fix typing

Embeddings integration

example.py

@@ -1,37 +0,0 @@
-import os
-
-import openai
-
-client = openai.OpenAI(
-    api_key=os.environ["CASHU_TOKEN"],
-    base_url=os.environ.get("ROUTSTR_API_URL", "https://api.routstr.com/v1"),
-    # base_url="http://roustrjfsdgfiueghsklchg.onion/v1",
-    # client=httpx.AsyncClient(
-    #     proxies={"http": "socks5://localhost:9050"},
-    # ),  # to use onion proxy (tor)
-)
-history: list = []
-
-
-def chat() -> None:
-    while True:
-        user_msg = {"role": "user", "content": input("\nYou: ")}
-        history.append(user_msg)
-        ai_msg = {"role": "assistant", "content": ""}
-
-        for chunk in client.chat.completions.create(
-            model=os.environ.get("MODEL", "openai/gpt-4o-mini"),
-            messages=history,
-            stream=True,
-        ):
-            if len(chunk.choices) > 0:
-                content = chunk.choices[0].delta.content
-                if content is not None:
-                    ai_msg["content"] += content
-                    print(content, end="", flush=True)
-        print()
-        history.append(ai_msg)
-
-
-if __name__ == "__main__":
-    chat()

examples/balance/check_balance.py

@@ -0,0 +1,11 @@
+import os
+
+import httpx
+
+# Use your Cashu token or API key as the Bearer token,
+# cashu token is hashed on the server and acts as an Temporary API key
+headers = {"Authorization": f"Bearer {os.environ.get('TOKEN')}"}
+base_url = os.environ.get("API_URL", "https://api.routstr.com/v1")
+
+resp = httpx.get(f"{base_url}/balance/info", headers=headers)
+print(resp.json())

examples/balance/create_balance.py

@@ -0,0 +1,15 @@
+import os
+
+import httpx
+
+# Send a Cashu token to the /create endpoint to get a persistent API key
+token = os.environ.get("TOKEN")
+if not token:
+    print("Please set TOKEN environment variable with a Cashu token")
+    exit(1)
+
+base_url = os.environ.get("API_URL", "https://api.routstr.com/v1")
+
+resp = httpx.get(f"{base_url}/balance/create", params={"initial_balance_token": token})
+
+print(resp.json())

examples/balance/refund_balance.py

@@ -0,0 +1,12 @@
+import os
+
+import httpx
+
+# Use your Cashu token or API key as the Bearer token
+headers = {"Authorization": f"Bearer {os.environ.get('TOKEN')}"}
+base_url = os.environ.get("API_URL", "https://api.routstr.com/v1")
+
+resp = httpx.post(f"{base_url}/balance/refund", headers=headers)
+
+print("Refund successful!")
+print(resp.json())

examples/balance/topup_balance.py

@@ -0,0 +1,16 @@
+import os
+
+import httpx
+
+# Use your Cashu token or API key as the Bearer token
+headers = {"Authorization": f"Bearer {os.environ.get('TOKEN')}"}
+base_url = os.environ.get("API_URL", "https://api.routstr.com/v1")
+
+# The Cashu token to top up with
+cashu_token = input("Enter Cashu token to top up: ")
+
+resp = httpx.post(
+    f"{base_url}/balance/topup", headers=headers, json={"cashu_token": cashu_token}
+)
+
+print(resp.json())

examples/chat_completions.py

@@ -0,0 +1,15 @@
+import os
+
+from openai import OpenAI
+
+client = OpenAI(
+    api_key=os.environ.get("TOKEN"),
+    base_url=os.environ.get("API_URL", "https://api.routstr.com/v1"),
+)
+
+response = client.chat.completions.create(
+    model=os.environ.get("MODEL", "gpt-5-nano"),
+    messages=[{"role": "user", "content": "Hello!"}],
+)
+
+print(response.choices[0].message.content)

examples/list_models.py

@@ -0,0 +1,19 @@
+import os
+
+import httpx
+from openai import OpenAI
+
+client = OpenAI(
+    api_key=os.environ.get("TOKEN", ""),
+    base_url=os.environ.get("API_URL", "https://api.routstr.com/v1"),
+)
+
+for model in client.models.list():
+    print(model.id)
+
+# OR
+
+models = httpx.get(
+    f"{client.base_url}/v1/models",
+    headers={"Authorization": f"Bearer {client.api_key}"},
+).json()

examples/responses/conversation.py

@@ -0,0 +1,31 @@
+import os
+
+from openai import OpenAI
+
+client = OpenAI(
+    api_key=os.environ.get("TOKEN"),
+    base_url=os.environ.get("API_URL", "https://api.routstr.com/v1"),
+)
+
+conversation = []  # type: ignore
+
+# First turn
+response1 = client.responses.create(  # type: ignore
+    model="o4-mini",
+    input="Hi, my name is Alice.",
+    conversation=conversation,
+)
+print("Response 1:", response1.output)
+
+# Note: The 'conversation' parameter might need to be constructed differently
+# depending on exact SDK/API spec. Typically, you pass back the previous turn's data.
+# Assuming the SDK manages or returns a conversation object/ID:
+# conversation.append(response1)
+
+# Second turn - demonstrating intent, actual implementation depends on strict API spec
+# response2 = client.responses.create(
+#     model="openai/gpt-4o-mini",
+#     input="What is my name?",
+#     conversation=conversation,
+# )
+# print("Response 2:", response2.output)

examples/responses/create.py

@@ -0,0 +1,17 @@
+import os
+
+from openai import OpenAI
+
+# The OpenAI SDK handles the 'responses' endpoint if it's updated to the latest version
+# and the base_url points to a compatible proxy like Routstr.
+client = OpenAI(
+    api_key=os.environ.get("TOKEN"),
+    base_url=os.environ.get("API_URL", "https://api.routstr.com/v1"),
+)
+
+response = client.responses.create(
+    model="gpt-5-mini",
+    input="Tell me a three sentence bedtime story about a unicorn.",
+)
+
+print(response.output)

examples/responses/streaming_response.py

@@ -0,0 +1,20 @@
+import os
+
+from openai import OpenAI
+
+client = OpenAI(
+    api_key=os.environ.get("TOKEN"),
+    base_url=os.environ.get("API_URL", "https://api.routstr.com/v1"),
+)
+
+stream = client.responses.create(
+    model="claude-4.5-sonnet",
+    input="Write a short poem about rust.",
+    stream=True,
+)
+
+for event in stream:
+    # Note: Depending on the SDK version and response structure,
+    # you might access event.output_delta or similar fields
+    print(event, end="", flush=True)
+print()

examples/responses/web_search.py

@@ -0,0 +1,16 @@
+import os
+
+from openai import OpenAI
+
+client = OpenAI(
+    api_key=os.environ.get("TOKEN"),
+    base_url=os.environ.get("API_URL", "https://api.routstr.com/v1"),
+)
+
+response = client.responses.create(
+    model="gpt-5-mini",
+    input="What is the latest news about AI?",
+    tools=[{"type": "web_search"}],  # type: ignore
+)
+
+print(response.output)

examples/streaming.py

@@ -0,0 +1,28 @@
+import os
+
+from openai import OpenAI
+
+client = OpenAI(
+    api_key=os.environ.get("TOKEN"),
+    base_url=os.environ.get("API_URL", "https://api.routstr.com/v1"),
+)
+
+messages = []
+while True:
+    messages.append({"role": "user", "content": input("\nYou: ")})
+
+    stream = client.chat.completions.create(
+        model=os.environ.get("MODEL", "gpt-5.1-mini"),
+        messages=messages,  # type: ignore
+        stream=True,
+    )
+
+    print("AI: ", end="")
+    response_content = ""
+    for chunk in stream:
+        if content := chunk.choices[0].delta.content:  # type: ignore
+            print(content, end="", flush=True)
+            response_content += content
+    print()
+
+    messages.append({"role": "assistant", "content": response_content})

examples/tor.py

@@ -0,0 +1,20 @@
+import os
+
+import httpx
+from openai import OpenAI
+
+# Requires `pip install "httpx[socks]"` and a running Tor proxy on port 9050
+client = OpenAI(
+    api_key=os.environ.get("TOKEN"),
+    base_url=os.environ.get("ONION_URL", "http://roustrjfsdgfiueghsklchg.onion/v1"),
+    http_client=httpx.Client(proxies="socks5://localhost:9050"),
+)
+
+print(
+    client.chat.completions.create(
+        model="openai/gpt-4o-mini",
+        messages=[{"role": "user", "content": "Hello from Tor!"}],
+    )
+    .choices[0]
+    .message.content
+)

pyproject.toml

@@ -73,6 +73,7 @@ packages = ["routstr"]
 [tool.ruff.lint]
 select = ["E", "F", "I"]
 ignore = ["E501"]
+exclude = ["examples"]
 
 [tool.mypy]
 python_version = "3.11"

routstr/payment/models.py

@@ -93,12 +93,32 @@ async def async_fetch_openrouter_models(source_filter: str | None = None) -> lis
 
     try:
         async with httpx.AsyncClient() as client:
-            response = await client.get(f"{base_url}/models", timeout=30)
-            response.raise_for_status()
-            data = response.json()
+            models_response, embeddings_response = await asyncio.gather(
+                client.get(f"{base_url}/models", timeout=30),
+                client.get(f"{base_url}/embeddings/models", timeout=30),
+                return_exceptions=True,
+            )
+
+            def process_models_response(
+                response: httpx.Response | BaseException,
+            ) -> list[dict]:
+                if not isinstance(response, BaseException):
+                    response.raise_for_status()
+                    data = response.json()
+                    return [
+                        model
+                        for model in data.get("data", [])
+                        if ":free" not in model.get("id", "").lower()
+                    ]
+                return []
 
             models_data: list[dict] = []
-            for model in data.get("data", []):
+            models_data.extend(process_models_response(models_response))
+            models_data.extend(process_models_response(embeddings_response))
+
+            # Apply source filter and exclusions
+            filtered_models = []
+            for model in models_data:
                 model_id = model.get("id", "")
 
                 if source_filter:
@@ -116,9 +136,9 @@ async def async_fetch_openrouter_models(source_filter: str | None = None) -> lis
                 if not _has_valid_pricing(model):
                     continue
 
-                models_data.append(model)
+                filtered_models.append(model)
 
-            return models_data
+            return filtered_models
     except Exception as e:
         logger.error(f"Error (async) fetching models from OpenRouter API: {e}")
         return []

routstr/proxy.py

@@ -65,7 +65,7 @@ def get_upstreams() -> list[BaseUpstreamProvider]:
 
 def get_model_instance(model_id: str) -> Model | None:
     """Get Model instance by ID from global cache."""
-    return _model_instances.get(model_id)
+    return _model_instances.get(model_id.lower())
 
 
 def get_provider_for_model(model_id: str) -> BaseUpstreamProvider | None:

routstr/upstream/base.py

@@ -734,51 +734,53 @@ class BaseUpstreamProvider:
                     await client.aclose()
                 return mapped_error
 
-            if path.endswith("chat/completions"):
-                client_wants_streaming = False
-                if request_body:
-                    try:
-                        request_data = json.loads(request_body)
-                        client_wants_streaming = request_data.get("stream", False)
-                        logger.debug(
-                            "Chat completion request analysis",
-                            extra={
-                                "client_wants_streaming": client_wants_streaming,
-                                "model": request_data.get("model", "unknown"),
-                                "key_hash": key.hashed_key[:8] + "...",
-                            },
-                        )
-                    except json.JSONDecodeError:
-                        logger.warning(
-                            "Failed to parse request body JSON for streaming detection"
-                        )
+            if path.endswith("chat/completions") or path.endswith("embeddings"):
+                if path.endswith("chat/completions"):
+                    client_wants_streaming = False
+                    if request_body:
+                        try:
+                            request_data = json.loads(request_body)
+                            client_wants_streaming = request_data.get("stream", False)
+                            logger.debug(
+                                "Chat completion request analysis",
+                                extra={
+                                    "client_wants_streaming": client_wants_streaming,
+                                    "model": request_data.get("model", "unknown"),
+                                    "key_hash": key.hashed_key[:8] + "...",
+                                },
+                            )
+                        except json.JSONDecodeError:
+                            logger.warning(
+                                "Failed to parse request body JSON for streaming detection"
+                            )
 
-                content_type = response.headers.get("content-type", "")
-                upstream_is_streaming = "text/event-stream" in content_type
-                is_streaming = client_wants_streaming and upstream_is_streaming
+                    content_type = response.headers.get("content-type", "")
+                    upstream_is_streaming = "text/event-stream" in content_type
+                    is_streaming = client_wants_streaming and upstream_is_streaming
 
-                logger.debug(
-                    "Response type analysis",
-                    extra={
-                        "is_streaming": is_streaming,
-                        "client_wants_streaming": client_wants_streaming,
-                        "upstream_is_streaming": upstream_is_streaming,
-                        "content_type": content_type,
-                        "key_hash": key.hashed_key[:8] + "...",
-                    },
-                )
-
-                if is_streaming and response.status_code == 200:
-                    result = await self.handle_streaming_chat_completion(
-                        response, key, max_cost_for_model
+                    logger.debug(
+                        "Response type analysis",
+                        extra={
+                            "is_streaming": is_streaming,
+                            "client_wants_streaming": client_wants_streaming,
+                            "upstream_is_streaming": upstream_is_streaming,
+                            "content_type": content_type,
+                            "key_hash": key.hashed_key[:8] + "...",
+                        },
                     )
-                    background_tasks = BackgroundTasks()
-                    background_tasks.add_task(response.aclose)
-                    background_tasks.add_task(client.aclose)
-                    result.background = background_tasks
-                    return result
 
-                elif response.status_code == 200:
+                    if is_streaming and response.status_code == 200:
+                        result = await self.handle_streaming_chat_completion(
+                            response, key, max_cost_for_model
+                        )
+                        background_tasks = BackgroundTasks()
+                        background_tasks.add_task(response.aclose)
+                        background_tasks.add_task(client.aclose)
+                        result.background = background_tasks
+                        return result
+
+                # Handle both non-streaming chat completions and embeddings
+                if response.status_code == 200:
                     try:
                         return await self.handle_non_streaming_chat_completion(
                             response, key, session, max_cost_for_model
@@ -1519,9 +1521,9 @@ class BaseUpstreamProvider:
                     error_response.headers["X-Cashu"] = refund_token
                     return error_response
 
-                if path.endswith("chat/completions"):
+                if path.endswith("chat/completions") or path.endswith("embeddings"):
                     logger.debug(
-                        "Processing chat completion response",
+                        "Processing completion/embeddings response",
                         extra={"path": path, "amount": amount, "unit": unit},
                     )
 
@@ -1770,15 +1772,32 @@ class BaseUpstreamProvider:
     async def _fetch_openrouter_models(self) -> list[dict]:
         """Fetch models from OpenRouter API."""
         url = "https://openrouter.ai/api/v1/models"
+        embeddings_url = "https://openrouter.ai/api/v1/embeddings/models"
+
         async with httpx.AsyncClient(timeout=30.0) as client:
-            response = await client.get(url)
-            response.raise_for_status()
-            models = response.json()
-            return [
-                model
-                for model in models.get("data", [])
-                if ":free" not in model.get("id", "").lower()
-            ]
+            models_response, embeddings_response = await asyncio.gather(
+                client.get(url), client.get(embeddings_url), return_exceptions=True
+            )
+
+            all_models = []
+
+            def process_models_response(
+                response: httpx.Response | BaseException,
+            ) -> list[dict]:
+                if not isinstance(response, BaseException):
+                    response.raise_for_status()
+                    data = response.json()
+                    return [
+                        model
+                        for model in data.get("data", [])
+                        if ":free" not in model.get("id", "").lower()
+                    ]
+                return []
+
+            all_models.extend(process_models_response(models_response))
+            all_models.extend(process_models_response(embeddings_response))
+
+            return all_models
 
     async def _fetch_provider_models(self) -> dict:
         """Fetch models from provider's API."""

routstr/upstream/openrouter.py

@@ -50,7 +50,13 @@ class OpenRouterUpstreamProvider(BaseUpstreamProvider):
     async def fetch_models(self) -> list[Model]:
         """Fetch all OpenRouter models."""
         models_data = await async_fetch_openrouter_models()
-        return [Model(**model) for model in models_data]  # type: ignore
+        models = [Model(**model) for model in models_data]  # type: ignore
+        # manual alias for openai/text-embedding-ada-002 due to openrouter api bug
+        for model in models:
+            if model.id == "openai/text-embedding-ada-002":
+                model.alias_ids = ["text-embedding-ada-002-v2"]
+                break
+        return models
 
     async def get_balance(self) -> float | None:
         """Get the current account balance from OpenRouter.

tests/integration/test_embeddings.py

@@ -0,0 +1,97 @@
+import json
+from typing import Any
+from unittest.mock import AsyncMock, MagicMock, patch
+
+import pytest
+from httpx import AsyncClient
+
+
+@pytest.mark.integration
+@pytest.mark.asyncio
+async def test_proxy_embeddings_endpoint(authenticated_client: AsyncClient) -> None:
+    """Test the embeddings endpoint proxy functionality"""
+
+    test_payload = {
+        "model": "text-embedding-ada-002",
+        "input": "The quick brown fox",
+    }
+
+    mock_response_data = {
+        "object": "list",
+        "data": [
+            {"object": "embedding", "embedding": [0.0023, -0.0012, 0.0045], "index": 0}
+        ],
+        "model": "text-embedding-ada-002",
+        "usage": {"prompt_tokens": 5, "total_tokens": 5},
+    }
+
+    with patch("httpx.AsyncClient.send") as mock_send:
+        # Create a proper async generator for iter_bytes
+        async def mock_iter_bytes(*args: Any, **kwargs: Any) -> Any:
+            yield json.dumps(mock_response_data).encode()
+
+        mock_response = AsyncMock()
+        mock_response.status_code = 200
+        mock_response.headers = {"content-type": "application/json"}
+        mock_response.text = json.dumps(mock_response_data)
+        # Use MagicMock for synchronous .json() method
+        mock_response.json = MagicMock(return_value=mock_response_data)
+        mock_response.iter_bytes = mock_iter_bytes
+        mock_response.aiter_bytes = mock_iter_bytes
+        mock_send.return_value = mock_response
+
+        # Make POST request to embeddings endpoint
+        response = await authenticated_client.post("/v1/embeddings", json=test_payload)
+
+        assert response.status_code == 200
+        response_data = response.json()
+        assert response_data["object"] == "list"
+        assert len(response_data["data"]) == 1
+        assert response_data["data"][0]["object"] == "embedding"
+
+        # Verify request was forwarded
+        mock_send.assert_called_once()
+        forwarded_request = mock_send.call_args[0][0]
+        # Verify the path ends with embeddings
+        # Note: forwarded path might be full URL
+        assert str(forwarded_request.url).endswith("embeddings")
+
+
+@pytest.mark.integration
+@pytest.mark.asyncio
+async def test_model_case_insensitivity(authenticated_client: AsyncClient) -> None:
+    """Test that model lookups are case insensitive"""
+
+    # We'll use a mixed-case model ID that should match the lowercase one in the system
+    # We assume 'gpt-3.5-turbo' is available in the mock env/database
+
+    test_payload = {
+        "model": "GPT-3.5-TURBO",
+        "messages": [{"role": "user", "content": "Hello"}],
+    }
+
+    with patch("httpx.AsyncClient.send") as mock_send:
+        mock_response_data = {
+            "id": "chatcmpl-123",
+            "object": "chat.completion",
+            "choices": [{"message": {"content": "Hi"}}],
+            "usage": {"total_tokens": 10},
+        }
+
+        async def mock_iter_bytes(*args: Any, **kwargs: Any) -> Any:
+            yield json.dumps(mock_response_data).encode()
+
+        mock_response = AsyncMock()
+        mock_response.status_code = 200
+        mock_response.headers = {"content-type": "application/json"}
+        mock_response.text = json.dumps(mock_response_data)
+        mock_response.json = MagicMock(return_value=mock_response_data)
+        mock_response.iter_bytes = mock_iter_bytes
+        mock_response.aiter_bytes = mock_iter_bytes
+        mock_send.return_value = mock_response
+
+        response = await authenticated_client.post(
+            "/v1/chat/completions", json=test_payload
+        )
+
+        assert response.status_code == 200