Optimize streaming usage extraction and Cashu refund buffering

example.py

@@ -0,0 +1,37 @@
+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

@@ -1,11 +0,0 @@
-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

@@ -1,15 +0,0 @@
-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

@@ -1,12 +0,0 @@
-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

@@ -1,16 +0,0 @@
-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

@@ -1,15 +0,0 @@
-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

@@ -1,19 +0,0 @@
-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

@@ -1,31 +0,0 @@
-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

@@ -1,17 +0,0 @@
-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

@@ -1,20 +0,0 @@
-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

@@ -1,16 +0,0 @@
-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

@@ -1,28 +0,0 @@
-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

@@ -1,20 +0,0 @@
-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
-)

migrations/versions/b9667ffc5701_alias_ids.py

@@ -1,37 +0,0 @@
-"""alias-ids
-
-Revision ID: b9667ffc5701
-Revises: lightning_invoices
-Create Date: 2025-12-25 19:30:44.673350
-"""
-
-import sqlalchemy as sa
-import sqlmodel
-from alembic import op
-
-# revision identifiers, used by Alembic.
-revision = "b9667ffc5701"
-down_revision = "lightning_invoices"
-branch_labels = None
-depends_on = None
-
-
-def upgrade() -> None:
-    # ### commands auto generated by Alembic ###
-    op.add_column(
-        "models",
-        sa.Column("canonical_slug", sqlmodel.sql.sqltypes.AutoString(), nullable=True),
-    )
-    op.add_column(
-        "models",
-        sa.Column("alias_ids", sqlmodel.sql.sqltypes.AutoString(), nullable=True),
-    )
-
-    # ### end Alembic commands ###
-
-
-def downgrade() -> None:
-    # ### commands auto generated by Alembic - please adjust! ###
-    op.drop_column("models", "alias_ids")
-    op.drop_column("models", "canonical_slug")
-    # ### end Alembic commands ###

pyproject.toml

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

routstr/algorithm.py

@@ -206,20 +206,19 @@ def create_model_mappings(
         alias: str, model: "Model", provider: "BaseUpstreamProvider"
     ) -> None:
         """Set alias to model/provider if not set or if new model is preferred."""
-        alias_lower = alias.lower()
-        existing_model = model_instances.get(alias_lower)
+        existing_model = model_instances.get(alias)
         if not existing_model:
             # No existing mapping, set it
-            model_instances[alias_lower] = model
-            provider_map[alias_lower] = provider
+            model_instances[alias] = model
+            provider_map[alias] = provider
         else:
             # Check if candidate should replace existing
-            existing_provider = provider_map[alias_lower]
+            existing_provider = provider_map[alias]
             if should_prefer_model(
                 model, provider, existing_model, existing_provider, alias
             ):
-                model_instances[alias_lower] = model
-                provider_map[alias_lower] = provider
+                model_instances[alias] = model
+                provider_map[alias] = provider
 
     def process_provider_models(
         upstream: "BaseUpstreamProvider", is_openrouter: bool = False

routstr/core/admin.py

@@ -1493,8 +1493,20 @@ class ModelCreate(BaseModel):
     per_request_limits: dict[str, object] | None = None
     top_provider: dict[str, object] | None = None
     upstream_provider_id: int | None = None
-    canonical_slug: str | None = None
-    alias_ids: list[str] | None = None
+    enabled: bool = True
+
+
+class ModelUpdate(BaseModel):
+    id: str
+    name: str
+    description: str
+    created: int
+    context_length: int
+    architecture: dict[str, object]
+    pricing: dict[str, object]
+    per_request_limits: dict[str, object] | None = None
+    top_provider: dict[str, object] | None = None
+    upstream_provider_id: int | None = None
     enabled: bool = True
 
 
@@ -2404,80 +2416,44 @@ async def admin_upstream_providers(request: Request) -> str:
     "/api/upstream-providers/{provider_id}/models",
     dependencies=[Depends(require_admin_api)],
 )
-async def upsert_provider_model(
+async def create_provider_model(
     provider_id: int, payload: ModelCreate
 ) -> dict[str, object]:
-    print(payload)
-    logger.info(
-        f"UPSERT_PROVIDER_MODEL called: provider_id={provider_id}, model_id={payload.id}"
-    )
     async with create_session() as session:
         provider = await session.get(UpstreamProviderRow, provider_id)
         if not provider:
             raise HTTPException(status_code=404, detail="Provider not found")
 
-        # Try to get existing model
-        existing_row = await session.get(ModelRow, (payload.id, provider_id))
+        exists = await session.get(ModelRow, (payload.id, provider_id))
+        if exists:
+            raise HTTPException(
+                status_code=409,
+                detail="Model with this ID already exists for this provider",
+            )
 
-        if existing_row:
-            # Update existing model
-            logger.info(f"Updating existing model: {payload.id}")
-            existing_row.name = payload.name
-            existing_row.description = payload.description
-            existing_row.created = int(payload.created)
-            existing_row.context_length = int(payload.context_length)
-            existing_row.architecture = json.dumps(payload.architecture)
-            existing_row.pricing = json.dumps(payload.pricing)
-            existing_row.sats_pricing = None
-            existing_row.per_request_limits = (
+        row = ModelRow(
+            id=payload.id,
+            name=payload.name,
+            description=payload.description,
+            created=int(payload.created),
+            context_length=int(payload.context_length),
+            architecture=json.dumps(payload.architecture),
+            pricing=json.dumps(payload.pricing),
+            sats_pricing=None,
+            per_request_limits=(
                 json.dumps(payload.per_request_limits)
                 if payload.per_request_limits is not None
                 else None
-            )
-            existing_row.top_provider = (
+            ),
+            top_provider=(
                 json.dumps(payload.top_provider) if payload.top_provider else None
-            )
-            existing_row.canonical_slug = payload.canonical_slug
-            existing_row.alias_ids = (
-                json.dumps(payload.alias_ids) if payload.alias_ids else None
-            )
-            existing_row.enabled = payload.enabled
-
-            session.add(existing_row)
-            await session.commit()
-            await session.refresh(existing_row)
-            row = existing_row
-
-        else:
-            # Create new model
-            logger.info(f"Creating new model: {payload.id}")
-            row = ModelRow(
-                id=payload.id,
-                name=payload.name,
-                description=payload.description,
-                created=int(payload.created),
-                context_length=int(payload.context_length),
-                architecture=json.dumps(payload.architecture),
-                pricing=json.dumps(payload.pricing),
-                sats_pricing=None,
-                per_request_limits=(
-                    json.dumps(payload.per_request_limits)
-                    if payload.per_request_limits is not None
-                    else None
-                ),
-                top_provider=(
-                    json.dumps(payload.top_provider) if payload.top_provider else None
-                ),
-                canonical_slug=payload.canonical_slug,
-                alias_ids=(
-                    json.dumps(payload.alias_ids) if payload.alias_ids else None
-                ),
-                upstream_provider_id=provider_id,
-                enabled=payload.enabled,
-            )
-            session.add(row)
-            await session.commit()
-            await session.refresh(row)
+            ),
+            upstream_provider_id=provider_id,
+            enabled=payload.enabled,
+        )
+        session.add(row)
+        await session.commit()
+        await session.refresh(row)
 
     await refresh_model_maps()
     return _row_to_model(
@@ -2485,20 +2461,6 @@ async def upsert_provider_model(
     ).dict()  # type: ignore
 
 
-@admin_router.patch(
-    "/api/upstream-providers/{provider_id}/models/{model_id:path}",
-    dependencies=[Depends(require_admin_api)],
-)
-async def update_provider_model_legacy(
-    provider_id: int, model_id: str, payload: ModelCreate
-) -> dict[str, object]:
-    """Legacy PATCH endpoint - redirects to upsert POST endpoint for backward compatibility."""
-    logger.info(
-        f"LEGACY_PATCH_UPDATE called: provider_id={provider_id}, model_id={model_id}"
-    )
-    return await upsert_provider_model(provider_id, payload)
-
-
 @admin_router.get(
     "/api/upstream-providers/{provider_id}/models/{model_id:path}",
     dependencies=[Depends(require_admin_api)],
@@ -2519,6 +2481,76 @@ async def get_provider_model(provider_id: int, model_id: str) -> dict[str, objec
         ).dict()  # type: ignore
 
 
+@admin_router.patch(
+    "/api/upstream-providers/{provider_id}/models/{model_id:path}",
+    dependencies=[Depends(require_admin_api)],
+)
+async def update_provider_model(
+    provider_id: int, model_id: str, payload: ModelUpdate
+) -> dict[str, object]:
+    if payload.id != model_id:
+        raise HTTPException(status_code=400, detail="Path id does not match payload id")
+
+    async with create_session() as session:
+        provider = await session.get(UpstreamProviderRow, provider_id)
+        if not provider:
+            raise HTTPException(status_code=404, detail="Provider not found")
+
+        row = await session.get(ModelRow, (model_id, provider_id))
+        if not row:
+            raise HTTPException(
+                status_code=404, detail="Model not found for this provider"
+            )
+
+        row.name = payload.name
+        row.description = payload.description
+        row.created = int(payload.created)
+        row.context_length = int(payload.context_length)
+        row.architecture = json.dumps(payload.architecture)
+        row.pricing = json.dumps(payload.pricing)
+        row.sats_pricing = None
+        row.per_request_limits = (
+            json.dumps(payload.per_request_limits)
+            if payload.per_request_limits is not None
+            else None
+        )
+        row.top_provider = (
+            json.dumps(payload.top_provider) if payload.top_provider else None
+        )
+        was_disabled = not row.enabled
+        row.enabled = payload.enabled
+
+        session.add(row)
+        await session.commit()
+        await session.refresh(row)
+
+    if was_disabled and payload.enabled:
+        from ..payment.models import _cleanup_enabled_models_once
+
+        try:
+            await _cleanup_enabled_models_once()
+        except Exception as e:
+            logger.warning(
+                f"Failed to run model cleanup after enabling: {e}",
+                extra={"model_id": model_id, "error": str(e)},
+            )
+
+    await refresh_model_maps()
+    return _row_to_model(
+        row, apply_provider_fee=True, provider_fee=provider.provider_fee
+    ).dict()  # type: ignore
+
+
+@admin_router.put(
+    "/api/upstream-providers/{provider_id}/models/{model_id:path}",
+    dependencies=[Depends(require_admin_api)],
+)
+async def update_provider_model_put(
+    provider_id: int, model_id: str, payload: ModelUpdate
+) -> dict[str, object]:
+    return await update_provider_model(provider_id, model_id, payload)
+
+
 @admin_router.delete(
     "/api/upstream-providers/{provider_id}/models/{model_id:path}",
     dependencies=[Depends(require_admin_api)],

routstr/core/db.py

@@ -68,10 +68,6 @@ class ModelRow(SQLModel, table=True):  # type: ignore
     sats_pricing: str | None = Field(default=None)
     per_request_limits: str | None = Field(default=None)
     top_provider: str | None = Field(default=None)
-    canonical_slug: str | None = Field(default=None, description="Canonical model slug")
-    alias_ids: str | None = Field(
-        default=None, description="JSON array of model alias IDs"
-    )
     enabled: bool = Field(default=True, description="Whether this model is enabled")
     upstream_provider: "UpstreamProviderRow" = Relationship(back_populates="models")
 

routstr/core/main.py

@@ -14,6 +14,7 @@ from ..balance import balance_router, deprecated_wallet_router
 from ..discovery import providers_cache_refresher, providers_router
 from ..nip91 import announce_provider
 from ..payment.models import (
+    cleanup_enabled_models_periodically,
     models_router,
     update_sats_pricing,
 )
@@ -48,6 +49,7 @@ async def lifespan(_: FastAPI) -> AsyncGenerator[None, None]:
     nip91_task = None
     providers_task = None
     models_refresh_task = None
+    models_cleanup_task = None
     model_maps_refresh_task = None
 
     try:
@@ -78,17 +80,13 @@ async def lifespan(_: FastAPI) -> AsyncGenerator[None, None]:
         _update_prices_task = asyncio.create_task(_update_prices())
         _initialize_upstreams_task = asyncio.create_task(initialize_upstreams())
 
-        # ensure both setup tasks complete
-        await asyncio.gather(
-            _update_prices_task, _initialize_upstreams_task, return_exceptions=True
-        )
-
         btc_price_task = asyncio.create_task(update_prices_periodically())
         pricing_task = asyncio.create_task(update_sats_pricing())
         if global_settings.models_refresh_interval_seconds > 0:
             models_refresh_task = asyncio.create_task(
                 refresh_upstreams_models_periodically(get_upstreams())
             )
+        models_cleanup_task = asyncio.create_task(cleanup_enabled_models_periodically())
         model_maps_refresh_task = asyncio.create_task(refresh_model_maps_periodically())
         payout_task = asyncio.create_task(periodic_payout())
         if global_settings.nsec:
@@ -96,6 +94,11 @@ async def lifespan(_: FastAPI) -> AsyncGenerator[None, None]:
         if global_settings.providers_refresh_interval_seconds > 0:
             providers_task = asyncio.create_task(providers_cache_refresher())
 
+        # ensure both setup tasks complete
+        await asyncio.gather(
+            _update_prices_task, _initialize_upstreams_task, return_exceptions=True
+        )
+
         yield
 
     except asyncio.CancelledError:
@@ -122,6 +125,8 @@ async def lifespan(_: FastAPI) -> AsyncGenerator[None, None]:
             providers_task.cancel()
         if models_refresh_task is not None:
             models_refresh_task.cancel()
+        if models_cleanup_task is not None:
+            models_cleanup_task.cancel()
         if model_maps_refresh_task is not None:
             model_maps_refresh_task.cancel()
 
@@ -139,6 +144,8 @@ async def lifespan(_: FastAPI) -> AsyncGenerator[None, None]:
                 tasks_to_wait.append(providers_task)
             if models_refresh_task is not None:
                 tasks_to_wait.append(models_refresh_task)
+            if models_cleanup_task is not None:
+                tasks_to_wait.append(models_cleanup_task)
             if model_maps_refresh_task is not None:
                 tasks_to_wait.append(model_maps_refresh_task)
 

routstr/core/middleware.py

@@ -55,16 +55,7 @@ class LoggingMiddleware(BaseHTTPMiddleware):
                 "headers": {
                     k: v
                     for k, v in request.headers.items()
-                    if k.lower()
-                    not in [
-                        "authorization",
-                        "x-cashu",
-                        "cookie",
-                        "cf-connecting-ip",
-                        "cf-ipcountry",
-                        "x-forwarded-for",
-                        "x-real-ip",
-                    ]
+                    if k.lower() not in ["authorization", "x-cashu", "cookie"]
                 },
                 "body_size": len(request_body) if request_body else 0,
             },

routstr/core/settings.py

@@ -59,7 +59,7 @@ class Settings(BaseSettings):
     cors_origins: list[str] = Field(default_factory=lambda: ["*"], env="CORS_ORIGINS")
     tor_proxy_url: str = Field(default="socks5://127.0.0.1:9050", env="TOR_PROXY_URL")
     providers_refresh_interval_seconds: int = Field(
-        default=0, env="PROVIDERS_REFRESH_INTERVAL_SECONDS"
+        default=300, env="PROVIDERS_REFRESH_INTERVAL_SECONDS"
     )
     pricing_refresh_interval_seconds: int = Field(
         default=120, env="PRICING_REFRESH_INTERVAL_SECONDS"

routstr/discovery.py

@@ -6,7 +6,7 @@ from typing import Any
 
 import httpx
 import websockets
-from fastapi import APIRouter, HTTPException
+from fastapi import APIRouter
 
 from .core.logging import get_logger
 from .core.settings import settings
@@ -389,9 +389,6 @@ async def get_providers(
     Return cached providers. If include_json, return provider+health; otherwise provider only.
     Optional filter by pubkey.
     """
-    if settings.providers_refresh_interval_seconds == 0:
-        raise HTTPException(status_code=404, detail="Provider discovery is disabled")
-
     cache = await get_cache()
     if not cache:
         await refresh_providers_cache(pubkey=pubkey)

routstr/payment/cost_calculation.py

@@ -5,7 +5,6 @@ from pydantic.v1 import BaseModel
 from ..core import get_logger
 from ..core.db import AsyncSession
 from ..core.settings import settings
-from .price import sats_usd_price
 
 logger = get_logger(__name__)
 
@@ -65,56 +64,6 @@ async def calculate_cost(  # todo: can be sync
         )
         return cost_data
 
-    usage_data = response_data["usage"]
-
-    usd_cost = 0.0
-
-    # Prioritize cost_details.upstream_inference_cost
-    if "cost_details" in usage_data:
-        usd_cost = float(
-            usage_data["cost_details"].get("upstream_inference_cost", 0) or 0
-        )
-
-    # Fallback to cost field if upstream_inference_cost is 0
-    if usd_cost == 0 and "cost" in usage_data:
-        try:
-            usd_cost = float(usage_data.get("cost", 0) or 0)
-        except Exception:
-            pass
-
-    if usd_cost > 0:
-        try:
-            sats_per_usd = 1.0 / sats_usd_price()
-            cost_in_sats = usd_cost * sats_per_usd
-            cost_in_msats = math.ceil(cost_in_sats * 1000)
-
-            logger.info(
-                "Using cost from usage data/details",
-                extra={
-                    "usd_cost": usd_cost,
-                    "cost_in_sats": cost_in_sats,
-                    "cost_in_msats": cost_in_msats,
-                    "model": response_data.get("model", "unknown"),
-                },
-            )
-
-            return CostData(
-                base_msats=-1,
-                input_msats=-1,  # Cost field doesn't break down by token type
-                output_msats=-1,
-                total_msats=cost_in_msats,
-            )
-        except Exception as e:
-            logger.warning(
-                "Error calculating cost from usage data",
-                extra={
-                    "error": str(e),
-                    "usd_cost": usd_cost,
-                    "model": response_data.get("model", "unknown"),
-                },
-            )
-            # Fall through to token-based calculation
-
     MSATS_PER_1K_INPUT_TOKENS: float = (
         float(settings.fixed_per_1k_input_tokens) * 1000.0
     )
@@ -180,21 +129,21 @@ async def calculate_cost(  # todo: can be sync
         )
         return cost_data
 
-    input_tokens = usage_data.get("prompt_tokens", 0)
-    output_tokens = usage_data.get("completion_tokens", 0)
+    input_tokens = response_data.get("usage", {}).get("prompt_tokens", 0)
+    output_tokens = response_data.get("usage", {}).get("completion_tokens", 0)
 
     # added for response api
     input_tokens = (
-        input_tokens if input_tokens != 0 else usage_data.get("input_tokens", 0)
+        input_tokens
+        if input_tokens != 0
+        else response_data.get("usage", {}).get("input_tokens", 0)
     )
     output_tokens = (
-        output_tokens if output_tokens != 0 else usage_data.get("output_tokens", 0)
+        output_tokens
+        if output_tokens != 0
+        else response_data.get("usage", {}).get("output_tokens", 0)
     )
 
-    # added for response api
-    input_tokens = input_tokens if input_tokens != 0 else  response_data.get("usage", {}).get("input_tokens", 0)
-    output_tokens = output_tokens if output_tokens != 0 else  response_data.get("usage", {}).get("output_tokens", 0)
-
     input_msats = round(input_tokens / 1000 * MSATS_PER_1K_INPUT_TOKENS, 3)
 
     output_msats = round(output_tokens / 1000 * MSATS_PER_1K_OUTPUT_TOKENS, 3)

routstr/payment/models.py

@@ -5,9 +5,10 @@ import random
 import httpx
 from fastapi import APIRouter, Depends
 from pydantic.v1 import BaseModel
+from sqlmodel import select
 from sqlmodel.ext.asyncio.session import AsyncSession
 
-from ..core.db import ModelRow, get_session
+from ..core.db import ModelRow, create_session, get_session
 from ..core.logging import get_logger
 from ..core.settings import settings
 from .price import sats_usd_price
@@ -92,32 +93,12 @@ async def async_fetch_openrouter_models(source_filter: str | None = None) -> lis
 
     try:
         async with httpx.AsyncClient() as client:
-            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 []
+            response = await client.get(f"{base_url}/models", timeout=30)
+            response.raise_for_status()
+            data = response.json()
 
             models_data: list[dict] = []
-            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:
+            for model in data.get("data", []):
                 model_id = model.get("id", "")
 
                 if source_filter:
@@ -135,9 +116,9 @@ async def async_fetch_openrouter_models(source_filter: str | None = None) -> lis
                 if not _has_valid_pricing(model):
                     continue
 
-                filtered_models.append(model)
+                models_data.append(model)
 
-            return filtered_models
+            return models_data
     except Exception as e:
         logger.error(f"Error (async) fetching models from OpenRouter API: {e}")
         return []
@@ -184,7 +165,6 @@ def _row_to_model(
         enabled=row.enabled,
         upstream_provider_id=row.upstream_provider_id,
         canonical_slug=getattr(row, "canonical_slug", None),
-        alias_ids=json.loads(row.alias_ids) if row.alias_ids else None,
     )
 
     if apply_provider_fee:
@@ -382,7 +362,7 @@ def _update_model_sats_pricing(model: Model, sats_to_usd: float) -> Model:
 
 async def _update_sats_pricing_once() -> None:
     """Update sats pricing once for all provider models (in-memory only)."""
-    from ..proxy import get_upstreams, refresh_model_maps
+    from ..proxy import get_upstreams
 
     upstreams = get_upstreams()
     sats_to_usd = sats_usd_price()
@@ -399,7 +379,6 @@ async def _update_sats_pricing_once() -> None:
 
     if updated_count > 0:
         logger.info("Updated sats pricing", extra={"models_updated": updated_count})
-        await refresh_model_maps()
 
 
 async def update_sats_pricing() -> None:
@@ -410,13 +389,7 @@ async def update_sats_pricing() -> None:
     except Exception:
         pass
 
-    try:
-        await _update_sats_pricing_once()
-    except Exception as e:
-        logger.warning(
-            "Initial sats pricing update failed (will retry in loop)",
-            extra={"error": str(e)},
-        )
+    await _update_sats_pricing_once()
 
     while True:
         try:
@@ -440,6 +413,114 @@ async def update_sats_pricing() -> None:
             logger.error(f"Error updating sats pricing: {e}")
 
 
+async def cleanup_enabled_models_periodically() -> None:
+    """Background task to clean up enabled models that match upstream pricing.
+
+    When model is enabled (enabled=True), remove it from DB if it matches upstream pricing.
+    Keep it in DB only if pricing differs from upstream or if it's disabled.
+    """
+    interval = getattr(
+        settings, "models_cleanup_interval_seconds", 300
+    )  # 5 minutes default
+    if not interval or interval <= 0:
+        return
+
+    while True:
+        try:
+            await _cleanup_enabled_models_once()
+        except asyncio.CancelledError:
+            break
+        except Exception as e:
+            logger.error(
+                "Error during enabled models cleanup",
+                extra={"error": str(e), "error_type": type(e).__name__},
+            )
+
+        try:
+            jitter = max(0.0, float(interval) * 0.1)
+            await asyncio.sleep(interval + random.uniform(0, jitter))
+        except asyncio.CancelledError:
+            break
+
+
+async def _cleanup_enabled_models_once() -> None:
+    """Clean up enabled models that match upstream pricing."""
+    from ..proxy import get_upstreams
+
+    async with create_session() as session:
+        # Get all enabled models from DB
+        result = await session.exec(
+            select(ModelRow).where(
+                ModelRow.enabled,  # Only enabled models
+            )
+        )
+        db_models = result.all()
+
+        if not db_models:
+            return
+
+        upstreams = get_upstreams()
+        models_to_remove = []
+
+        for db_model in db_models:
+            # Find corresponding upstream model
+            upstream_model = None
+            for upstream in upstreams:
+                upstream_model = upstream.get_cached_model_by_id(db_model.id)
+                if upstream_model:
+                    break
+
+            if not upstream_model:
+                continue
+
+            # Compare pricing to see if they match
+            db_pricing = json.loads(db_model.pricing)
+            upstream_pricing = upstream_model.pricing.dict()
+
+            # Check if pricing matches (with small tolerance for float comparison)
+            pricing_matches = _pricing_matches(db_pricing, upstream_pricing)
+
+            if pricing_matches:
+                models_to_remove.append(db_model)
+                logger.info(
+                    f"Removing enabled model {db_model.id} - matches upstream pricing",
+                    extra={"model_id": db_model.id},
+                )
+
+        # Remove models that match upstream pricing
+        for model in models_to_remove:
+            await session.delete(model)
+
+        if models_to_remove:
+            await session.commit()
+            logger.info(
+                f"Cleaned up {len(models_to_remove)} enabled models that match upstream pricing"
+            )
+
+
+def _pricing_matches(
+    db_pricing: dict, upstream_pricing: dict, tolerance: float = 0.0
+) -> bool:
+    """Check if pricing dictionaries match within tolerance."""
+    keys_to_compare = [
+        "prompt",
+        "completion",
+        "request",
+        "image",
+        "web_search",
+        "internal_reasoning",
+    ]
+
+    for key in keys_to_compare:
+        db_val = int(float(db_pricing.get(key, 0.0)) * 1000000)
+        upstream_val = int(float(upstream_pricing.get(key, 0.0)) * 1000000)
+
+        if abs(db_val - upstream_val) > tolerance:
+            return False
+
+    return True
+
+
 @models_router.get("/v1/models")
 @models_router.get("/models", include_in_schema=False)
 async def models(session: AsyncSession = Depends(get_session)) -> dict:

routstr/proxy.py

@@ -65,12 +65,12 @@ 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.lower())
+    return _model_instances.get(model_id)
 
 
 def get_provider_for_model(model_id: str) -> BaseUpstreamProvider | None:
     """Get UpstreamProvider for model ID from global cache."""
-    return _provider_map.get(model_id.lower())
+    return _provider_map.get(model_id)
 
 
 def get_unique_models() -> list[Model]:
@@ -80,27 +80,31 @@ def get_unique_models() -> list[Model]:
 
 async def refresh_model_maps() -> None:
     """Refresh global model and provider maps using the cost-based algorithm."""
-    from sqlalchemy.orm import selectinload
-
     global _model_instances, _provider_map, _unique_models
 
+    # Gather database overrides and disabled models
     async with create_session() as session:
-        # Fetch all providers with their models in a single logical operation
-        query = select(UpstreamProviderRow).options(
-            selectinload(UpstreamProviderRow.models)  # type: ignore
+        result = await session.exec(select(ModelRow).where(ModelRow.enabled))
+        override_rows = result.all()
+
+        provider_result = await session.exec(select(UpstreamProviderRow))
+        providers_by_id = {p.id: p for p in provider_result.all()}
+
+        overrides_by_id: dict[str, tuple[ModelRow, float]] = {
+            row.id: (
+                row,
+                providers_by_id[row.upstream_provider_id].provider_fee
+                if row.upstream_provider_id in providers_by_id
+                else 1.01,
+            )
+            for row in override_rows
+            if row.upstream_provider_id is not None
+        }
+
+        disabled_result = await session.exec(
+            select(ModelRow.id).where(ModelRow.enabled == False)  # noqa: E712
         )
-        result = await session.exec(query)
-        provider_rows = result.all()
-
-    overrides_by_id: dict[str, tuple[ModelRow, float]] = {}
-    disabled_model_ids: set[str] = set()
-
-    for provider in provider_rows:
-        for model in provider.models:
-            if model.enabled:
-                overrides_by_id[model.id] = (model, provider.provider_fee)
-            else:
-                disabled_model_ids.add(model.id)
+        disabled_model_ids = {row for row in disabled_result.all()}
 
     _model_instances, _provider_map, _unique_models = create_model_mappings(
         upstreams=_upstreams,
@@ -142,7 +146,7 @@ async def proxy(
     request_body_dict = parse_request_body_json(request_body, path)
 
     if is_responses_api:
-        model_id = extract_model_from_responses_request(request_body_dict)
+        model_id = extract_model_from_responses_api_request(request_body_dict)
     else:
         model_id = request_body_dict.get("model", "unknown")
 
@@ -171,7 +175,7 @@ async def proxy(
 
     if x_cashu := headers.get("x-cashu", None):
         if is_responses_api:
-            return await upstream.handle_x_cashu_responses(
+            return await upstream.handle_x_cashu_responses_api(
                 request, x_cashu, path, max_cost_for_model, model_obj
             )
         else:
@@ -201,7 +205,7 @@ async def proxy(
     headers = upstream.prepare_headers(dict(request.headers))
 
     if is_responses_api:
-        response = await upstream.forward_responses_request(
+        response = await upstream.forward_responses_api_request(
             request,
             path,
             headers,
@@ -324,7 +328,7 @@ async def get_bearer_token_key(
         raise
 
 
-def extract_model_from_responses_request(request_body_dict: dict[str, Any]) -> str:
+def extract_model_from_responses_api_request(request_body_dict: dict[str, Any]) -> str:
     if model := request_body_dict.get("model"):
         return model
 
@@ -337,7 +341,7 @@ def extract_model_from_responses_request(request_body_dict: dict[str, Any]) -> s
 
     logger.warning(
         "No model found in Responses API request",
-        extra={"body_keys": list(request_body_dict.keys())}
+        extra={"body_keys": list(request_body_dict.keys())},
     )
     return "unknown"
 

routstr/upstream/handlers/cashu_payment_handler.py

@@ -0,0 +1,566 @@
+"""X-Cashu payment request handlers for different API types."""
+
+from __future__ import annotations
+
+import json
+from typing import TYPE_CHECKING, Callable
+
+import httpx
+from fastapi import BackgroundTasks, Request
+from fastapi.responses import Response, StreamingResponse
+
+from ...core import get_logger
+from ...wallet import send_token
+from ..payment.cashu_handler import (
+    CashuTokenError,
+    create_cashu_error_response,
+    validate_and_redeem_token,
+)
+from ..processing.http_client import HttpForwarder
+from ..processing.response_handler import ChatCompletionProcessor, ResponsesApiProcessor
+
+if TYPE_CHECKING:
+    from ...payment.models import Model
+
+logger = get_logger(__name__)
+
+
+class BaseCashuHandler:
+    """Base handler for X-Cashu payment requests."""
+
+    def __init__(self, base_url: str):
+        self.base_url = base_url
+        self.http_forwarder = HttpForwarder(base_url)
+
+    async def send_refund(self, amount: int, unit: str, mint: str | None = None) -> str:
+        """Create and send a refund token to the user."""
+        logger.debug(
+            "Creating refund token",
+            extra={"amount": amount, "unit": unit, "mint": mint},
+        )
+
+        max_retries = 3
+        last_exception = None
+
+        for attempt in range(max_retries):
+            try:
+                refund_token = await send_token(amount, unit=unit, mint_url=mint)
+
+                logger.info(
+                    "Refund token created successfully",
+                    extra={
+                        "amount": amount,
+                        "unit": unit,
+                        "mint": mint,
+                        "attempt": attempt + 1,
+                        "token_preview": refund_token[:20] + "..."
+                        if len(refund_token) > 20
+                        else refund_token,
+                    },
+                )
+
+                return refund_token
+            except Exception as e:
+                last_exception = e
+                if attempt < max_retries - 1:
+                    logger.warning(
+                        "Refund token creation failed, retrying",
+                        extra={
+                            "error": str(e),
+                            "error_type": type(e).__name__,
+                            "attempt": attempt + 1,
+                            "max_retries": max_retries,
+                            "amount": amount,
+                            "unit": unit,
+                            "mint": mint,
+                        },
+                    )
+                else:
+                    logger.error(
+                        "Failed to create refund token after all retries",
+                        extra={
+                            "error": str(e),
+                            "error_type": type(e).__name__,
+                            "attempt": attempt + 1,
+                            "max_retries": max_retries,
+                            "amount": amount,
+                            "unit": unit,
+                            "mint": mint,
+                        },
+                    )
+
+        raise Exception(
+            f"failed to create refund after {max_retries} attempts: {str(last_exception)}"
+        )
+
+    def _calculate_refund_amount(self, amount: int, unit: str, cost_msats: int) -> int:
+        """Calculate refund amount based on unit and cost."""
+        if unit == "msat":
+            return amount - cost_msats
+        elif unit == "sat":
+            return amount - (cost_msats + 999) // 1000
+        else:
+            raise ValueError(f"Invalid unit: {unit}")
+
+    async def _handle_upstream_error(
+        self,
+        response: httpx.Response,
+        amount: int,
+        unit: str,
+        mint: str | None,
+        api_type: str = "API",
+    ) -> Response:
+        """Handle upstream service errors with refund."""
+        logger.warning(
+            f"Upstream {api_type} request failed, processing refund",
+            extra={
+                "status_code": response.status_code,
+                "amount": amount,
+                "unit": unit,
+            },
+        )
+
+        refund_token = await self.send_refund(amount - 60, unit, mint)
+
+        logger.info(
+            f"Refund processed for failed upstream {api_type} request",
+            extra={
+                "status_code": response.status_code,
+                "refund_amount": amount,
+                "unit": unit,
+                "refund_token_preview": refund_token[:20] + "..."
+                if len(refund_token) > 20
+                else refund_token,
+            },
+        )
+
+        error_response = Response(
+            content=json.dumps(
+                {
+                    "error": {
+                        "message": f"Error forwarding {api_type} request to upstream",
+                        "type": "upstream_error",
+                        "code": response.status_code,
+                        "refund_token": refund_token,
+                    }
+                }
+            ),
+            status_code=response.status_code,
+            media_type="application/json",
+        )
+        error_response.headers["X-Cashu"] = refund_token
+        return error_response
+
+
+class ChatCompletionCashuHandler(BaseCashuHandler):
+    """Handler for X-Cashu paid chat completion requests."""
+
+    async def handle_request(
+        self,
+        request: Request,
+        x_cashu_token: str,
+        path: str,
+        headers: dict,
+        max_cost_for_model: int,
+        model_obj: Model,
+        prepare_request_body_func: Callable,
+        get_x_cashu_cost_func: Callable,
+        query_params_func: Callable,
+    ) -> Response | StreamingResponse:
+        """Handle chat completion request with X-Cashu payment."""
+        try:
+            amount, unit, mint = await validate_and_redeem_token(x_cashu_token, request)
+
+            # Forward request to upstream
+            request_body = await request.body()
+            response = await self.http_forwarder.forward_request(
+                request=request,
+                path=path,
+                headers=headers,
+                request_body=request_body,
+                query_params=query_params_func(path, request.query_params),
+                transform_body_func=prepare_request_body_func,
+                model_obj=model_obj,
+            )
+
+            # Handle upstream errors
+            if response.status_code != 200:
+                return await self._handle_upstream_error(
+                    response, amount, unit, mint, "chat completion"
+                )
+
+            # Process chat completion response
+            if path.endswith("chat/completions"):
+                result = await self._handle_chat_completion_response(
+                    response,
+                    amount,
+                    unit,
+                    max_cost_for_model,
+                    mint,
+                    get_x_cashu_cost_func,
+                )
+                background_tasks = BackgroundTasks()
+                background_tasks.add_task(response.aclose)
+                result.background = background_tasks
+                return result
+
+            # Default streaming response for other endpoints
+            return self._create_default_streaming_response(response)
+
+        except CashuTokenError as e:
+            return create_cashu_error_response(e, request, x_cashu_token)
+
+    async def _handle_chat_completion_response(
+        self,
+        response: httpx.Response,
+        amount: int,
+        unit: str,
+        max_cost_for_model: int,
+        mint: str | None,
+        get_x_cashu_cost_func: Callable,
+    ) -> Response | StreamingResponse:
+        """Handle chat completion response processing."""
+        # For X-Cashu requests, we force non-streaming response even if upstream is streaming,
+        # because we need to calculate the refund and include it in the headers.
+        # Refund token must be in the headers, which are sent before the body.
+
+        content = await response.aread()
+        content_str = content.decode("utf-8") if isinstance(content, bytes) else content
+
+        # Check if it was streaming content to process usage correctly
+        is_streaming_content = ChatCompletionProcessor.is_streaming_response(
+            content_str
+        )
+
+        if is_streaming_content:
+            # Buffer streaming content to calculate refund before sending headers.
+            # Headers (containing refund token) must be sent before body, preventing
+            # true streaming while supporting dynamic refunds.
+
+            # Extract usage data from buffered content
+            usage_data, model = ChatCompletionProcessor.extract_usage_from_streaming(
+                content_str
+            )
+
+            if usage_data and model:
+                try:
+                    response_data = {"usage": usage_data, "model": model}
+                    cost_data = await get_x_cashu_cost_func(
+                        response_data, max_cost_for_model
+                    )
+
+                    if cost_data:
+                        refund_amount = self._calculate_refund_amount(
+                            amount, unit, cost_data.total_msats
+                        )
+
+                        if refund_amount > 0:
+                            refund_token = await self.send_refund(
+                                refund_amount, unit, mint
+                            )
+                            # Refund token will be added to new response headers below
+                except Exception as e:
+                    logger.error(
+                        "Error calculating refund for buffered streaming response",
+                        extra={"error": str(e)},
+                    )
+
+            # We return a StreamingResponse that yields the buffered content
+            # This satisfies the client expecting a stream, while allowing us to set headers.
+
+            response_headers = ChatCompletionProcessor.clean_response_headers(
+                dict(response.headers)
+            )
+
+            if "refund_token" in locals() and refund_token:
+                response_headers["X-Cashu"] = refund_token
+
+            lines = content_str.strip().split("\n")
+            return StreamingResponse(
+                ChatCompletionProcessor.create_streaming_generator(lines),
+                status_code=response.status_code,
+                headers=response_headers,
+                media_type="text/event-stream",  # Keep original media type if possible, or text/event-stream
+            )
+        else:
+            return await self._handle_non_streaming_response(
+                content_str,
+                response,
+                amount,
+                unit,
+                max_cost_for_model,
+                mint,
+                get_x_cashu_cost_func,
+            )
+
+    async def _handle_non_streaming_response(
+        self,
+        content_str: str,
+        response: httpx.Response,
+        amount: int,
+        unit: str,
+        max_cost_for_model: int,
+        mint: str | None,
+        get_x_cashu_cost_func: Callable,
+    ) -> Response:
+        """Handle non-streaming chat completion response with refund calculation."""
+        response_headers = ChatCompletionProcessor.clean_response_headers(
+            dict(response.headers)
+        )
+
+        try:
+            response_json = json.loads(content_str)
+            cost_data = await get_x_cashu_cost_func(response_json, max_cost_for_model)
+
+            if cost_data:
+                refund_amount = self._calculate_refund_amount(
+                    amount, unit, cost_data.total_msats
+                )
+
+                if refund_amount > 0:
+                    refund_token = await self.send_refund(refund_amount, unit, mint)
+                    response_headers["X-Cashu"] = refund_token
+
+            return Response(
+                content=content_str,
+                status_code=response.status_code,
+                headers=response_headers,
+                media_type="application/json",
+            )
+
+        except json.JSONDecodeError:
+            # Emergency refund on parse error
+            emergency_refund = amount
+            refund_token = await send_token(emergency_refund, unit=unit, mint_url=mint)
+            response_headers["X-Cashu"] = refund_token
+
+            logger.warning(
+                "Emergency refund issued due to JSON parse error",
+                extra={
+                    "original_amount": amount,
+                    "refund_amount": emergency_refund,
+                    "deduction": 60,
+                },
+            )
+
+            return Response(
+                content=content_str,
+                status_code=response.status_code,
+                headers=response_headers,
+                media_type="application/json",
+            )
+
+    def _create_default_streaming_response(
+        self, response: httpx.Response
+    ) -> StreamingResponse:
+        """Create default streaming response for non-chat endpoints."""
+        background_tasks = BackgroundTasks()
+        background_tasks.add_task(response.aclose)
+
+        return StreamingResponse(
+            response.aiter_bytes(),
+            status_code=response.status_code,
+            headers=dict(response.headers),
+            background=background_tasks,
+        )
+
+
+class ResponsesApiCashuHandler(BaseCashuHandler):
+    """Handler for X-Cashu paid Responses API requests."""
+
+    async def handle_request(
+        self,
+        request: Request,
+        x_cashu_token: str,
+        path: str,
+        headers: dict,
+        max_cost_for_model: int,
+        model_obj: Model,
+        prepare_responses_api_request_body_func: Callable,
+        get_x_cashu_cost_func: Callable,
+        query_params_func: Callable,
+    ) -> Response | StreamingResponse:
+        """Handle Responses API request with X-Cashu payment."""
+        try:
+            amount, unit, mint = await validate_and_redeem_token(x_cashu_token, request)
+
+            # Forward request to upstream
+            request_body = await request.body()
+            response = await self.http_forwarder.forward_request(
+                request=request,
+                path=path,
+                headers=headers,
+                request_body=request_body,
+                query_params=query_params_func(path, request.query_params),
+                transform_body_func=prepare_responses_api_request_body_func,
+                model_obj=model_obj,
+            )
+
+            # Handle upstream errors
+            if response.status_code != 200:
+                return await self._handle_upstream_error(
+                    response, amount, unit, mint, "Responses API"
+                )
+
+            # Process Responses API response
+            if path.startswith("responses"):
+                result = await self._handle_responses_api_completion(
+                    response,
+                    amount,
+                    unit,
+                    max_cost_for_model,
+                    mint,
+                    get_x_cashu_cost_func,
+                )
+                background_tasks = BackgroundTasks()
+                background_tasks.add_task(response.aclose)
+                result.background = background_tasks
+                return result
+
+            # Default streaming response for other endpoints
+            return self._create_default_streaming_response(response)
+
+        except CashuTokenError as e:
+            return create_cashu_error_response(e, request, x_cashu_token)
+
+    async def _handle_responses_api_completion(
+        self,
+        response: httpx.Response,
+        amount: int,
+        unit: str,
+        max_cost_for_model: int,
+        mint: str | None,
+        get_x_cashu_cost_func: Callable,
+    ) -> Response | StreamingResponse:
+        """Handle Responses API completion response processing."""
+        # Force non-streaming for X-Cashu to handle refunds in headers
+        content = await response.aread()
+        content_str = content.decode("utf-8") if isinstance(content, bytes) else content
+
+        is_streaming_content = ResponsesApiProcessor.is_streaming_response(content_str)
+
+        if is_streaming_content:
+            usage_data, model = (
+                ResponsesApiProcessor.extract_usage_with_reasoning_tokens(content_str)
+            )
+
+            refund_token = None
+            if usage_data and model:
+                try:
+                    response_data = {"usage": usage_data, "model": model}
+                    cost_data = await get_x_cashu_cost_func(
+                        response_data, max_cost_for_model
+                    )
+
+                    if cost_data:
+                        refund_amount = self._calculate_refund_amount(
+                            amount, unit, cost_data.total_msats
+                        )
+
+                        if refund_amount > 0:
+                            refund_token = await self.send_refund(
+                                refund_amount, unit, mint
+                            )
+                except Exception as e:
+                    logger.error(
+                        "Error calculating refund for buffered Responses API response",
+                        extra={"error": str(e)},
+                    )
+
+            response_headers = ResponsesApiProcessor.clean_response_headers(
+                dict(response.headers)
+            )
+
+            if refund_token:
+                response_headers["X-Cashu"] = refund_token
+
+            lines = content_str.strip().split("\n")
+            return StreamingResponse(
+                ResponsesApiProcessor.create_streaming_generator(lines),
+                status_code=response.status_code,
+                headers=response_headers,
+                media_type="text/event-stream",
+            )
+        else:
+            return await self._handle_non_streaming_responses_api_response(
+                content_str,
+                response,
+                amount,
+                unit,
+                max_cost_for_model,
+                mint,
+                get_x_cashu_cost_func,
+            )
+
+    async def _handle_non_streaming_responses_api_response(
+        self,
+        content_str: str,
+        response: httpx.Response,
+        amount: int,
+        unit: str,
+        max_cost_for_model: int,
+        mint: str | None,
+        get_x_cashu_cost_func: Callable,
+    ) -> Response:
+        """Handle non-streaming Responses API response with refund calculation."""
+        response_headers = ResponsesApiProcessor.clean_response_headers(
+            dict(response.headers)
+        )
+
+        try:
+            response_json = json.loads(content_str)
+            cost_data = await get_x_cashu_cost_func(response_json, max_cost_for_model)
+
+            if cost_data:
+                refund_amount = self._calculate_refund_amount(
+                    amount, unit, cost_data.total_msats
+                )
+
+                if refund_amount > 0:
+                    refund_token = await self.send_refund(refund_amount, unit, mint)
+                    response_headers["X-Cashu"] = refund_token
+
+            return Response(
+                content=content_str,
+                status_code=response.status_code,
+                headers=response_headers,
+                media_type="application/json",
+            )
+
+        except json.JSONDecodeError:
+            # Emergency refund on parse error
+            emergency_refund = amount
+            refund_token = await self.send_refund(
+                emergency_refund, unit=unit, mint=mint
+            )
+            response_headers["X-Cashu"] = refund_token
+
+            logger.warning(
+                "Emergency refund issued due to JSON parse error in Responses API",
+                extra={
+                    "original_amount": amount,
+                    "refund_amount": emergency_refund,
+                    "deduction": 60,
+                },
+            )
+
+            return Response(
+                content=content_str,
+                status_code=response.status_code,
+                headers=response_headers,
+                media_type="application/json",
+            )
+
+    def _create_default_streaming_response(
+        self, response: httpx.Response
+    ) -> StreamingResponse:
+        """Create default streaming response for non-chat endpoints."""
+        background_tasks = BackgroundTasks()
+        background_tasks.add_task(response.aclose)
+
+        return StreamingResponse(
+            response.aiter_bytes(),
+            status_code=response.status_code,
+            headers=dict(response.headers),
+            background=background_tasks,
+        )

routstr/upstream/helpers.py

@@ -1,6 +1,5 @@
 from __future__ import annotations
 
-import asyncio
 import os
 import re
 from typing import TYPE_CHECKING
@@ -8,7 +7,6 @@ from typing import TYPE_CHECKING
 if TYPE_CHECKING:
     from ..core.settings import Settings
 
-from sqlmodel import select
 
 from ..core import get_logger
 from ..core.db import AsyncSession, ModelRow, UpstreamProviderRow, create_session
@@ -147,17 +145,6 @@ async def refresh_upstreams_models_periodically(
                         f"Error refreshing models for {upstream.base_url}",
                         extra={"error": str(e), "error_type": type(e).__name__},
                     )
-
-            try:
-                from ..payment.models import _update_sats_pricing_once
-
-                await _update_sats_pricing_once()
-            except Exception as e:
-                logger.warning(f"Failed to update pricing after model refresh: {e}")
-                from ..proxy import refresh_model_maps
-
-                await refresh_model_maps()
-
         except asyncio.CancelledError:
             break
         except Exception as e:
@@ -179,6 +166,8 @@ async def init_upstreams() -> list[BaseUpstreamProvider]:
     Seeds database with providers from settings if empty, then loads and instantiates
     provider instances from database records, and refreshes their models cache.
     """
+    from sqlmodel import select
+
     from ..core.settings import settings
 
     async with create_session() as session:
@@ -194,16 +183,16 @@ async def init_upstreams() -> list[BaseUpstreamProvider]:
             result = await session.exec(select(UpstreamProviderRow))
             existing_providers = result.all()
 
-        async def _init_single_provider(
-            provider_row: UpstreamProviderRow,
-        ) -> BaseUpstreamProvider | None:
+        upstreams: list[BaseUpstreamProvider] = []
+        for provider_row in existing_providers:
             if not provider_row.enabled:
                 logger.debug(f"Skipping disabled provider: {provider_row.base_url}")
-                return None
+                continue
 
             provider = _instantiate_provider(provider_row)
             if provider:
                 await provider.refresh_models_cache()
+                upstreams.append(provider)
                 logger.debug(
                     f"Initialized {provider_row.provider_type} provider",
                     extra={
@@ -211,12 +200,6 @@ async def init_upstreams() -> list[BaseUpstreamProvider]:
                         "models_cached": len(provider.get_cached_models()),
                     },
                 )
-                return provider
-            return None
-
-        tasks = [_init_single_provider(row) for row in existing_providers]
-        results = await asyncio.gather(*tasks)
-        upstreams = [p for p in results if p is not None]
 
         return upstreams
 

routstr/upstream/openrouter.py

@@ -50,13 +50,7 @@ class OpenRouterUpstreamProvider(BaseUpstreamProvider):
     async def fetch_models(self) -> list[Model]:
         """Fetch all OpenRouter models."""
         models_data = await async_fetch_openrouter_models()
-        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
+        return [Model(**model) for model in models_data]  # type: ignore
 
     async def get_balance(self) -> float | None:
         """Get the current account balance from OpenRouter.

routstr/upstream/payment/cashu_handler.py

@@ -0,0 +1,111 @@
+"""X-Cashu payment handling utilities."""
+
+from __future__ import annotations
+
+from typing import TYPE_CHECKING
+
+from fastapi import Request
+from fastapi.responses import Response
+
+from ...core import get_logger
+from ...payment.helpers import create_error_response
+from ...wallet import recieve_token
+
+if TYPE_CHECKING:
+    pass
+
+logger = get_logger(__name__)
+
+
+class CashuTokenError(Exception):
+    """Exception raised for X-Cashu token processing errors."""
+
+    def __init__(self, error_type: str, message: str, status_code: int = 400):
+        self.error_type = error_type
+        self.message = message
+        self.status_code = status_code
+        super().__init__(message)
+
+
+async def validate_and_redeem_token(
+    x_cashu_token: str, request: Request
+) -> tuple[int, str, str | None]:
+    """Validate and redeem X-Cashu token.
+
+    Args:
+        x_cashu_token: The X-Cashu token to validate and redeem
+        request: Original FastAPI request (for error responses)
+
+    Returns:
+        Tuple of (amount, unit, mint_url)
+
+    Raises:
+        CashuTokenError: If token validation or redemption fails
+    """
+    logger.info(
+        "Processing X-Cashu token redemption",
+        extra={
+            "token_preview": x_cashu_token[:20] + "..."
+            if len(x_cashu_token) > 20
+            else x_cashu_token,
+        },
+    )
+
+    try:
+        amount, unit, mint = await recieve_token(x_cashu_token)
+        logger.info(
+            "X-Cashu token redeemed successfully",
+            extra={"amount": amount, "unit": unit, "mint": mint},
+        )
+        return amount, unit, mint
+    except Exception as e:
+        error_message = str(e)
+        logger.error(
+            "X-Cashu token redemption failed",
+            extra={
+                "error": error_message,
+                "error_type": type(e).__name__,
+            },
+        )
+
+        # Determine specific error type
+        if "already spent" in error_message.lower():
+            raise CashuTokenError(
+                "token_already_spent",
+                "The provided CASHU token has already been spent",
+                400,
+            )
+        elif "invalid token" in error_message.lower():
+            raise CashuTokenError(
+                "invalid_token", "The provided CASHU token is invalid", 400
+            )
+        elif "mint error" in error_message.lower():
+            raise CashuTokenError(
+                "mint_error", f"CASHU mint error: {error_message}", 422
+            )
+        else:
+            raise CashuTokenError(
+                "cashu_error", f"CASHU token processing failed: {error_message}", 400
+            )
+
+
+def create_cashu_error_response(
+    error: CashuTokenError, request: Request, x_cashu_token: str
+) -> Response:
+    """Create error response for X-Cashu token errors.
+
+    Args:
+        error: The CashuTokenError that occurred
+        request: Original FastAPI request
+        x_cashu_token: The token that caused the error
+
+    Returns:
+        Error response with appropriate status code and message
+    """
+    return create_error_response(
+        error.error_type,
+        error.message,
+        error.status_code,
+        request=request,
+        token=x_cashu_token,
+    )

routstr/upstream/processing/http_client.py

@@ -0,0 +1,200 @@
+"""HTTP client utilities for upstream requests."""
+
+from __future__ import annotations
+
+import traceback
+from typing import TYPE_CHECKING, Callable, Mapping
+
+import httpx
+from fastapi import BackgroundTasks, Request
+from fastapi.responses import StreamingResponse
+
+from ...core import get_logger
+
+if TYPE_CHECKING:
+    from ...payment.models import Model
+
+logger = get_logger(__name__)
+
+
+class HttpForwarder:
+    """Handles HTTP request forwarding to upstream services."""
+
+    def __init__(self, base_url: str):
+        self.base_url = base_url
+
+    def _prepare_path(self, path: str) -> str:
+        """Prepare path by removing v1/ prefix if present."""
+        if path.startswith("v1/"):
+            path = path.replace("v1/", "")
+        return path
+
+    async def forward_request(
+        self,
+        request: Request,
+        path: str,
+        headers: dict,
+        request_body: bytes | None,
+        query_params: Mapping[str, str] | None,
+        transform_body_func: Callable | None = None,
+        model_obj: Model | None = None,
+    ) -> httpx.Response:
+        """Forward HTTP request to upstream service.
+
+        Args:
+            request: Original FastAPI request
+            path: Request path
+            headers: Prepared headers for upstream
+            request_body: Request body bytes, if any
+            query_params: Query parameters for the request
+            transform_body_func: Optional function to transform request body
+            model_obj: Model object for request body transformation
+
+        Returns:
+            Response from upstream service
+
+        Raises:
+            httpx.RequestError: If request fails
+            Exception: For other unexpected errors
+        """
+        path = self._prepare_path(path)
+        url = f"{self.base_url}/{path}"
+
+        # Transform body if function provided
+        transformed_body = request_body
+        if transform_body_func and request_body and model_obj:
+            transformed_body = transform_body_func(request_body, model_obj)
+
+        logger.info(
+            "Forwarding request to upstream",
+            extra={
+                "url": url,
+                "method": request.method,
+                "path": path,
+                "has_request_body": request_body is not None,
+            },
+        )
+
+        client = httpx.AsyncClient(
+            transport=httpx.AsyncHTTPTransport(retries=1),
+            timeout=None,
+        )
+
+        try:
+            if transformed_body is not None:
+                response = await client.send(
+                    client.build_request(
+                        request.method,
+                        url,
+                        headers=headers,
+                        content=transformed_body,
+                        params=query_params,
+                    ),
+                    stream=True,
+                )
+            else:
+                response = await client.send(
+                    client.build_request(
+                        request.method,
+                        url,
+                        headers=headers,
+                        content=request.stream(),
+                        params=query_params,
+                    ),
+                    stream=True,
+                )
+
+            logger.info(
+                "Received upstream response",
+                extra={
+                    "status_code": response.status_code,
+                    "path": path,
+                    "content_type": response.headers.get("content-type", "unknown"),
+                },
+            )
+
+            return response
+
+        except httpx.RequestError as exc:
+            await client.aclose()
+            error_type = type(exc).__name__
+            error_details = str(exc)
+
+            logger.error(
+                "HTTP request error to upstream",
+                extra={
+                    "error_type": error_type,
+                    "error_details": error_details,
+                    "method": request.method,
+                    "url": url,
+                    "path": path,
+                    "query_params": dict(request.query_params),
+                },
+            )
+
+            if isinstance(exc, httpx.ConnectError):
+                error_message = "Unable to connect to upstream service"
+            elif isinstance(exc, httpx.TimeoutException):
+                error_message = "Upstream service request timed out"
+            elif isinstance(exc, httpx.NetworkError):
+                error_message = "Network error while connecting to upstream service"
+            else:
+                error_message = f"Error connecting to upstream service: {error_type}"
+
+            raise httpx.RequestError(error_message) from exc
+
+        except Exception as exc:
+            await client.aclose()
+            tb = traceback.format_exc()
+
+            logger.error(
+                "Unexpected error in upstream forwarding",
+                extra={
+                    "error": str(exc),
+                    "error_type": type(exc).__name__,
+                    "method": request.method,
+                    "url": url,
+                    "path": path,
+                    "query_params": dict(request.query_params),
+                    "traceback": tb,
+                },
+            )
+
+            raise
+
+
+class StreamingResponseWrapper:
+    """Wrapper for creating streaming responses with proper cleanup."""
+
+    @staticmethod
+    def create_streaming_response(
+        response: httpx.Response,
+        client: httpx.AsyncClient,
+    ) -> StreamingResponse:
+        """Create a streaming response with background cleanup tasks.
+
+        Args:
+            response: httpx response to stream
+            client: httpx client to clean up
+
+        Returns:
+            StreamingResponse with background cleanup
+        """
+        background_tasks = BackgroundTasks()
+        background_tasks.add_task(response.aclose)
+        background_tasks.add_task(client.aclose)
+
+        logger.debug(
+            "Creating streaming response",
+            extra={
+                "status_code": response.status_code,
+                "content_type": response.headers.get("content-type", "unknown"),
+            },
+        )
+
+        return StreamingResponse(
+            response.aiter_bytes(),
+            status_code=response.status_code,
+            headers=dict(response.headers),
+            background=background_tasks,
+        )

routstr/upstream/processing/response_handler.py

@@ -0,0 +1,170 @@
+"""Response processing utilities for different API types."""
+
+from __future__ import annotations
+
+import json
+import re
+from collections.abc import AsyncGenerator
+from typing import TYPE_CHECKING
+
+from ...core import get_logger
+
+if TYPE_CHECKING:
+    pass
+
+logger = get_logger(__name__)
+
+
+class ResponseProcessor:
+    """Processes responses from upstream services."""
+
+    @staticmethod
+    def is_streaming_response(content_str: str) -> bool:
+        """Determine if response content indicates streaming.
+
+        Args:
+            content_str: Response content as string
+
+        Returns:
+            True if response is streaming, False otherwise
+        """
+        return content_str.startswith("data:") or "data:" in content_str
+
+    @staticmethod
+    def extract_usage_from_streaming(
+        content_str: str,
+    ) -> tuple[dict | None, str | None]:
+        """Extract usage data and model from streaming response content.
+
+        Args:
+            content_str: Streaming response content as string
+
+        Returns:
+            Tuple of (usage_data, model) or (None, None) if not found
+        """
+        usage_data = None
+        model = None
+
+        lines = content_str.strip().split("\n")
+        for line in lines:
+            if line.startswith("data: "):
+                try:
+                    data_json = json.loads(line[6:])
+                    if "usage" in data_json:
+                        usage_data = data_json["usage"]
+                        model = data_json.get("model")
+                    elif "model" in data_json and not model:
+                        model = data_json["model"]
+                except json.JSONDecodeError:
+                    continue
+
+        return usage_data, model
+
+    @staticmethod
+    def clean_response_headers(headers: dict) -> dict:
+        """Clean response headers by removing encoding-related headers.
+
+        Args:
+            headers: Original response headers
+
+        Returns:
+            Cleaned headers dict
+        """
+        cleaned_headers = dict(headers)
+        headers_to_remove = ["transfer-encoding", "content-encoding", "content-length"]
+
+        for header in headers_to_remove:
+            cleaned_headers.pop(header, None)
+
+        return cleaned_headers
+
+    @staticmethod
+    async def create_streaming_generator(
+        lines: list[str],
+    ) -> AsyncGenerator[bytes, None]:
+        """Create async generator for streaming response lines.
+
+        Args:
+            lines: List of response lines to stream
+
+        Yields:
+            Encoded response lines
+        """
+        for line in lines:
+            yield (line + "\n").encode("utf-8")
+
+
+class ChatCompletionProcessor(ResponseProcessor):
+    """Processor specifically for chat completion responses."""
+
+    @staticmethod
+    def extract_model_from_chunks(stored_chunks: list[bytes]) -> str | None:
+        """Extract model name from stored response chunks.
+
+        Args:
+            stored_chunks: List of response chunks from streaming
+
+        Returns:
+            Model name if found, None otherwise
+        """
+        last_model_seen = None
+
+        for i in range(len(stored_chunks) - 1, -1, -1):
+            chunk = stored_chunks[i]
+            if not chunk:
+                continue
+            try:
+                events = re.split(b"data: ", chunk)
+                for event_data in events:
+                    if not event_data or event_data.strip() in (b"[DONE]", b""):
+                        continue
+                    try:
+                        data = json.loads(event_data)
+                        if isinstance(data, dict) and data.get("model"):
+                            return str(data.get("model"))
+                    except json.JSONDecodeError:
+                        continue
+            except Exception as e:
+                logger.debug(
+                    "Error processing chunk for model extraction",
+                    extra={"error": str(e), "error_type": type(e).__name__},
+                )
+
+        return last_model_seen
+
+
+class ResponsesApiProcessor(ResponseProcessor):
+    """Processor specifically for Responses API responses."""
+
+    @staticmethod
+    def extract_usage_with_reasoning_tokens(
+        content_str: str,
+    ) -> tuple[dict | None, str | None]:
+        """Extract usage data including reasoning tokens from Responses API content.
+
+        Args:
+            content_str: Streaming response content as string
+
+        Returns:
+            Tuple of (usage_data, model) with reasoning tokens preserved
+        """
+        usage_data = None
+        model = None
+
+        lines = content_str.strip().split("\n")
+        for line in lines:
+            if line.startswith("data: "):
+                try:
+                    data_json = json.loads(line[6:])
+                    if "usage" in data_json:
+                        usage_data = data_json["usage"]
+                        model = data_json.get("model")
+                        # Note: We now only track input_tokens and output_tokens
+                        # reasoning_tokens are ignored per the requirement
+                        break
+                    elif "model" in data_json and not model:
+                        model = data_json["model"]
+                except json.JSONDecodeError:
+                    continue
+
+        return usage_data, model

tests/integration/test_embeddings.py

@@ -1,97 +0,0 @@
-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

tests/integration/test_performance_load.py

@@ -9,7 +9,6 @@ import gc
 import statistics
 import time
 from typing import Any, Dict, List
-from unittest.mock import patch
 
 import psutil
 import pytest
@@ -106,46 +105,42 @@ class TestPerformanceBaseline:
             ("GET", "/v1/wallet/info", authenticated_client, None),
         ]
 
-        # Enable provider discovery for this test
-        with patch(
-            "routstr.core.settings.settings.providers_refresh_interval_seconds", 300
-        ):
-            # Warm up
-            for _ in range(10):
-                await integration_client.get("/")
+        # Warm up
+        for _ in range(10):
+            await integration_client.get("/")
 
-            # Test each endpoint
-            for method, path, client, data in endpoints:
-                response_times = []
+        # Test each endpoint
+        for method, path, client, data in endpoints:
+            response_times = []
 
-                for i in range(100):
-                    start = time.time()
+            for i in range(100):
+                start = time.time()
 
-                    if method == "GET":
-                        response = await client.get(path)
-                    else:
-                        response = await client.post(path, json=data)
+                if method == "GET":
+                    response = await client.get(path)
+                else:
+                    response = await client.post(path, json=data)
 
-                    duration = time.time() - start
-                    response_times.append(duration * 1000)  # Convert to ms
+                duration = time.time() - start
+                response_times.append(duration * 1000)  # Convert to ms
 
-                    assert response.status_code in [200, 201]
+                assert response.status_code in [200, 201]
 
-                    if i % 10 == 0:
-                        metrics.record_system_metrics()
+                if i % 10 == 0:
+                    metrics.record_system_metrics()
 
-                # Verify 95th percentile < 500ms
-                p95 = sorted(response_times)[int(len(response_times) * 0.95)]
-                assert p95 < 500, (
-                    f"{method} {path} p95 response time {p95}ms exceeds 500ms limit"
-                )
+            # Verify 95th percentile < 500ms
+            p95 = sorted(response_times)[int(len(response_times) * 0.95)]
+            assert p95 < 500, (
+                f"{method} {path} p95 response time {p95}ms exceeds 500ms limit"
+            )
 
-                print(f"\n{method} {path}:")
-                print(f"  Mean: {statistics.mean(response_times):.2f}ms")
-                print(f"  P95: {p95:.2f}ms")
-                print(
-                    f"  P99: {sorted(response_times)[int(len(response_times) * 0.99)]:.2f}ms"
-                )
+            print(f"\n{method} {path}:")
+            print(f"  Mean: {statistics.mean(response_times):.2f}ms")
+            print(f"  P95: {p95:.2f}ms")
+            print(
+                f"  P99: {sorted(response_times)[int(len(response_times) * 0.99)]:.2f}ms"
+            )
 
 
 @pytest.mark.integration

tests/integration/test_provider_management.py

@@ -3,7 +3,7 @@ Integration tests for provider management functionality.
 Tests GET /v1/providers/ endpoint for listing and managing providers.
 """
 
-from typing import Any, Generator
+from typing import Any
 from unittest.mock import patch
 
 import pytest
@@ -19,15 +19,6 @@ def _clear_providers_cache() -> None:
     _PROVIDERS_CACHE.clear()
 
 
-@pytest.fixture(autouse=True)
-def _enable_provider_discovery() -> Generator[None, Any, Any]:
-    """Enable provider discovery for all tests in this module"""
-    with patch(
-        "routstr.core.settings.settings.providers_refresh_interval_seconds", 300
-    ):
-        yield
-
-
 @pytest.mark.integration
 @pytest.mark.asyncio
 async def test_providers_endpoint_default_response(

tests/integration/test_wallet_authentication.py

@@ -3,6 +3,7 @@ Integration tests for wallet authentication system including API key generation
 Tests POST /v1/wallet/topup endpoint and authorization header validation.
 """
 
+import hashlib
 from datetime import datetime, timedelta
 from typing import Any
 
@@ -14,6 +15,7 @@ from routstr.core.db import ApiKey
 
 from .utils import (
     CashuTokenGenerator,
+    ConcurrencyTester,
     ResponseValidator,
 )
 
@@ -387,6 +389,69 @@ async def test_api_key_with_expiry_time(
     # The expiry time and refund address functionality is tested elsewhere
 
 
+@pytest.mark.integration
+@pytest.mark.asyncio
+async def test_concurrent_token_submissions(
+    integration_client: AsyncClient, testmint_wallet: Any, integration_session: Any
+) -> None:
+    """Test concurrent submissions of different tokens"""
+
+    # Generate multiple unique tokens with known amounts
+    num_tokens = 10
+    tokens = []
+    expected_balances = {}
+
+    for i in range(num_tokens):
+        amount = 100 + i * 10
+        token = await testmint_wallet.mint_tokens(amount)
+        tokens.append(token)
+        # Store expected balance by token hash
+        hashed_key = hashlib.sha256(token.encode()).hexdigest()
+        expected_balances[hashed_key] = amount * 1000  # msats
+
+    # Create concurrent requests
+    requests = [
+        {
+            "method": "GET",
+            "url": "/v1/wallet/info",
+            "headers": {"Authorization": f"Bearer {token}"},
+        }
+        for token in tokens
+    ]
+
+    # Execute concurrently
+    tester = ConcurrencyTester()
+    responses = await tester.run_concurrent_requests(
+        integration_client, requests, max_concurrent=5
+    )
+
+    # All should succeed
+    assert len(responses) == num_tokens
+    api_keys = set()
+
+    for response in responses:
+        assert response.status_code == 200
+        data = response.json()
+        api_key = data["api_key"]
+        api_keys.add(api_key)
+
+        # Verify balance matches the expected amount
+        hashed_key = api_key[3:]  # Remove "sk-" prefix
+        assert data["balance"] == expected_balances[hashed_key]
+
+    # Should have created unique API keys
+    assert len(api_keys) == num_tokens
+
+    # Verify all keys exist in database
+    for api_key in api_keys:
+        hashed_key = api_key[3:]  # Remove "sk-" prefix
+        result = await integration_session.execute(
+            select(ApiKey).where(ApiKey.hashed_key == hashed_key)  # type: ignore[arg-type]
+        )
+        db_key = result.scalar_one()
+        assert db_key.balance == expected_balances[hashed_key]
+
+
 @pytest.mark.integration
 @pytest.mark.asyncio
 async def test_authorization_with_cashu_token_directly(
@@ -439,6 +504,48 @@ async def test_x_cashu_header_support(
     assert response.status_code == 200
 
 
+@pytest.mark.integration
+@pytest.mark.asyncio
+@pytest.mark.slow
+async def test_api_key_consistency_under_load(
+    integration_client: AsyncClient, testmint_wallet: Any, integration_session: Any
+) -> None:
+    """Test API key generation consistency under concurrent load"""
+
+    # Generate a single token
+    token = await testmint_wallet.mint_tokens(1000)
+
+    # First request to create the API key
+    integration_client.headers["Authorization"] = f"Bearer {token}"
+    initial_response = await integration_client.get("/v1/wallet/info")
+    assert initial_response.status_code == 200
+    expected_api_key = initial_response.json()["api_key"]
+    expected_balance = initial_response.json()["balance"]
+
+    # Try to use the same token concurrently multiple times
+    # All should return the same API key since it's already created
+    requests = [
+        {
+            "method": "GET",
+            "url": "/v1/wallet/info",
+            "headers": {"Authorization": f"Bearer {token}"},
+        }
+        for _ in range(20)  # 20 concurrent attempts
+    ]
+
+    tester = ConcurrencyTester()
+    responses = await tester.run_concurrent_requests(
+        integration_client, requests, max_concurrent=10
+    )
+
+    # All should succeed and return the same API key
+    for response in responses:
+        assert response.status_code == 200
+        data = response.json()
+        assert data["api_key"] == expected_api_key
+        assert data["balance"] == expected_balance
+
+
 @pytest.mark.integration
 @pytest.mark.asyncio
 async def test_database_timestamp_accuracy(

tests/integration/test_wallet_information.py

@@ -13,7 +13,7 @@ from sqlmodel import select, update
 
 from routstr.core.db import ApiKey
 
-from .utils import ResponseValidator
+from .utils import ConcurrencyTester, ResponseValidator
 
 
 @pytest.mark.integration
@@ -204,6 +204,45 @@ async def test_expired_api_key_behavior(
     assert db_key.refund_address == "test@lightning.address"
 
 
+@pytest.mark.integration
+@pytest.mark.asyncio
+async def test_concurrent_access_same_api_key(
+    integration_client: AsyncClient, authenticated_client: AsyncClient
+) -> None:
+    """Test concurrent access with the same API key"""
+
+    # Get the API key from authenticated client
+    response = await authenticated_client.get("/v1/wallet/")
+    api_key = response.json()["api_key"]
+    initial_balance = response.json()["balance"]
+
+    # Create multiple concurrent requests
+    requests = []
+    for i in range(20):
+        # Alternate between both endpoints
+        endpoint = "/v1/wallet/" if i % 2 == 0 else "/v1/wallet/info"
+        requests.append(
+            {
+                "method": "GET",
+                "url": endpoint,
+                "headers": {"Authorization": f"Bearer {api_key}"},
+            }
+        )
+
+    # Execute concurrently
+    tester = ConcurrencyTester()
+    responses = await tester.run_concurrent_requests(
+        integration_client, requests, max_concurrent=10
+    )
+
+    # All should succeed with consistent data
+    for response in responses:
+        assert response.status_code == 200
+        data = response.json()
+        assert data["api_key"] == api_key
+        assert data["balance"] == initial_balance
+
+
 @pytest.mark.integration
 @pytest.mark.asyncio
 async def test_wallet_info_data_consistency(

tests/integration/test_wallet_topup.py

@@ -15,6 +15,7 @@ from routstr.core.db import ApiKey
 
 from .utils import (
     CashuTokenGenerator,
+    ConcurrencyTester,
     ResponseValidator,
 )
 
@@ -283,6 +284,60 @@ async def test_transaction_history_tracking(  # type: ignore[no-untyped-def]
     assert response.status_code == 400
 
 
+@pytest.mark.integration
+@pytest.mark.asyncio
+async def test_concurrent_topups_same_api_key(  # type: ignore[no-untyped-def]
+    integration_client: AsyncClient,
+    authenticated_client: AsyncClient,
+    testmint_wallet: Any,
+) -> None:
+    """Test concurrent top-ups to the same API key"""
+
+    # Get API key
+    response = await authenticated_client.get("/v1/wallet/")
+    api_key = response.json()["api_key"]
+    initial_balance = response.json()["balance"]
+
+    # Generate multiple unique tokens
+    num_tokens = 10
+    tokens = []
+    total_amount = 0
+
+    for i in range(num_tokens):
+        amount = 100 + i * 10  # Different amounts
+        token = await testmint_wallet.mint_tokens(amount)
+        tokens.append(token)
+        total_amount += amount
+
+    # Create concurrent top-up requests
+    requests = [
+        {
+            "method": "POST",
+            "url": "/v1/wallet/topup",
+            "params": {"cashu_token": token},
+            "headers": {"Authorization": f"Bearer {api_key}"},
+        }
+        for token in tokens
+    ]
+
+    # Execute concurrently
+    tester = ConcurrencyTester()
+    responses = await tester.run_concurrent_requests(
+        integration_client, requests, max_concurrent=5
+    )
+
+    # All should succeed
+    for response in responses:
+        assert response.status_code == 200
+        assert "msats" in response.json()
+
+    # Verify final balance is correct
+    final_response = await authenticated_client.get("/v1/wallet/")
+    final_balance = final_response.json()["balance"]
+    expected_balance = initial_balance + (total_amount * 1000)
+    assert final_balance == expected_balance
+
+
 @pytest.mark.integration
 @pytest.mark.asyncio
 async def test_topup_during_active_proxy_request(  # type: ignore[no-untyped-def]

ui/components/ModelSearchFilter.tsx

@@ -41,10 +41,6 @@ export function ModelSearchFilter({
           model.name.toLowerCase().includes(query) ||
           model.full_name.toLowerCase().includes(query) ||
           model.provider.toLowerCase().includes(query) ||
-          (model.alias_ids &&
-            model.alias_ids.some((alias) =>
-              alias.toLowerCase().includes(query)
-            )) ||
           (model.description &&
             model.description.toLowerCase().includes(query)) ||
           model.modelType.toLowerCase().includes(query)

ui/components/ModelSelector.tsx

@@ -542,7 +542,6 @@ export function ModelSelector({
       top_provider: null,
       upstream_provider_id: providerId,
       enabled: model.isEnabled,
-      alias_ids: model.alias_ids || null,
     };
 
     setModelDialogState({
@@ -586,7 +585,6 @@ export function ModelSelector({
       top_provider: null,
       upstream_provider_id: providerId,
       enabled: model.isEnabled,
-      alias_ids: model.alias_ids || null,
     };
 
     setModelDialogState({

ui/lib/api/schemas/models.ts

@@ -31,7 +31,6 @@ export const ModelSchema = z.object({
   // API key type indicators
   has_own_api_key: z.boolean(),
   api_key_type: z.string(), // "individual" or "group"
-  alias_ids: z.array(z.string()).nullable().optional(),
 });
 
 // Schema for a model with additional provider-specific settings

ui/lib/api/services/admin.ts

@@ -121,7 +121,6 @@ export interface AdminModelAsModel {
   soft_deleted?: boolean;
   has_own_api_key: boolean;
   api_key_type: string;
-  alias_ids?: string[] | null;
 }
 
 export interface AdminModelGroup {
@@ -208,7 +207,6 @@ export class AdminService {
       soft_deleted: !adminModel.enabled,
       has_own_api_key: false,
       api_key_type: 'group',
-      alias_ids: adminModel.alias_ids,
     };
   }
 
@@ -356,9 +354,8 @@ export class AdminService {
       original: data.pricing,
       converted: payload.pricing,
     });
-    // Use the same POST endpoint for both create and update (upsert)
-    const model = await apiClient.post<AdminModel>(
-      `/admin/api/upstream-providers/${providerId}/models`,
+    const model = await apiClient.patch<AdminModel>(
+      `/admin/api/upstream-providers/${providerId}/models/${encodeURIComponent(modelId)}`,
       payload
     );
     return {