enforce payment finalization

prevent payout race condition

routstr/core/admin.py

@@ -2516,7 +2516,7 @@ async def get_provider_model(provider_id: int, model_id: str) -> dict[str, objec
                 status_code=404, detail="Model not found for this provider"
             )
         return _row_to_model(
-            row, apply_provider_fee=True, provider_fee=provider.provider_fee
+            row, apply_provider_fee=False, provider_fee=provider.provider_fee
         ).dict()  # type: ignore
 
 
@@ -2729,7 +2729,10 @@ async def get_provider_models(provider_id: int) -> dict[str, object]:
             raise HTTPException(status_code=404, detail="Provider not found")
 
         db_models = await list_models(
-            session=session, upstream_id=provider_id, include_disabled=True
+            session=session,
+            upstream_id=provider_id,
+            include_disabled=True,
+            apply_fees=False,
         )
 
         upstream_models = []
@@ -2737,10 +2740,7 @@ async def get_provider_models(provider_id: int) -> dict[str, object]:
         if upstream_instance:
             try:
                 raw_models = await upstream_instance.fetch_models()
-                upstream_models = [
-                    upstream_instance._apply_provider_fee_to_model(m)
-                    for m in raw_models
-                ]
+                upstream_models = raw_models
             except Exception as e:
                 logger.error(
                     f"Failed to fetch models from {provider.provider_type}: {e}"

routstr/payment/lnurl.py

@@ -291,9 +291,7 @@ async def raw_send_to_lnurl(
         lnurl_data["callback_url"], final_amount
     )
 
-    melt_quote_resp = await wallet.melt_quote(
-        invoice=bolt11_invoice, amount_msat=final_amount
-    )
+    melt_quote_resp = await wallet.melt_quote(invoice=bolt11_invoice)
 
     if amount:
         proofs, _ = await wallet.select_to_send(proofs, amount, set_reserved=True)

routstr/payment/models.py

@@ -230,6 +230,7 @@ async def list_models(
     session: AsyncSession,
     upstream_id: int,
     include_disabled: bool = False,
+    apply_fees: bool = True,
 ) -> list[Model]:
     from sqlmodel import select
 
@@ -247,7 +248,7 @@ async def list_models(
     return [
         _row_to_model(
             r,
-            apply_provider_fee=True,
+            apply_provider_fee=apply_fees,
             provider_fee=providers_by_id[r.upstream_provider_id].provider_fee
             if r.upstream_provider_id in providers_by_id
             else 1.01,

routstr/upstream/base.py

@@ -427,7 +427,11 @@ class BaseUpstreamProvider:
         )
 
     async def handle_streaming_chat_completion(
-        self, response: httpx.Response, key: ApiKey, max_cost_for_model: int
+        self,
+        response: httpx.Response,
+        key: ApiKey,
+        max_cost_for_model: int,
+        background_tasks: BackgroundTasks,
     ) -> StreamingResponse:
         """Handle streaming chat completion responses with token usage tracking and cost adjustment.
 
@@ -534,7 +538,14 @@ class BaseUpstreamProvider:
                                 }
                                 yield f"data: {json.dumps(usage_chunk_data)}\n\n".encode()
                                 usage_finalized = True
-                            except Exception:
+                            except Exception as e:
+                                logger.exception(
+                                    "Error during usage finalization",
+                                    extra={
+                                        "key_hash": key.hashed_key[:8] + "...",
+                                        "error": str(e),
+                                    },
+                                )
                                 # Fallback: yield original usage chunk if adjustment fails
                                 yield f"data: {json.dumps(usage_chunk_data)}\n\n".encode()
 
@@ -555,7 +566,9 @@ class BaseUpstreamProvider:
                 raise
             finally:
                 if not usage_finalized:
-                    await finalize_db_only()
+                    # Create a background task to ensure finalization happens
+                    # even if the generator is closed early
+                    background_tasks.add_task(finalize_db_only)
 
         # Remove inaccurate encoding headers from upstream response
         response_headers = dict(response.headers)
@@ -1136,12 +1149,12 @@ class BaseUpstreamProvider:
                     )
 
                     if is_streaming and response.status_code == 200:
-                        result = await self.handle_streaming_chat_completion(
-                            response, key, max_cost_for_model
-                        )
                         background_tasks = BackgroundTasks()
                         background_tasks.add_task(response.aclose)
                         background_tasks.add_task(client.aclose)
+                        result = await self.handle_streaming_chat_completion(
+                            response, key, max_cost_for_model, background_tasks
+                        )
                         result.background = background_tasks
                         return result
 

routstr/wallet.py

@@ -319,6 +319,7 @@ async def periodic_payout() -> None:
                             wallet, mint_url, unit, not_reserved=True
                         )
                         proofs = await slow_filter_spend_proofs(proofs, wallet)
+                        await asyncio.sleep(5)
                         user_balance = await db.balances_for_mint_and_unit(
                             session, mint_url, unit
                         )
@@ -344,8 +345,6 @@ async def periodic_payout() -> None:
                                     "amount_received": amount_received,
                                 },
                             )
-
-                        await asyncio.sleep(5)
         except Exception as e:
             logger.error(
                 f"Error sending payout: {type(e).__name__}",

ui/components/AddProviderModelDialog.tsx

@@ -808,45 +808,6 @@ export function AddProviderModelDialog({
                     </FormItem>
                   )}
                 />
-                <FormField
-                  control={form.control}
-                  name='max_prompt_cost'
-                  render={({ field }) => (
-                    <FormItem>
-                      <FormLabel>Max Prompt Cost</FormLabel>
-                      <FormControl>
-                        <Input type='number' step='0.0001' {...field} />
-                      </FormControl>
-                      <FormMessage />
-                    </FormItem>
-                  )}
-                />
-                <FormField
-                  control={form.control}
-                  name='max_completion_cost'
-                  render={({ field }) => (
-                    <FormItem>
-                      <FormLabel>Max Completion Cost</FormLabel>
-                      <FormControl>
-                        <Input type='number' step='0.0001' {...field} />
-                      </FormControl>
-                      <FormMessage />
-                    </FormItem>
-                  )}
-                />
-                <FormField
-                  control={form.control}
-                  name='max_cost'
-                  render={({ field }) => (
-                    <FormItem>
-                      <FormLabel>Max Total Cost</FormLabel>
-                      <FormControl>
-                        <Input type='number' step='0.0001' {...field} />
-                      </FormControl>
-                      <FormMessage />
-                    </FormItem>
-                  )}
-                />
               </div>
             </div>
 

ui/lib/api/services/admin.ts

@@ -139,18 +139,26 @@ export class AdminService {
   ): Record<string, unknown> {
     if (!pricing) return pricing;
     const result = { ...pricing };
-    if (typeof result.prompt === 'number') {
-      result.prompt = result.prompt * 1000000;
-    }
-    if (typeof result.completion === 'number') {
-      result.completion = result.completion * 1000000;
-    }
-    if (typeof result.request === 'number') {
-      result.request = result.request * 1000000;
-    }
-    if (typeof result.image === 'number') {
-      result.image = result.image * 1000000;
-    }
+
+    // Only prompt and completion are per-token and need scaling to per-1M
+    const convertField = (field: string) => {
+      const val = result[field];
+      if (val !== undefined && val !== null) {
+        const num = typeof val === 'string' ? parseFloat(val) : (val as number);
+        if (!isNaN(num)) {
+          // Multiply by 1M and round to avoid floating point artifacts (e.g. 0.40399999999999997)
+          // 9 decimals is plenty for USD/1M tokens (0.000000001)
+          result[field] = parseFloat((num * 1000000).toFixed(9));
+        }
+      }
+    };
+
+    convertField('prompt');
+    convertField('completion');
+
+    // Other fields (request, image, etc.) are already flat fees (per item)
+    // so we do NOT scale them.
+
     return result;
   }
 
@@ -159,18 +167,23 @@ export class AdminService {
   ): Record<string, unknown> {
     if (!pricing) return pricing;
     const result = { ...pricing };
-    if (typeof result.prompt === 'number') {
-      result.prompt = result.prompt / 1000000;
-    }
-    if (typeof result.completion === 'number') {
-      result.completion = result.completion / 1000000;
-    }
-    if (typeof result.request === 'number') {
-      result.request = result.request / 1000000;
-    }
-    if (typeof result.image === 'number') {
-      result.image = result.image / 1000000;
-    }
+
+    // Only prompt and completion are per-1M in UI and need scaling down to per-token
+    const convertField = (field: string) => {
+      const val = result[field];
+      if (val !== undefined && val !== null) {
+        const num = typeof val === 'string' ? parseFloat(val) : (val as number);
+        if (!isNaN(num)) {
+          result[field] = num / 1000000;
+        }
+      }
+    };
+
+    convertField('prompt');
+    convertField('completion');
+
+    // Other fields stay as flat fees
+
     return result;
   }
 
@@ -291,7 +304,19 @@ export class AdminService {
     const data = await apiClient.get<ProviderModels>(
       `/admin/api/upstream-providers/${providerId}/models`
     );
-    return data;
+
+    // Convert pricing for all models in the list so the UI receives "per 1M tokens" values
+    return {
+      ...data,
+      db_models: data.db_models.map((m) => ({
+        ...m,
+        pricing: this.convertPricingToPerMillionTokens(m.pricing),
+      })),
+      remote_models: data.remote_models.map((m) => ({
+        ...m,
+        pricing: this.convertPricingToPerMillionTokens(m.pricing),
+      })),
+    };
   }
 
   static async createProviderModel(
@@ -498,8 +523,8 @@ export class AdminService {
       : null;
 
     const pricing = {
-      prompt: (data.input_cost as number) / 1000000,
-      completion: (data.output_cost as number) / 1000000,
+      prompt: data.input_cost as number,
+      completion: data.output_cost as number,
       request: (data.min_cost_per_request as number) || 0,
       image: 0,
       web_search: 0,
@@ -553,8 +578,8 @@ export class AdminService {
     const existingModel = await this.getModel(modelId, providerId);
 
     const pricing = {
-      prompt: (data.input_cost as number) / 1000000,
-      completion: (data.output_cost as number) / 1000000,
+      prompt: data.input_cost as number,
+      completion: data.output_cost as number,
       request: (data.min_cost_per_request as number) || 0,
       image: 0,
       web_search: 0,