -81,6 +81,7 @@
   "use_anthropic": false,
   "use_google": false,
   "use_openrouter": false,
+  "use_orcarouter": false,
   "use_all_providers": false,
   "use_ai": true,
   "enable_conversation_persistence": true,

template/{{cookiecutter.project_slug}}/ENV_VARS.md

@@ -54,6 +54,9 @@ group is for and which are required vs optional.
 {%- if cookiecutter.use_openrouter %}
 | `OPENROUTER_API_KEY` | **required** | — | From openrouter.ai |
 {%- endif %}
+{%- if cookiecutter.use_orcarouter %}
+| `ORCAROUTER_API_KEY` | **required** | — | From orcarouter.ai |
+{%- endif %}
 {%- if cookiecutter.enable_logfire %}
 | `LOGFIRE_TOKEN` | optional | — | When set, ships traces to Logfire (logfire.pydantic.dev) |
 {%- endif %}

template/{{cookiecutter.project_slug}}/MANUAL_STEPS.md

@@ -60,6 +60,14 @@ These are used to sign JWTs and authenticate service-to-service calls. Rotate at
 - [ ] Set `OPENROUTER_API_KEY` in `.env`.
 {%- endif %}
 
+{%- if cookiecutter.use_orcarouter %}
+
+## OrcaRouter
+
+- [ ] Create API key at https://www.orcarouter.ai.
+- [ ] Set `ORCAROUTER_API_KEY` in `.env`.
+{%- endif %}
+
 {%- if cookiecutter.enable_oauth_google %}
 
 ## Google OAuth

template/{{cookiecutter.project_slug}}/backend/.env.example

@@ -208,9 +208,12 @@ GOOGLE_API_KEY=
 {%- if cookiecutter.use_openrouter %}
 OPENROUTER_API_KEY=
 {%- endif %}
+{%- if cookiecutter.use_orcarouter %}
+ORCAROUTER_API_KEY=
+{%- endif %}
 {%- if cookiecutter.use_all_providers %}
 # Multi-provider deployment — model name must include the provider prefix:
-#   openai/<model>, anthropic/<model>, google/<model>, openrouter/<provider>/<model>
+#   openai/<model>, anthropic/<model>, google/<model>, openrouter/<provider>/<model>, orcarouter/<provider>/<model>
 AI_MODEL=openai/gpt-5.5
 {%- elif cookiecutter.use_openai %}
 AI_MODEL=gpt-5.5
@@ -220,6 +223,8 @@ AI_MODEL=claude-opus-4-7
 AI_MODEL=gemini-2.5-flash
 {%- elif cookiecutter.use_openrouter %}
 AI_MODEL=anthropic/claude-opus-4-7
+{%- elif cookiecutter.use_orcarouter %}
+AI_MODEL=anthropic/claude-sonnet-4.6
 {%- endif %}
 AI_TEMPERATURE=0.7
 {%- if cookiecutter.use_pydantic_ai %}