Merge branch 'fix-x-cashu-misscalculation' into refresh-models

# Conflicts:
#	routstr/core/admin.py

.env.example

@@ -1,43 +1,39 @@
 # Core Configuration
 UPSTREAM_BASE_URL=https://api.openai.com/v1
 UPSTREAM_API_KEY=your-upstream-api-key
-ADMIN_PASSWORD=secure-admin-password
+
+# ADMIN_PASSWORD=secure-admin-password
 
 # Database
-DATABASE_URL=sqlite+aiosqlite:///keys.db
+# DATABASE_URL=sqlite+aiosqlite:///keys.db
 
 # Node Information
-NAME=My Routstr Node
-DESCRIPTION=Fast AI API access with Bitcoin payments
-NPUB=npub1...
-HTTP_URL=https://api.mynode.com
-ONION_URL=http://mynode.onion
+# NAME=My Routstr Node
+# DESCRIPTION=Fast AI API access with Bitcoin payments
+# NSEC=nsec1...
+# HTTP_URL=https://api.mynode.com
+# ONION_URL=http://mynode.onion (auto fetched from compose)
+# RELAYS="wss://relay.damus.io,wss://relay.nostr.band,wss://eden.nostr.land,wss://relay.routstr.com"
+# CASHU_MINTS="https://mint.minibits.cash/Bitcoin,https://mint.cubabitcoin.org,https://ecashmint.otrta.me"
+# RECEIVE_LN_ADDRESS=
 
-# Cashu Configuration
-CASHU_MINTS=https://mint.minibits.cash/Bitcoin
-RECEIVE_LN_ADDRESS=
-
-# Pricing Configuration
-MODEL_BASED_PRICING=true
-COST_PER_REQUEST=1
-COST_PER_1K_INPUT_TOKENS=0
-COST_PER_1K_OUTPUT_TOKENS=0
-EXCHANGE_FEE=1.005
-UPSTREAM_PROVIDER_FEE=1.05
+# Custom Pricing Configuration
+# MODEL_BASED_PRICING=true
+# COST_PER_REQUEST=1
+# COST_PER_1K_INPUT_TOKENS=0
+# COST_PER_1K_OUTPUT_TOKENS=0
+# EXCHANGE_FEE=1.005
+# UPSTREAM_PROVIDER_FEE=1.05
 
 # Network Configuration
-CORS_ORIGINS=*
-TOR_PROXY_URL=socks5://127.0.0.1:9050
+# CORS_ORIGINS=*
+# TOR_PROXY_URL=socks5://127.0.0.1:9050
 
 # Logging
-LOG_LEVEL=INFO
-ENABLE_CONSOLE_LOGGING=true
+# LOG_LEVEL=INFO
+# ENABLE_CONSOLE_LOGGING=true
 
-# Model Management
-MODELS_PATH=models.json
-BASE_URL=https://openrouter.ai/api/v1
-SOURCE=
-
-# Optional Features
-PREPAID_API_KEY=
-PREPAID_BALANCE=0
+# Custom Model Management
+# BASE_URL=https://openrouter.ai/api/v1
+# MODELS_PATH=models.json
+# SOURCE=

README.md

@@ -91,7 +91,7 @@ The most common settings are shown below. See `.env.example` for the full list.
 
 - `UPSTREAM_BASE_URL` – URL of the OpenAI-compatible service
 - `UPSTREAM_API_KEY` – API key for the upstream service (optional)
-- `MODEL_BASED_PRICING` – Set to `true` to use pricing from `models.json`
+- `FIXED_PRICING` – Set to `true` to use a fixed per-request price; `false` (default) uses model pricing from `models.json`
 - `ADMIN_PASSWORD` – Password for the `/admin/` dashboard
 - `CASHU_MINTS` – Comma-separated list of Cashu mint URLs
 - `NAME` – Name of the proxy

compose.testing.yml

@@ -19,10 +19,10 @@ services:
       - "ONION_URL=http://test.onion"
       - "CORS_ORIGINS=*"
       - "RECEIVE_LN_ADDRESS=test@routstr.com"
-      - "COST_PER_REQUEST=10"
-      - "COST_PER_1K_INPUT_TOKENS=0"
-      - "COST_PER_1K_OUTPUT_TOKENS=0"
-      - "MODEL_BASED_PRICING=true"
+      - "FIXED_COST_PER_REQUEST=10"
+      - "FIXED_PER_1K_INPUT_TOKENS=0"
+      - "FIXED_PER_1K_OUTPUT_TOKENS=0"
+      - "FIXED_PRICING=false"
       - "NSEC=nsec1testkey1234567890abcdef"
       - "REFUND_PROCESSING_INTERVAL=3600"
       - "MINIMUM_PAYOUT=1000"

compose.yml

@@ -1,5 +1,3 @@
-version: '3.8'
-
 services:
   routstr:
     build: .
@@ -26,12 +24,5 @@ services:
     depends_on:
       - routstr
 
-  # Legacy service definition to ensure cleanup of old container
-  router:
-    image: alpine:latest
-    command: /bin/true
-    profiles:
-      - cleanup
-
 volumes:
   tor-data:

docs/advanced/custom-pricing.md

@@ -14,11 +14,11 @@ Routstr supports three pricing models:
 
 ### Configuration
 
-Enable model-based pricing:
+Enable model-based pricing (default behavior):
 
 ```bash
 # .env
-MODEL_BASED_PRICING=true
+FIXED_PRICING=false
 MODELS_PATH=/app/config/models.json
 EXCHANGE_FEE=1.005         # 0.5% exchange fee
 UPSTREAM_PROVIDER_FEE=1.05  # 5% provider margin
@@ -118,14 +118,14 @@ if __name__ == "__main__":
 
 ### Configuration
 
-Set up token-based pricing:
+Set up token-based pricing overrides:
 
 ```bash
 # .env
-MODEL_BASED_PRICING=false
-COST_PER_REQUEST=1              # 1 sat base fee
-COST_PER_1K_INPUT_TOKENS=5      # 5 sats per 1K input
-COST_PER_1K_OUTPUT_TOKENS=15    # 15 sats per 1K output
+FIXED_PRICING=false                 # use model pricing
+FIXED_COST_PER_REQUEST=1            # optional base fee
+FIXED_PER_1K_INPUT_TOKENS=5         # optional override
+FIXED_PER_1K_OUTPUT_TOKENS=15       # optional override
 ```
 
 ### Custom Token Counting

docs/advanced/nostr.md

@@ -102,7 +102,7 @@ Configure which relays to publish to:
 DEFAULT_RELAYS = [
     "wss://relay.damus.io",
     "wss://relay.nostr.band",
-    "wss://nostr.mom",
+    "wss://relay.routstr.com",
     "wss://nos.lol"
 ]
 

docs/api/endpoints.md

@@ -436,6 +436,27 @@ Authorization: Bearer sk-...
 
 ## Provider Discovery
 
+## Admin Settings
+
+These endpoints are protected by the Admin cookie (`admin_password` set to your configured admin password).
+
+### Get Settings
+
+```http
+GET /admin/api/settings
+```
+
+Returns the current application settings (sensitive values may be redacted).
+
+### Update Settings
+
+```http
+PATCH /admin/api/settings
+Content-Type: application/json
+```
+
+Body is a partial JSON of settings fields to update. Validated and persisted to the database.
+
 ### List Providers
 
 Get available upstream providers.

docs/api/overview.md

@@ -347,7 +347,7 @@ GET /health
 Response:
 {
   "status": "healthy",
-  "version": "0.1.2",
+  "version": "0.1.4",
   "timestamp": "2024-01-01T00:00:00Z",
   "checks": {
     "database": "ok",

docs/contributing/code-structure.md

@@ -348,7 +348,7 @@ Project metadata and dependencies:
 ```toml
 [project]
 name = "routstr"
-version = "0.1.2"
+version = "0.1.4"
 dependencies = [
     "fastapi[standard]>=0.115",
     "sqlmodel>=0.0.24",

docs/getting-started/configuration.md

@@ -1,6 +1,6 @@
 # Configuration
 
-Routstr Core is configured through environment variables. This guide covers all available options.
+Routstr Core is configured via a single settings row in the database. Environment variables are only used on first run to seed that row (with a few computed defaults like `ONION_URL`). After that, the database is the source of truth. You can update settings at runtime via the admin API. `DATABASE_URL` is always env-only.
 
 ## Environment Variables
 
@@ -33,19 +33,21 @@ Routstr Core is configured through environment variables. This guide covers all
 
 | Variable | Description | Default | Required |
 |----------|-------------|---------|----------|
-| `MODEL_BASED_PRICING` | Enable model-specific pricing from models.json | `false` | ❌ |
-| `COST_PER_REQUEST` | Fixed cost per API request in sats | `1` | ❌ |
-| `COST_PER_1K_INPUT_TOKENS` | Cost per 1000 input tokens in sats | `0` | ❌ |
-| `COST_PER_1K_OUTPUT_TOKENS` | Cost per 1000 output tokens in sats | `0` | ❌ |
+| `FIXED_PRICING` | Force fixed per-request pricing (ignore model token pricing) | `false` | ❌ |
+| `FIXED_COST_PER_REQUEST` | Fixed cost per API request in sats | `1` | ❌ |
+| `FIXED_PER_1K_INPUT_TOKENS` | Optional override: sats per 1000 input tokens | `0` | ❌ |
+| `FIXED_PER_1K_OUTPUT_TOKENS` | Optional override: sats per 1000 output tokens | `0` | ❌ |
 | `EXCHANGE_FEE` | Exchange rate markup (1.005 = 0.5% fee) | `1.005` | ❌ |
 | `UPSTREAM_PROVIDER_FEE` | Provider fee markup (1.05 = 5% fee) | `1.05` | ❌ |
 
-### Network Configuration
+### Network & Discovery
 
 | Variable | Description | Default | Required |
 |----------|-------------|---------|----------|
 | `CORS_ORIGINS` | Comma-separated list of allowed CORS origins | `*` | ❌ |
 | `TOR_PROXY_URL` | SOCKS5 proxy URL for Tor connections | `socks5://127.0.0.1:9050` | ❌ |
+| `RELAYS` | Comma-separated nostr relays used for provider discovery | sane defaults | ❌ |
+| `PROVIDERS_REFRESH_INTERVAL_SECONDS` | Provider cache refresh interval | `300` | ❌ |
 
 ### Logging Configuration
 
@@ -60,6 +62,7 @@ Routstr Core is configured through environment variables. This guide covers all
 |----------|-------------|---------|----------|
 | `CHAT_COMPLETIONS_API_VERSION` | Append `api-version` to `/chat/completions` (Azure OpenAI) | - | ❌ |
 | `DATABASE_URL` | SQLite database connection string | `sqlite+aiosqlite:///keys.db` | ❌ |
+| `REFUND_CACHE_TTL_SECONDS` | Cache TTL for refund responses (seconds) | `3600` | ❌ |
 
 ## Configuration Examples
 
@@ -78,7 +81,6 @@ ADMIN_PASSWORD=my-secure-password
 # .env
 UPSTREAM_BASE_URL=https://api.anthropic.com/v1
 UPSTREAM_API_KEY=your-anthropic-key
-MODEL_BASED_PRICING=true
 MODELS_PATH=/app/config/anthropic-models.json
 ```
 
@@ -115,36 +117,21 @@ ONION_URL=http://lightningai.onion
 CASHU_MINTS=https://mint1.com,https://mint2.com
 ```
 
-## Pricing Models
+## Pricing
 
-### Fixed Pricing
+- Default: pricing comes from your `models.json`.
+- Force fixed per-request pricing: set `FIXED_PRICING=true` and `FIXED_COST_PER_REQUEST`.
+- Optional token overrides when using model pricing: set
+  `FIXED_PER_1K_INPUT_TOKENS` and/or `FIXED_PER_1K_OUTPUT_TOKENS`.
+- Legacy envs are still accepted and mapped automatically:
+  `MODEL_BASED_PRICING` → `!FIXED_PRICING`, `COST_PER_REQUEST` → `FIXED_COST_PER_REQUEST`,
+  `COST_PER_1K_*` → `FIXED_PER_1K_*`.
 
-Simple per-request pricing:
+Example fixed pricing:
 
 ```bash
-MODEL_BASED_PRICING=false
-COST_PER_REQUEST=10  # 10 sats per request
-```
-
-### Token-Based Pricing
-
-Charge based on token usage:
-
-```bash
-MODEL_BASED_PRICING=false
-COST_PER_REQUEST=1  # 1 sat base fee
-COST_PER_1K_INPUT_TOKENS=5   # 5 sats per 1k input
-COST_PER_1K_OUTPUT_TOKENS=15  # 15 sats per 1k output
-```
-
-### Model-Based Pricing
-
-Use dynamic pricing from models.json:
-
-```bash
-MODEL_BASED_PRICING=true
-EXCHANGE_FEE=1.01  # 1% exchange fee
-UPSTREAM_PROVIDER_FEE=1.00  # No additional markup
+FIXED_PRICING=true
+FIXED_COST_PER_REQUEST=10
 ```
 
 ## Custom Models Configuration

docs/getting-started/docker.md

@@ -115,8 +115,8 @@ NPUB=npub1...
 HTTP_URL=https://api.mynode.com
 ONION_URL=http://mynode.onion
 
-# Pricing
-MODEL_BASED_PRICING=true
+# Pricing (optional)
+FIXED_PRICING=false
 EXCHANGE_FEE=1.005
 UPSTREAM_PROVIDER_FEE=1.05
 ```

docs/getting-started/quickstart.md

@@ -67,7 +67,7 @@ You should see:
 {
   "name": "ARoutstrNode",
   "description": "A Routstr Node",
-  "version": "0.1.2",
+  "version": "0.1.4",
   "npub": "",
   "mints": ["https://mint.minibits.cash/Bitcoin"],
   "models": {...}

docs/user-guide/models-pricing.md

@@ -11,8 +11,8 @@ Routstr supports three pricing models:
 Simple per-request charging:
 
 ```bash
-MODEL_BASED_PRICING=false
-COST_PER_REQUEST=10  # 10 sats per request
+FIXED_PRICING=true
+FIXED_COST_PER_REQUEST=10  # 10 sats per request
 ```
 
 **Best for:**
@@ -26,10 +26,10 @@ COST_PER_REQUEST=10  # 10 sats per request
 Charge based on actual token usage:
 
 ```bash
-MODEL_BASED_PRICING=false
-COST_PER_REQUEST=1              # 1 sat base fee
-COST_PER_1K_INPUT_TOKENS=5      # 5 sats per 1K input
-COST_PER_1K_OUTPUT_TOKENS=15    # 15 sats per 1K output
+FIXED_PRICING=false             # use model pricing
+FIXED_COST_PER_REQUEST=1        # optional base fee
+FIXED_PER_1K_INPUT_TOKENS=5     # optional override
+FIXED_PER_1K_OUTPUT_TOKENS=15   # optional override
 ```
 
 **Best for:**
@@ -43,7 +43,7 @@ COST_PER_1K_OUTPUT_TOKENS=15    # 15 sats per 1K output
 Dynamic pricing based on model costs:
 
 ```bash
-MODEL_BASED_PRICING=true
+FIXED_PRICING=false
 EXCHANGE_FEE=1.005      # 0.5% exchange fee
 UPSTREAM_PROVIDER_FEE=1.05  # 5% provider fee
 ```

migrations/versions/a1b2c3d4e5f6_add_settings_table.py

@@ -0,0 +1,35 @@
+"""add settings table
+
+Revision ID: a1b2c3d4e5f6
+Revises: 042f6b77d69d
+Create Date: 2025-09-06 00:00:00.000000
+"""
+
+from __future__ import annotations
+
+import sqlalchemy as sa
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision = "a1b2c3d4e5f6"
+down_revision = "042f6b77d69d"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.create_table(
+        "settings",
+        sa.Column("id", sa.Integer(), primary_key=True, nullable=False),
+        sa.Column("data", sa.Text(), nullable=False),
+        sa.Column(
+            "updated_at",
+            sa.DateTime(),
+            nullable=True,
+            server_default=sa.text("CURRENT_TIMESTAMP"),
+        ),
+    )
+
+
+def downgrade() -> None:
+    op.drop_table("settings")

migrations/versions/c0ffee123456_create_models_table.py

@@ -0,0 +1,37 @@
+"""create models table
+
+Revision ID: c0ffee123456
+Revises: a1b2c3d4e5f6
+Create Date: 2025-09-10 00:00:00.000000
+"""
+
+from __future__ import annotations
+
+import sqlalchemy as sa
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision = "c0ffee123456"
+down_revision = "a1b2c3d4e5f6"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.create_table(
+        "models",
+        sa.Column("id", sa.String(), primary_key=True, nullable=False),
+        sa.Column("name", sa.String(), nullable=False),
+        sa.Column("created", sa.Integer(), nullable=False),
+        sa.Column("description", sa.Text(), nullable=False),
+        sa.Column("context_length", sa.Integer(), nullable=False),
+        sa.Column("architecture", sa.Text(), nullable=False),
+        sa.Column("pricing", sa.Text(), nullable=False),
+        sa.Column("sats_pricing", sa.Text(), nullable=True),
+        sa.Column("per_request_limits", sa.Text(), nullable=True),
+        sa.Column("top_provider", sa.Text(), nullable=True),
+    )
+
+
+def downgrade() -> None:
+    op.drop_table("models")

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "routstr"
-version = "0.1.2"
+version = "0.1.4"
 description = "Payment proxy for your LLM endpoint using cashu and nostr."
 readme = "README.md"
 requires-python = ">=3.11"
@@ -18,6 +18,7 @@ dependencies = [
     "marshmallow>=3.13,<4.0",
     "websockets>=12.0",
     "nostr>=0.0.2",
+    "mdurl==0.1.2",
 ]
 
 [dependency-groups]

routstr/__init__.py

@@ -1,7 +1,3 @@
-import dotenv
-
-dotenv.load_dotenv()
-
 from .core.main import app as fastapi_app  # noqa
 
 __all__ = ["fastapi_app"]

routstr/auth.py

@@ -7,18 +7,14 @@ from sqlmodel import col, update
 
 from .core import get_logger
 from .core.db import ApiKey, AsyncSession
+from .core.settings import settings
 from .payment.cost_caculation import (
     CostData,
     CostDataError,
     MaxCostData,
     calculate_cost,
 )
-from .wallet import (
-    PRIMARY_MINT_URL,
-    TRUSTED_MINTS,
-    credit_balance,
-    deserialize_token_from_string,
-)
+from .wallet import credit_balance, deserialize_token_from_string
 
 logger = get_logger(__name__)
 
@@ -165,12 +161,12 @@ async def validate_bearer_key(
                     "has_expiry_time": bool(key_expiry_time),
                 },
             )
-            if token_obj.mint in TRUSTED_MINTS:
+            if token_obj.mint in settings.cashu_mints:
                 refund_currency = token_obj.unit
                 refund_mint_url = token_obj.mint
             else:
                 refund_currency = "sat"
-                refund_mint_url = PRIMARY_MINT_URL
+                refund_mint_url = settings.primary_mint
 
             new_key = ApiKey(
                 hashed_key=hashed_key,
@@ -426,7 +422,7 @@ async def adjust_payment_for_tokens(
         },
     )
 
-    match calculate_cost(response_data, deducted_max_cost):
+    match await calculate_cost(response_data, deducted_max_cost, session):
         case MaxCostData() as cost:
             logger.debug(
                 "Using max cost data (no token adjustment)",
@@ -637,3 +633,10 @@ async def adjust_payment_for_tokens(
                     }
                 },
             )
+    # Fallback return to satisfy type checker; execution should not reach here
+    return {
+        "base_msats": deducted_max_cost,
+        "input_msats": 0,
+        "output_msats": 0,
+        "total_msats": deducted_max_cost,
+    }

routstr/balance.py

@@ -1,6 +1,5 @@
 import asyncio
 import hashlib
-import os
 from time import monotonic
 from typing import Annotated, NoReturn
 
@@ -9,11 +8,15 @@ from pydantic import BaseModel
 
 from .auth import validate_bearer_key
 from .core.db import ApiKey, AsyncSession, get_session
-from .wallet import PRIMARY_MINT_URL, credit_balance, send_to_lnurl, send_token
+from .core.logging import get_logger
+from .core.settings import settings
+from .wallet import credit_balance, recieve_token, send_to_lnurl, send_token
 
 router = APIRouter()
 balance_router = APIRouter(prefix="/v1/balance")
 
+logger = get_logger(__name__)
+
 
 async def get_key_from_header(
     authorization: Annotated[str, Header(...)],
@@ -34,6 +37,7 @@ async def account_info(key: ApiKey = Depends(get_key_from_header)) -> dict:
     return {
         "api_key": "sk-" + key.hashed_key,
         "balance": key.balance,
+        "reserved": key.reserved_balance,
     }
 
 
@@ -65,6 +69,7 @@ async def wallet_info(key: ApiKey = Depends(get_key_from_header)) -> dict:
     return {
         "api_key": "sk-" + key.hashed_key,
         "balance": key.balance,
+        "reserved": key.reserved_balance,
     }
 
 
@@ -102,7 +107,7 @@ async def topup_wallet_endpoint(
     return {"msats": amount_msats}
 
 
-_REFUND_CACHE_TTL_SECONDS: int = int(os.environ.get("REFUND_CACHE_TTL_SECONDS", "3600"))
+_REFUND_CACHE_TTL_SECONDS: int = settings.refund_cache_ttl_seconds
 _refund_cache_lock: asyncio.Lock = asyncio.Lock()
 _refund_cache: dict[str, tuple[float, dict[str, str]]] = {}
 
@@ -150,30 +155,32 @@ async def refund_wallet_endpoint(
     key: ApiKey = await validate_bearer_key(bearer_value, session)
     remaining_balance_msats: int = key.balance
 
-    if remaining_balance_msats <= 0:
+    if key.refund_currency == "sat":
+        remaining_balance = remaining_balance_msats // 1000
+    else:
+        remaining_balance = remaining_balance_msats
+
+    if remaining_balance_msats > 0 and remaining_balance <= 0:
+        raise HTTPException(status_code=400, detail="Balance too small to refund")
+    elif remaining_balance <= 0:
         raise HTTPException(status_code=400, detail="No balance to refund")
 
     # Perform refund operation first, before modifying balance
     try:
         if key.refund_address:
-            if key.refund_currency == "sat":
-                remaining_balance = remaining_balance_msats * 1000
+            from .core.settings import settings as global_settings
+
             await send_to_lnurl(
                 remaining_balance,
                 key.refund_currency or "sat",
-                key.refund_mint_url or PRIMARY_MINT_URL,
+                key.refund_mint_url or global_settings.primary_mint,
                 key.refund_address,
             )
             result = {"recipient": key.refund_address}
         else:
-            refund_amount = (
-                remaining_balance_msats // 1000
-                if key.refund_currency == "sat"
-                else remaining_balance_msats
-            )
             refund_currency = key.refund_currency or "sat"
             token = await send_token(
-                refund_amount, refund_currency, key.refund_mint_url
+                remaining_balance, refund_currency, key.refund_mint_url
             )
             result = {"token": token}
 
@@ -206,6 +213,19 @@ async def refund_wallet_endpoint(
     return result
 
 
+@router.post("/donate")
+async def donate(token: str, ref: str | None = None) -> str:
+    try:
+        amount, unit, _ = await recieve_token(token)
+        if ref:
+            logger.info(
+                "donation received", extra={"ref": ref, "amount": amount, "unit": unit}
+            )
+        return "Thanks!"
+    except Exception:
+        return "Invalid token."
+
+
 @router.api_route(
     "/{path:path}",
     methods=["GET", "POST", "PUT", "DELETE"],

routstr/core/db.py

@@ -52,6 +52,20 @@ class ApiKey(SQLModel, table=True):  # type: ignore
         return self.balance - self.reserved_balance
 
 
+class ModelRow(SQLModel, table=True):  # type: ignore
+    __tablename__ = "models"
+    id: str = Field(primary_key=True)
+    name: str = Field()
+    created: int = Field()
+    description: str = Field()
+    context_length: int = Field()
+    architecture: str = Field()
+    pricing: str = Field()
+    sats_pricing: str | None = Field(default=None)
+    per_request_limits: str | None = Field(default=None)
+    top_provider: str | None = Field(default=None)
+
+
 async def balances_for_mint_and_unit(
     db_session: AsyncSession, mint_url: str, unit: str
 ) -> int:

routstr/core/logging.py

@@ -181,7 +181,12 @@ class SecurityFilter(logging.Filter):
 
 def get_log_level() -> str:
     """Get log level from environment variable."""
-    level = os.environ.get("LOG_LEVEL", "INFO").upper()
+    try:
+        from .settings import settings
+
+        level = settings.log_level.upper()
+    except Exception:
+        level = os.environ.get("LOG_LEVEL", "INFO").upper()
     # Validate log level - if invalid, default to INFO
     valid_levels = {"TRACE", "DEBUG", "INFO", "WARNING", "ERROR", "CRITICAL"}
     if level not in valid_levels:
@@ -191,11 +196,16 @@ def get_log_level() -> str:
 
 def should_enable_console_logging() -> bool:
     """Check if console logging should be enabled."""
-    return os.environ.get("ENABLE_CONSOLE_LOGGING", "true").lower() in (
-        "true",
-        "1",
-        "yes",
-    )
+    try:
+        from .settings import settings
+
+        return bool(settings.enable_console_logging)
+    except Exception:
+        return os.environ.get("ENABLE_CONSOLE_LOGGING", "true").lower() in (
+            "true",
+            "1",
+            "yes",
+        )
 
 
 def setup_logging() -> None:

routstr/core/main.py

@@ -1,5 +1,4 @@
 import asyncio
-import os
 from contextlib import asynccontextmanager
 from typing import AsyncGenerator
 
@@ -11,20 +10,27 @@ from starlette.exceptions import HTTPException
 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 MODELS, models_router, update_sats_pricing
+from ..payment.models import (
+    ensure_models_bootstrapped,
+    models_router,
+    refresh_models_periodically,
+    update_sats_pricing,
+)
 from ..proxy import proxy_router
 from ..wallet import periodic_payout
 from .admin import admin_router
-from .db import init_db, run_migrations
+from .db import create_session, init_db, run_migrations
 from .exceptions import general_exception_handler, http_exception_handler
 from .logging import get_logger, setup_logging
 from .middleware import LoggingMiddleware
+from .settings import SettingsService
+from .settings import settings as global_settings
 
 # Initialize logging first
 setup_logging()
 logger = get_logger(__name__)
 
-__version__ = "0.1.2"
+__version__ = "0.1.4-dev"
 
 
 @asynccontextmanager
@@ -35,6 +41,7 @@ async def lifespan(_: FastAPI) -> AsyncGenerator[None, None]:
     payout_task = None
     nip91_task = None
     providers_task = None
+    models_refresh_task = None
 
     try:
         # Run database migrations on startup
@@ -47,7 +54,20 @@ async def lifespan(_: FastAPI) -> AsyncGenerator[None, None]:
         # This creates any tables that might not be tracked by migrations yet
         await init_db()
 
+        # Initialize application settings (env -> computed -> DB precedence)
+        async with create_session() as session:
+            s = await SettingsService.initialize(session)
+
+        # Apply app metadata from settings
+        try:
+            app.title = s.name
+            app.description = s.description
+        except Exception:
+            pass
+
+        await ensure_models_bootstrapped()
         pricing_task = asyncio.create_task(update_sats_pricing())
+        models_refresh_task = asyncio.create_task(refresh_models_periodically())
         payout_task = asyncio.create_task(periodic_payout())
         nip91_task = asyncio.create_task(announce_provider())
         providers_task = asyncio.create_task(providers_cache_refresher())
@@ -71,6 +91,8 @@ async def lifespan(_: FastAPI) -> AsyncGenerator[None, None]:
             nip91_task.cancel()
         if providers_task is not None:
             providers_task.cancel()
+        if models_refresh_task is not None:
+            models_refresh_task.cancel()
 
         try:
             tasks_to_wait = []
@@ -82,6 +104,8 @@ async def lifespan(_: FastAPI) -> AsyncGenerator[None, None]:
                 tasks_to_wait.append(nip91_task)
             if providers_task is not None:
                 tasks_to_wait.append(providers_task)
+            if models_refresh_task is not None:
+                tasks_to_wait.append(models_refresh_task)
 
             if tasks_to_wait:
                 await asyncio.gather(*tasks_to_wait, return_exceptions=True)
@@ -93,18 +117,12 @@ async def lifespan(_: FastAPI) -> AsyncGenerator[None, None]:
             )
 
 
-app = FastAPI(
-    version=__version__,
-    title=os.environ.get("NAME", "ARoutstrNode" + __version__),
-    description=os.environ.get("DESCRIPTION", "A Routstr Node"),
-    contact={"name": os.environ.get("NAME", ""), "npub": os.environ.get("NPUB", "")},
-    lifespan=lifespan,
-)
+app = FastAPI(version=__version__, lifespan=lifespan)
+
 
-# Configure CORS
 app.add_middleware(
     CORSMiddleware,
-    allow_origins=os.environ.get("CORS_ORIGINS", "*").split(","),
+    allow_origins=global_settings.cors_origins,
     allow_credentials=True,
     allow_methods=["*"],
     allow_headers=["*"],
@@ -123,14 +141,14 @@ app.add_exception_handler(Exception, general_exception_handler)
 @app.get("/v1/info")
 async def info() -> dict:
     return {
-        "name": app.title,
-        "description": app.description,
+        "name": global_settings.name,
+        "description": global_settings.description,
         "version": __version__,
-        "npub": os.environ.get("NPUB", ""),
-        "mints": os.environ.get("CASHU_MINTS", "").split(","),
-        "http_url": os.environ.get("HTTP_URL", ""),
-        "onion_url": os.environ.get("ONION_URL", ""),
-        "models": MODELS,
+        "npub": global_settings.npub,
+        "mints": global_settings.cashu_mints,
+        "http_url": global_settings.http_url,
+        "onion_url": global_settings.onion_url,
+        "models": [],  # kept for back-compat; prefer /v1/models
     }
 
 
@@ -139,6 +157,11 @@ async def admin_redirect() -> RedirectResponse:
     return RedirectResponse("/admin/")
 
 
+@app.get("/v1/providers")
+async def providers() -> RedirectResponse:
+    return RedirectResponse("/v1/providers/")
+
+
 app.include_router(models_router)
 app.include_router(admin_router)
 app.include_router(balance_router)

routstr/core/settings.py

@@ -0,0 +1,308 @@
+from __future__ import annotations
+
+import asyncio
+import json
+import os
+from datetime import datetime, timezone
+from typing import Any
+
+from pydantic.v1 import BaseModel, BaseSettings, Field
+from sqlmodel.ext.asyncio.session import AsyncSession
+
+
+class Settings(BaseSettings):
+    class Config:
+        case_sensitive = True
+
+        @classmethod
+        def parse_env_var(cls, field_name: str, raw_value: str) -> Any:  # type: ignore[override]
+            if field_name in {"cashu_mints", "cors_origins", "relays"}:
+                v = str(raw_value).strip()
+                if v == "":
+                    return []
+                return [p.strip() for p in v.split(",") if p.strip()]
+            return raw_value
+
+    # Core
+    upstream_base_url: str = Field(default="", env="UPSTREAM_BASE_URL")
+    upstream_api_key: str = Field(default="", env="UPSTREAM_API_KEY")
+    admin_password: str = Field(default="", env="ADMIN_PASSWORD")
+
+    # Node info
+    name: str = Field(default="ARoutstrNode", env="NAME")
+    description: str = Field(default="A Routstr Node", env="DESCRIPTION")
+    npub: str = Field(default="", env="NPUB")
+    http_url: str = Field(default="", env="HTTP_URL")
+    onion_url: str = Field(default="", env="ONION_URL")
+
+    # Cashu
+    cashu_mints: list[str] = Field(default_factory=list, env="CASHU_MINTS")
+    receive_ln_address: str = Field(default="", env="RECEIVE_LN_ADDRESS")
+    primary_mint: str = Field(default="", env="PRIMARY_MINT_URL")
+
+    # Pricing
+    # Default behavior: derive pricing from MODELS
+    # If fixed_pricing is True -> use fixed_cost_per_request and ignore tokens
+    # If fixed_per_1k_* are set (non-zero) -> override model token pricing when model-based
+    fixed_pricing: bool = Field(default=False, env="FIXED_PRICING")
+    fixed_cost_per_request: int = Field(default=1, env="FIXED_COST_PER_REQUEST")
+    fixed_per_1k_input_tokens: int = Field(default=0, env="FIXED_PER_1K_INPUT_TOKENS")
+    fixed_per_1k_output_tokens: int = Field(default=0, env="FIXED_PER_1K_OUTPUT_TOKENS")
+    exchange_fee: float = Field(default=1.005, env="EXCHANGE_FEE")
+    upstream_provider_fee: float = Field(default=1.05, env="UPSTREAM_PROVIDER_FEE")
+    tolerance_percentage: float = Field(default=1.0, env="TOLERANCE_PERCENTAGE")
+    # Minimum per-request charge in millisatoshis when model pricing is free/zero
+    min_request_msat: int = Field(default=1, env="MIN_REQUEST_MSAT")
+
+    # Network
+    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=300, env="PROVIDERS_REFRESH_INTERVAL_SECONDS"
+    )
+    pricing_refresh_interval_seconds: int = Field(
+        default=120, env="PRICING_REFRESH_INTERVAL_SECONDS"
+    )
+    models_refresh_interval_seconds: int = Field(
+        default=30, env="MODELS_REFRESH_INTERVAL_SECONDS"
+    )
+    enable_pricing_refresh: bool = Field(default=True, env="ENABLE_PRICING_REFRESH")
+    enable_models_refresh: bool = Field(default=True, env="ENABLE_MODELS_REFRESH")
+    delete_removed_models: bool = Field(default=False, env="DELETE_REMOVED_MODELS")
+    refund_cache_ttl_seconds: int = Field(default=3600, env="REFUND_CACHE_TTL_SECONDS")
+
+    # Logging
+    log_level: str = Field(default="INFO", env="LOG_LEVEL")
+    enable_console_logging: bool = Field(default=True, env="ENABLE_CONSOLE_LOGGING")
+
+    # Other
+    chat_completions_api_version: str = Field(
+        default="", env="CHAT_COMPLETIONS_API_VERSION"
+    )
+    models_path: str = Field(default="models.json", env="MODELS_PATH")
+    source: str = Field(default="", env="SOURCE")
+
+    # Secrets / optional runtime controls
+    provider_id: str = Field(default="", env="PROVIDER_ID")
+    nsec: str = Field(default="", env="NSEC")
+
+    # Discovery
+    relays: list[str] = Field(default_factory=list, env="RELAYS")
+
+
+def _compute_primary_mint(cashu_mints: list[str]) -> str:
+    return cashu_mints[0] if cashu_mints else "https://mint.minibits.cash/Bitcoin"
+
+
+def resolve_bootstrap() -> Settings:
+    base = Settings()  # Reads env with custom parse_env_var
+    # Back-compat env mapping
+    try:
+        # Map MODEL_BASED_PRICING -> fixed_pricing (inverted)
+        if "MODEL_BASED_PRICING" in os.environ and "FIXED_PRICING" not in os.environ:
+            mbp_raw = os.environ.get("MODEL_BASED_PRICING", "").strip().lower()
+            mbp = mbp_raw in {"1", "true", "yes", "on"}
+            base.fixed_pricing = not mbp
+        # Map COST_PER_REQUEST -> fixed_cost_per_request if new not provided
+        if (
+            "COST_PER_REQUEST" in os.environ
+            and "FIXED_COST_PER_REQUEST" not in os.environ
+        ):
+            try:
+                base.fixed_cost_per_request = int(
+                    os.environ["COST_PER_REQUEST"].strip()
+                )
+            except Exception:
+                pass
+        # Map COST_PER_1K_* -> CUSTOM_PER_1K_*
+        if (
+            "COST_PER_1K_INPUT_TOKENS" in os.environ
+            and "FIXED_PER_1K_INPUT_TOKENS" not in os.environ
+        ):
+            try:
+                base.fixed_per_1k_input_tokens = int(
+                    os.environ["COST_PER_1K_INPUT_TOKENS"].strip()
+                )
+            except Exception:
+                pass
+        if (
+            "COST_PER_1K_OUTPUT_TOKENS" in os.environ
+            and "FIXED_PER_1K_OUTPUT_TOKENS" not in os.environ
+        ):
+            try:
+                base.fixed_per_1k_output_tokens = int(
+                    os.environ["COST_PER_1K_OUTPUT_TOKENS"].strip()
+                )
+            except Exception:
+                pass
+    except Exception:
+        pass
+    if not base.onion_url:
+        try:
+            from ..nip91 import discover_onion_url_from_tor  # type: ignore
+
+            discovered = discover_onion_url_from_tor()
+            if discovered:
+                base.onion_url = discovered
+        except Exception:
+            pass
+    # Derive NPUB from NSEC if not provided
+    if not base.npub and base.nsec:
+        try:
+            from nostr.key import PrivateKey  # type: ignore
+
+            if base.nsec.startswith("nsec"):
+                pk = PrivateKey.from_nsec(base.nsec)
+            elif len(base.nsec) == 64:
+                pk = PrivateKey(bytes.fromhex(base.nsec))
+            else:
+                pk = None
+            if pk is not None:
+                try:
+                    base.npub = pk.public_key.bech32()
+                except Exception:
+                    # Fallback to hex if bech32 not available
+                    base.npub = pk.public_key.hex()
+        except Exception:
+            pass
+    if not base.cors_origins:
+        base.cors_origins = ["*"]
+    if not base.primary_mint:
+        base.primary_mint = _compute_primary_mint(base.cashu_mints)
+    return base
+
+
+class SettingsRow(BaseModel):
+    id: int
+    data: dict[str, Any]
+    updated_at: datetime | None = None
+
+
+# Single, concrete settings instance that callers import directly
+settings: Settings = resolve_bootstrap()
+
+
+class SettingsService:
+    _current: Settings | None = None
+    _lock: asyncio.Lock = asyncio.Lock()
+
+    @classmethod
+    def get(cls) -> Settings:
+        if cls._current is None:
+            raise RuntimeError("SettingsService not initialized")
+        return cls._current
+
+    @classmethod
+    async def initialize(cls, db_session: AsyncSession) -> Settings:
+        async with cls._lock:
+            from sqlmodel import text
+
+            await db_session.exec(  # type: ignore
+                text(
+                    "CREATE TABLE IF NOT EXISTS settings (id INTEGER PRIMARY KEY, data TEXT NOT NULL, updated_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP)"
+                )
+            )
+
+            row = await db_session.exec(  # type: ignore
+                text("SELECT id, data, updated_at FROM settings WHERE id = 1")
+            )
+            row = row.first()
+            env_resolved = resolve_bootstrap()
+
+            if row is None:
+                await db_session.exec(  # type: ignore
+                    text(
+                        "INSERT INTO settings (id, data, updated_at) VALUES (1, :data, :updated_at)"
+                    ).bindparams(
+                        data=json.dumps(env_resolved.dict()),
+                        updated_at=datetime.now(timezone.utc),
+                    )
+                )
+                await db_session.commit()
+                cls._current = settings
+                # Update the existing instance in-place for all live importers
+                for k, v in env_resolved.dict().items():
+                    setattr(settings, k, v)
+                return cls._current
+
+            db_id, db_data, _updated_at = row
+            try:
+                db_json = (
+                    json.loads(db_data) if isinstance(db_data, str) else dict(db_data)
+                )
+            except Exception:
+                db_json = {}
+
+            merged_dict: dict[str, Any] = dict(env_resolved.dict())
+            merged_dict.update(
+                {k: v for k, v in db_json.items() if v not in (None, "")}
+            )
+
+            # Ensure primary_mint is consistent with cashu_mints if not explicitly set
+            if not merged_dict.get("primary_mint"):
+                merged_dict["primary_mint"] = _compute_primary_mint(
+                    merged_dict.get("cashu_mints", [])
+                )
+
+            if any(k not in db_json for k in merged_dict.keys()):
+                await db_session.exec(  # type: ignore
+                    text(
+                        "UPDATE settings SET data = :data, updated_at = :updated_at WHERE id = 1"
+                    ).bindparams(
+                        data=json.dumps(merged_dict),
+                        updated_at=datetime.now(timezone.utc),
+                    )
+                )
+                await db_session.commit()
+
+            # Update the existing instance in-place for all live importers
+            for k, v in merged_dict.items():
+                setattr(settings, k, v)
+            cls._current = settings
+            return cls._current
+
+    @classmethod
+    async def update(
+        cls, partial: dict[str, Any], db_session: AsyncSession
+    ) -> Settings:
+        async with cls._lock:
+            current = cls.get()
+            candidate_dict = {**current.dict(), **partial}
+            candidate = Settings(**candidate_dict)
+            from sqlmodel import text
+
+            # Ensure primary_mint reflects candidate mints if missing
+            if not candidate.primary_mint:
+                candidate.primary_mint = _compute_primary_mint(candidate.cashu_mints)
+
+            await db_session.exec(  # type: ignore
+                text(
+                    "UPDATE settings SET data = :data, updated_at = :updated_at WHERE id = 1"
+                ).bindparams(
+                    data=json.dumps(candidate.dict()),
+                    updated_at=datetime.now(timezone.utc),
+                )
+            )
+            await db_session.commit()
+            # Update in-place
+            for k, v in candidate.dict().items():
+                setattr(settings, k, v)
+            cls._current = settings
+            return settings
+
+    @classmethod
+    async def reload_from_db(cls, db_session: AsyncSession) -> Settings:
+        async with cls._lock:
+            from sqlmodel import text
+
+            row = await db_session.exec(text("SELECT data FROM settings WHERE id = 1"))  # type: ignore
+            row = row.first()
+            if row is None:
+                raise RuntimeError("Settings row missing")
+            (data_str,) = row
+            data = json.loads(data_str) if isinstance(data_str, str) else dict(data_str)
+            # Update in-place
+            for k, v in data.items():
+                setattr(settings, k, v)
+            cls._current = settings
+            return settings

routstr/discovery.py

@@ -1,6 +1,5 @@
 import asyncio
 import json
-import os
 import random
 import string
 from typing import Any
@@ -10,6 +9,7 @@ import websockets
 from fastapi import APIRouter
 
 from .core.logging import get_logger
+from .core.settings import settings
 
 logger = get_logger(__name__)
 
@@ -196,15 +196,18 @@ async def get_cache() -> list[dict[str, Any]]:
 
 
 def _get_discovery_relays() -> list[str]:
-    relays_env = os.getenv("RELAYS") or ""
-    discovery_relays = [r.strip() for r in relays_env.split(",") if r.strip()]
-    if not discovery_relays:
-        discovery_relays = [
+    try:
+        relays = settings.relays
+    except Exception:
+        relays = []
+    if not relays:
+        relays = [
             "wss://relay.nostr.band",
             "wss://relay.damus.io",
             "wss://relay.routstr.com",
+            "wss://nos.lol",
         ]
-    return discovery_relays
+    return relays
 
 
 async def _discover_providers(pubkey: str | None = None) -> list[dict[str, Any]]:
@@ -297,10 +300,8 @@ async def providers_cache_refresher(
 ) -> None:
     if interval_seconds is None:
         try:
-            interval_seconds = int(
-                os.getenv("PROVIDERS_REFRESH_INTERVAL_SECONDS", "300")
-            )
-        except ValueError:
+            interval_seconds = settings.providers_refresh_interval_seconds
+        except Exception:
             interval_seconds = 300
 
     await refresh_providers_cache(pubkey=pubkey)
@@ -321,8 +322,10 @@ async def fetch_provider_health(endpoint_url: str) -> dict[str, Any]:
         # Set up client arguments conditionally
         proxies = None
         if is_onion:
-            # Get Tor proxy URL from environment variable
-            tor_proxy = os.getenv("TOR_PROXY_URL", "socks5://127.0.0.1:9050")
+            try:
+                tor_proxy = settings.tor_proxy_url
+            except Exception:
+                tor_proxy = "socks5://127.0.0.1:9050"
             proxies = {"http://": tor_proxy, "https://": tor_proxy}  # type: ignore[assignment]
 
         async with httpx.AsyncClient(

routstr/nip91.py

@@ -19,6 +19,7 @@ from nostr.message_type import ClientMessageType
 from nostr.relay_manager import RelayManager
 
 from .core import get_logger
+from .core.settings import settings
 
 logger = get_logger(__name__)
 
@@ -286,23 +287,32 @@ def discover_onion_url_from_tor(base_dir: str = "/var/lib/tor") -> str | None:
 
 
 async def _determine_provider_id(public_key_hex: str, relay_urls: list[str]) -> str:
-    explicit = os.getenv("PROVIDER_ID") or os.getenv("NIP91_PROVIDER_ID")
+    explicit = settings.provider_id
     if explicit:
         logger.info(f"Using configured provider_id from env: {explicit}")
         return explicit
 
-    latest_event: dict[str, Any] | None = None
-    latest_ts = -1
-    for relay_url in relay_urls:
+    async def query_single_relay(relay_url: str) -> list[dict[str, Any]]:
         try:
             events, _ok = await query_nip91_events(relay_url, public_key_hex, None)
-            for ev in events:
-                ts = int(ev.get("created_at", 0))
-                if ts > latest_ts:
-                    latest_event = ev
-                    latest_ts = ts
+            return events
         except Exception:
-            continue
+            return []
+
+    # Query all relays concurrently
+    all_events_lists = await asyncio.gather(
+        *[query_single_relay(relay_url) for relay_url in relay_urls]
+    )
+
+    latest_event: dict[str, Any] | None = None
+    latest_ts = -1
+
+    for events_list in all_events_lists:
+        for ev in events_list:
+            ts = int(ev.get("created_at", 0))
+            if ts > latest_ts:
+                latest_event = ev
+                latest_ts = ts
 
     existing_d = _get_single_tag_value(latest_event, "d") if latest_event else None
     if existing_d:
@@ -352,7 +362,7 @@ async def announce_provider() -> None:
     Checks for existing announcements and creates new ones if needed.
     """
     # Check for NSEC in environment (use NSEC only)
-    nsec = os.getenv("NSEC")
+    nsec = settings.nsec
     if not nsec:
         logger.info("Nostr private key not found (NSEC), skipping NIP-91 announcement")
         return
@@ -366,38 +376,26 @@ async def announce_provider() -> None:
     private_key_hex, public_key_hex = keypair
     logger.info(f"Using Nostr pubkey: {public_key_hex}")
 
-    # Configure relays first (RELAYS only)
-    relay_urls_env = os.getenv("RELAYS") or ""
-    logger.debug(f"Configured relays: {relay_urls_env}")
-    relay_urls = [url.strip() for url in relay_urls_env.split(",") if url.strip()]
-    if not relay_urls:
-        relay_urls = [
-            "wss://relay.nostr.band",
-            "wss://relay.damus.io",
-            "wss://nos.lol",
-        ]
-
-    # Determine a stable provider_id
-    provider_id = await _determine_provider_id(public_key_hex, relay_urls)
-    logger.info(f"Using provider_id: {provider_id}")
-
-    # Core settings only (no ROUTSTR_* vars)
-    base_url = os.getenv("HTTP_URL")
-    onion_url = os.getenv("ONION_URL")
+    # Resolve settings and determine if we can publish BEFORE touching relays
+    try:
+        base_url: str | None = settings.http_url
+        onion_url: str | None = settings.onion_url
+        provider_name = settings.name or "Routstr Proxy"
+        provider_about = settings.description or "Privacy-preserving AI proxy via Nostr"
+        cashu_mints = [m.strip() for m in settings.cashu_mints if m.strip()]
+    except Exception:
+        base_url = settings.http_url or None
+        onion_url = settings.onion_url or None
+        provider_name = settings.name or "Routstr Proxy"
+        provider_about = settings.description or "Privacy-preserving AI proxy via Nostr"
+        cashu_mints = [m.strip() for m in settings.cashu_mints if m.strip()]
     if not onion_url:
         discovered = discover_onion_url_from_tor()
         if discovered:
             onion_url = discovered
             logger.info(f"Discovered onion URL via Tor volume: {onion_url}")
-    provider_name = os.getenv("NAME", "Routstr Proxy")
-    provider_about = os.getenv("DESCRIPTION", "Privacy-preserving AI proxy via Nostr")
-    # Mint URLs optional: include all CASHU_MINTS entries if available
-    cashu_mints = [
-        m.strip() for m in os.getenv("CASHU_MINTS", "").split(",") if m.strip()
-    ]
     mint_urls = cashu_mints if cashu_mints else None
 
-    # Build endpoint URLs (skip defaults like localhost)
     endpoint_urls: list[str] = []
     if base_url and base_url.strip() and base_url.strip() != "http://localhost:8000":
         endpoint_urls.append(base_url.strip())
@@ -415,6 +413,19 @@ async def announce_provider() -> None:
         )
         return
 
+    # Only now configure relays and determine provider_id (may query relays)
+    relay_urls = [u.strip() for u in getattr(settings, "relays", []) if u.strip()]
+    if not relay_urls:
+        relay_urls = [
+            "wss://relay.nostr.band",
+            "wss://relay.damus.io",
+            "wss://relay.routstr.com",
+            "wss://nos.lol",
+        ]
+
+    provider_id = await _determine_provider_id(public_key_hex, relay_urls)
+    logger.info(f"Using provider_id: {provider_id}")
+
     # Build metadata
     metadata = {
         "name": provider_name,
@@ -432,10 +443,10 @@ async def announce_provider() -> None:
         metadata=metadata,
     )
 
-    # Backoff configuration and state
-    backoff_base = float(os.getenv("NIP91_BACKOFF_BASE_SECONDS", "5"))
-    backoff_max = float(os.getenv("NIP91_BACKOFF_MAX_SECONDS", "900"))
-    backoff_jitter_ratio = float(os.getenv("NIP91_BACKOFF_JITTER_RATIO", "0.2"))
+    # Backoff configuration and state (sensible defaults)
+    backoff_base = 5.0
+    backoff_max = 900.0
+    backoff_jitter_ratio = 0.2
     relay_next_allowed: dict[str, float] = {}
     relay_current_delay: dict[str, float] = {}
 
@@ -499,9 +510,7 @@ async def announce_provider() -> None:
         )
 
     # Re-announce periodically (every 24 hours)
-    announcement_interval = int(
-        os.getenv("NIP91_ANNOUNCEMENT_INTERVAL", str(24 * 60 * 60))
-    )
+    announcement_interval = 24 * 60 * 60
 
     while True:
         try:

routstr/payment/cost_caculation.py

@@ -1,34 +1,16 @@
+import json
 import math
-import os
 
 from pydantic.v1 import BaseModel
+from sqlmodel import select
+from sqlmodel.ext.asyncio.session import AsyncSession
 
 from ..core import get_logger
-from .models import MODELS
+from ..core.db import ModelRow
+from ..core.settings import settings
 
 logger = get_logger(__name__)
 
-COST_PER_REQUEST = (
-    int(os.environ.get("COST_PER_REQUEST", "1")) * 1000
-)  # Convert to msats
-COST_PER_1K_INPUT_TOKENS = (
-    int(os.environ.get("COST_PER_1K_INPUT_TOKENS", "0")) * 1000
-)  # Convert to msats
-COST_PER_1K_OUTPUT_TOKENS = (
-    int(os.environ.get("COST_PER_1K_OUTPUT_TOKENS", "0")) * 1000
-)  # Convert to msats
-MODEL_BASED_PRICING = os.environ.get("MODEL_BASED_PRICING", "false").lower() == "true"
-
-logger.info(
-    "Cost calculation initialized",
-    extra={
-        "cost_per_request_msats": COST_PER_REQUEST,
-        "cost_per_1k_input_tokens_msats": COST_PER_1K_INPUT_TOKENS,
-        "cost_per_1k_output_tokens_msats": COST_PER_1K_OUTPUT_TOKENS,
-        "model_based_pricing": MODEL_BASED_PRICING,
-    },
-)
-
 
 class CostData(BaseModel):
     base_msats: int
@@ -46,8 +28,8 @@ class CostDataError(BaseModel):
     code: str
 
 
-def calculate_cost(
-    response_data: dict, max_cost: int
+async def calculate_cost(
+    response_data: dict, max_cost: int, session: AsyncSession | None = None
 ) -> CostData | MaxCostData | CostDataError:
     """
     Calculate the cost of an API request based on token usage.
@@ -85,44 +67,53 @@ def calculate_cost(
         )
         return cost_data
 
-    MSATS_PER_1K_INPUT_TOKENS = COST_PER_1K_INPUT_TOKENS
-    MSATS_PER_1K_OUTPUT_TOKENS = COST_PER_1K_OUTPUT_TOKENS
+    MSATS_PER_1K_INPUT_TOKENS: float = (
+        float(settings.fixed_per_1k_input_tokens) * 1000.0
+    )
+    MSATS_PER_1K_OUTPUT_TOKENS: float = (
+        float(settings.fixed_per_1k_output_tokens) * 1000.0
+    )
 
-    if MODEL_BASED_PRICING and MODELS:
+    if not settings.fixed_pricing and session is not None:
         response_model = response_data.get("model", "")
         logger.debug(
             "Using model-based pricing",
-            extra={
-                "model": response_model,
-                "available_models": [model.id for model in MODELS],
-            },
+            extra={"model": response_model},
         )
 
-        if response_model not in [model.id for model in MODELS]:
+        result = await session.exec(select(ModelRow.id))  # type: ignore
+        available_ids = [
+            row[0] if isinstance(row, tuple) else row for row in result.all()
+        ]
+        if response_model not in available_ids:
             logger.error(
                 "Invalid model in response",
-                extra={
-                    "response_model": response_model,
-                    "available_models": [model.id for model in MODELS],
-                },
+                extra={"response_model": response_model},
             )
             return CostDataError(
                 message=f"Invalid model in response: {response_model}",
                 code="model_not_found",
             )
 
-        model = next(model for model in MODELS if model.id == response_model)
-        if model.sats_pricing is None:
+        row = await session.get(ModelRow, response_model)
+        if row is None or not row.sats_pricing:
             logger.error(
                 "Model pricing not defined",
-                extra={"model": response_model, "model_id": model.id},
+                extra={"model": response_model, "model_id": response_model},
             )
             return CostDataError(
                 message="Model pricing not defined", code="pricing_not_found"
             )
 
-        MSATS_PER_1K_INPUT_TOKENS = model.sats_pricing.prompt * 1_000_000  # type: ignore
-        MSATS_PER_1K_OUTPUT_TOKENS = model.sats_pricing.completion * 1_000_000  # type: ignore
+        try:
+            sats_pricing = json.loads(row.sats_pricing)
+            mspp = float(sats_pricing.get("prompt", 0))
+            mspc = float(sats_pricing.get("completion", 0))
+        except Exception:
+            return CostDataError(message="Invalid pricing data", code="pricing_invalid")
+
+        MSATS_PER_1K_INPUT_TOKENS = mspp * 1_000_000.0
+        MSATS_PER_1K_OUTPUT_TOKENS = mspc * 1_000_000.0
 
         logger.info(
             "Applied model-specific pricing",

routstr/payment/helpers.py

@@ -1,26 +1,21 @@
 import json
-import os
+import math
 from typing import Mapping
 
 from fastapi import HTTPException, Response
 from fastapi.requests import Request
+from sqlmodel import select
+from sqlmodel.ext.asyncio.session import AsyncSession
 
 from ..core import get_logger
+from ..core.db import ModelRow
+from ..core.settings import settings
 from ..wallet import deserialize_token_from_string
-from .cost_caculation import COST_PER_REQUEST, MODEL_BASED_PRICING
-from .models import MODELS
+from .models import Pricing, compute_effective_max_cost_msats
 
 logger = get_logger(__name__)
 
 
-UPSTREAM_BASE_URL = os.environ.get("UPSTREAM_BASE_URL", "")
-UPSTREAM_API_KEY = os.environ.get("UPSTREAM_API_KEY", "")
-CHAT_COMPLETIONS_API_VERSION = os.environ.get("CHAT_COMPLETIONS_API_VERSION", "")
-
-if not UPSTREAM_BASE_URL:
-    raise ValueError("Please set the UPSTREAM_BASE_URL environment variable")
-
-
 def check_token_balance(headers: dict, body: dict, max_cost_for_model: int) -> None:
     if x_cashu := headers.get("x-cashu", None):
         cashu_token = x_cashu
@@ -77,7 +72,8 @@ def check_token_balance(headers: dict, body: dict, max_cost_for_model: int) -> N
         token_obj.amount if token_obj.unit == "msat" else token_obj.amount * 1000
     )
 
-    if max_cost_for_model > amount_msat:
+    fee_buffer = 60
+    if max_cost_for_model > amount_msat + fee_buffer:
         raise HTTPException(
             status_code=413,
             detail={
@@ -89,49 +85,157 @@ def check_token_balance(headers: dict, body: dict, max_cost_for_model: int) -> N
         )
 
 
-def get_max_cost_for_model(model: str, tolerance_percentage: int = 1) -> int:
+async def get_max_cost_for_model(
+    model: str, session: AsyncSession | None = None
+) -> int:
     """Get the maximum cost for a specific model."""
     logger.debug(
         "Getting max cost for model",
         extra={
             "model": model,
-            "model_based_pricing": MODEL_BASED_PRICING,
-            "has_models": bool(MODELS),
+            "fixed_pricing": settings.fixed_pricing,
+            "has_models": True,
         },
     )
 
-    if not MODEL_BASED_PRICING or not MODELS:
+    # Fixed pricing: always use fixed_cost_per_request
+    if settings.fixed_pricing:
+        default_cost_msats = settings.fixed_cost_per_request * 1000
         logger.debug(
-            "Using default cost (no model-based pricing)",
-            extra={"cost_msats": COST_PER_REQUEST, "model": model},
+            "Using fixed cost pricing",
+            extra={"cost_msats": default_cost_msats, "model": model},
         )
-        return COST_PER_REQUEST
+        return max(settings.min_request_msat, default_cost_msats)
 
-    if model not in [model.id for model in MODELS]:
+    if session is None:
+        # Without a DB session, we can't resolve model pricing; fall back to fixed cost
+        fallback_msats = settings.fixed_cost_per_request * 1000
+        logger.warning(
+            "No DB session provided for model pricing; using fixed cost",
+            extra={"requested_model": model, "using_default_cost": fallback_msats},
+        )
+        return max(settings.min_request_msat, fallback_msats)
+
+    result = await session.exec(select(ModelRow.id))  # type: ignore
+    available_ids = [row[0] if isinstance(row, tuple) else row for row in result.all()]
+    if model not in available_ids:
+        # If no models or unknown model, fall back to fixed cost if provided, else minimal default
+        fallback_msats = settings.fixed_cost_per_request * 1000
         logger.warning(
             "Model not found in available models",
             extra={
                 "requested_model": model,
-                "available_models": [m.id for m in MODELS],
-                "using_default_cost": COST_PER_REQUEST,
+                "available_models": available_ids,
+                "using_default_cost": fallback_msats,
             },
         )
-        return COST_PER_REQUEST
+        return max(settings.min_request_msat, fallback_msats)
 
-    for m in MODELS:
-        if m.id == model:
-            max_cost = m.sats_pricing.max_cost * 1000 * (1 - tolerance_percentage / 100)  # type: ignore
-            logger.debug(
-                "Found model-specific max cost",
-                extra={"model": model, "max_cost_msats": max_cost},
-            )
-            return int(max_cost)
+    row = await session.get(ModelRow, model)
+    if row and row.sats_pricing:
+        try:
+            sats_dict = json.loads(row.sats_pricing)
+        except Exception:
+            sats_dict = None
+        if isinstance(sats_dict, dict):
+            effective_msats = compute_effective_max_cost_msats(sats_dict)
+            if effective_msats > 0:
+                logger.debug(
+                    "Found model-specific max cost",
+                    extra={"model": model, "max_cost_msats": effective_msats},
+                )
+                return effective_msats
 
     logger.warning(
-        "Model pricing not found, using default",
-        extra={"model": model, "default_cost_msats": COST_PER_REQUEST},
+        "Model pricing not found, using fixed cost",
+        extra={
+            "model": model,
+            "default_cost_msats": settings.fixed_cost_per_request * 1000,
+        },
     )
-    return COST_PER_REQUEST
+    return max(settings.min_request_msat, settings.fixed_cost_per_request * 1000)
+
+
+async def calculate_discounted_max_cost(
+    max_cost_for_model: int, body: dict, session: AsyncSession | None = None
+) -> int:
+    """Calculate the discounted max cost for a request using model pricing when available."""
+    if settings.fixed_pricing or session is None:
+        return max_cost_for_model
+
+    model = body.get("model", "unknown")
+    model_pricing = await get_model_cost_info(model, session=session)
+    if not model_pricing:
+        return max_cost_for_model
+
+    tol = settings.tolerance_percentage
+    tol_factor = max(0.0, 1 - float(tol) / 100.0)
+    max_prompt_allowed_sats = model_pricing.max_prompt_cost * tol_factor
+    max_completion_allowed_sats = model_pricing.max_completion_cost * tol_factor
+
+    adjusted = max_cost_for_model
+
+    if messages := body.get("messages"):
+        prompt_tokens = estimate_tokens(messages)
+        estimated_prompt_delta_sats = (
+            max_prompt_allowed_sats - prompt_tokens * model_pricing.prompt
+        )
+        if estimated_prompt_delta_sats >= 0:
+            adjusted = adjusted - math.floor(estimated_prompt_delta_sats * 1000)
+        else:
+            adjusted = adjusted + math.ceil(-estimated_prompt_delta_sats * 1000)
+
+    max_tokens_raw = body.get("max_tokens", None)
+    if max_tokens_raw is not None:
+        try:
+            max_tokens_int = int(max_tokens_raw)
+        except (TypeError, ValueError):
+            logger.warning(
+                "Invalid max_tokens; ignoring in cost adjustment",
+                extra={"max_tokens": str(max_tokens_raw)[:64], "model": model},
+            )
+        else:
+            estimated_completion_delta_sats = (
+                max_completion_allowed_sats - max_tokens_int * model_pricing.completion
+            )
+            if estimated_completion_delta_sats >= 0:
+                adjusted = adjusted - math.floor(estimated_completion_delta_sats * 1000)
+            else:
+                adjusted = adjusted + math.ceil(-estimated_completion_delta_sats * 1000)
+
+    adjusted = min(max_cost_for_model, adjusted)
+
+    logger.debug(
+        "Discounted max cost computed",
+        extra={
+            "model": model,
+            "original_msats": max_cost_for_model,
+            "adjusted_msats": adjusted,
+            "tolerance_pct": tol,
+        },
+    )
+
+    return max(0, adjusted)
+
+
+def estimate_tokens(messages: list) -> int:
+    return len(str(messages)) // 3
+
+
+async def get_model_cost_info(
+    model_id: str, session: AsyncSession | None = None
+) -> Pricing | None:
+    if not model_id or model_id == "unknown":
+        return None
+    if session is None:
+        return None
+    row = await session.get(ModelRow, model_id)
+    if row and row.sats_pricing:
+        try:
+            return Pricing(**json.loads(row.sats_pricing))  # type: ignore
+        except Exception:
+            return None
+    return None
 
 
 def create_error_response(
@@ -161,11 +265,12 @@ def create_error_response(
 
 def prepare_upstream_headers(request_headers: dict) -> dict:
     """Prepare headers for upstream request, removing sensitive/problematic ones."""
+    upstream_api_key = settings.upstream_api_key
     logger.debug(
         "Preparing upstream headers",
         extra={
             "original_headers_count": len(request_headers),
-            "has_upstream_api_key": bool(UPSTREAM_API_KEY),
+            "has_upstream_api_key": bool(upstream_api_key),
         },
     )
 
@@ -184,8 +289,8 @@ def prepare_upstream_headers(request_headers: dict) -> dict:
             removed_headers.append(header)
 
     # Handle authorization
-    if UPSTREAM_API_KEY:
-        headers["Authorization"] = f"Bearer {UPSTREAM_API_KEY}"
+    if upstream_api_key:
+        headers["Authorization"] = f"Bearer {upstream_api_key}"
         if headers.pop("authorization", None) is not None:
             removed_headers.append("authorization (replaced with upstream key)")
     else:
@@ -198,7 +303,7 @@ def prepare_upstream_headers(request_headers: dict) -> dict:
         extra={
             "final_headers_count": len(headers),
             "removed_headers": removed_headers,
-            "added_upstream_auth": bool(UPSTREAM_API_KEY),
+            "added_upstream_auth": bool(upstream_api_key),
         },
     )
 
@@ -210,6 +315,7 @@ def prepare_upstream_params(
 ) -> dict[str, str]:
     """Prepare query params for upstream request, optionally adding api-version for chat/completions."""
     params: dict[str, str] = dict(query_params or {})
-    if path.endswith("chat/completions") and CHAT_COMPLETIONS_API_VERSION:
-        params["api-version"] = CHAT_COMPLETIONS_API_VERSION
+    chat_api_version = settings.chat_completions_api_version
+    if path.endswith("chat/completions") and chat_api_version:
+        params["api-version"] = chat_api_version
     return params

routstr/payment/lnurl.py

@@ -230,7 +230,11 @@ async def get_lnurl_invoice(
 
 
 async def raw_send_to_lnurl(
-    wallet: Wallet, proofs: list[Proof], lnurl: str, unit: str
+    wallet: Wallet,
+    proofs: list[Proof],
+    lnurl: str,
+    unit: str,
+    amount: int | None = None,
 ) -> int:
     """Send funds to an LNURL address.
 
@@ -255,6 +259,11 @@ async def raw_send_to_lnurl(
         paid = await wallet.send_to_lnurl("user@getalby.com", 50, unit="usd")
     """
     total_balance = sum(proof.amount for proof in proofs)
+    if amount and total_balance < amount:
+        raise ValueError("Amount to send is higher than available proofs.")
+    else:
+        assert isinstance(amount, int)
+        total_balance = amount
     lnurl_data = await get_lnurl_data(lnurl)
 
     if unit == "sat":
@@ -285,6 +294,10 @@ async def raw_send_to_lnurl(
     melt_quote_resp = await wallet.melt_quote(
         invoice=bolt11_invoice, amount_msat=final_amount
     )
+
+    if amount:
+        proofs, _ = await wallet.select_to_send(proofs, amount, set_reserved=True)
+
     _ = await wallet.melt(
         proofs=proofs,
         invoice=bolt11_invoice,

routstr/payment/models.py

@@ -1,13 +1,17 @@
 import asyncio
 import json
-import os
+import random
 from pathlib import Path
 from urllib.request import urlopen
 
-from fastapi import APIRouter
+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, create_session, get_session
 from ..core.logging import get_logger
+from ..core.settings import settings
 from .price import sats_usd_ask_price
 
 logger = get_logger(__name__)
@@ -30,6 +34,8 @@ class Pricing(BaseModel):
     image: float
     web_search: float
     internal_reasoning: float
+    max_prompt_cost: float = 0.0  # in sats not msats
+    max_completion_cost: float = 0.0  # in sats not msats
     max_cost: float = 0.0  # in sats not msats
 
 
@@ -52,12 +58,42 @@ class Model(BaseModel):
     top_provider: TopProvider | None = None
 
 
-MODELS: list[Model] = []
+def compute_effective_max_cost_msats(
+    pricing: Pricing | dict | None,
+) -> int:
+    if pricing is None:
+        try:
+            return max(1, int(settings.min_request_msat))
+        except Exception:
+            return 1
+
+    pricing_obj = (
+        pricing if isinstance(pricing, Pricing) else Pricing.parse_obj(pricing)  # type: ignore[arg-type]
+    )
+
+    try:
+        tolerance = float(getattr(settings, "tolerance_percentage", 0.0))
+    except Exception:
+        tolerance = 0.0
+
+    tolerance_factor = max(0.0, 1.0 - tolerance / 100.0)
+
+    try:
+        min_request_msat = max(1, int(settings.min_request_msat))
+    except Exception:
+        min_request_msat = 1
+
+    base_msats = int(float(pricing_obj.max_cost or 0.0) * 1000.0 * tolerance_factor)
+
+    if base_msats <= 0:
+        return min_request_msat
+
+    return max(min_request_msat, base_msats)
 
 
 def fetch_openrouter_models(source_filter: str | None = None) -> list[dict]:
     """Fetches model information from OpenRouter API."""
-    base_url = os.getenv("BASE_URL", "https://openrouter.ai/api/v1")
+    base_url = "https://openrouter.ai/api/v1"
 
     try:
         with urlopen(f"{base_url}/models") as response:
@@ -80,6 +116,9 @@ def fetch_openrouter_models(source_filter: str | None = None) -> list[dict]:
                     "(free)" in model.get("name", "")
                     or model_id == "openrouter/auto"
                     or model_id == "google/gemini-2.5-pro-exp-03-25"
+                    or model_id == "opengvlab/internvl3-78b"
+                    or model_id == "openrouter/sonoma-dusk-alpha"
+                    or model_id == "openrouter/sonoma-sky-alpha"
                 ):
                     continue
 
@@ -91,6 +130,14 @@ def fetch_openrouter_models(source_filter: str | None = None) -> list[dict]:
         return []
 
 
+def is_openrouter_upstream() -> bool:
+    try:
+        base = (settings.upstream_base_url or "").strip().rstrip("/")
+    except Exception:
+        return False
+    return base.lower() == "https://openrouter.ai/api/v1"
+
+
 def load_models() -> list[Model]:
     """Load model definitions from a JSON file or auto-generate from OpenRouter API.
 
@@ -100,7 +147,10 @@ def load_models() -> list[Model]:
     and no user file is provided, it will be used as a fallback.
     """
 
-    models_path = Path(os.environ.get("MODELS_PATH", "models.json"))
+    try:
+        models_path = Path(settings.models_path)
+    except Exception:
+        models_path = Path("models.json")
 
     # Check if user has actively provided a models.json file
     if models_path.exists():
@@ -108,14 +158,23 @@ def load_models() -> list[Model]:
         try:
             with models_path.open("r") as f:
                 data = json.load(f)
-            return [Model(**model) for model in data.get("models", [])]
+            return [Model(**model) for model in data.get("models", [])]  # type: ignore
         except Exception as e:
             logger.error(f"Error loading models from {models_path}: {e}")
             # Fall through to auto-generation
 
-    # Auto-generate models from OpenRouter API
+    # Only auto-generate from OpenRouter when upstream is OpenRouter
+    if not is_openrouter_upstream():
+        logger.info(
+            "Skipping auto-generation from OpenRouter because upstream_base_url is not https://openrouter.ai/api/v1"
+        )
+        return []
+
     logger.info("Auto-generating models from OpenRouter API")
-    source_filter = os.getenv("SOURCE")
+    try:
+        source_filter = settings.source or None
+    except Exception:
+        source_filter = None
     source_filter = source_filter if source_filter and source_filter.strip() else None
 
     models_data = fetch_openrouter_models(source_filter=source_filter)
@@ -124,58 +183,388 @@ def load_models() -> list[Model]:
         return []
 
     logger.info(f"Successfully fetched {len(models_data)} models from OpenRouter API")
-    return [Model(**model) for model in models_data]
+    return [Model(**model) for model in models_data]  # type: ignore
 
 
-MODELS = load_models()
+def _row_to_model(row: ModelRow) -> Model:
+    architecture = json.loads(row.architecture)
+    pricing = json.loads(row.pricing)
+    sats_pricing = json.loads(row.sats_pricing) if row.sats_pricing else None
+    per_request_limits = (
+        json.loads(row.per_request_limits) if row.per_request_limits else None
+    )
+    top_provider = json.loads(row.top_provider) if row.top_provider else None
+
+    # Enforce minimum per-request fee on free/zero-priced models in API output
+    try:
+        if isinstance(pricing, dict):
+            if float(pricing.get("request", 0.0)) <= 0.0:
+                pricing["request"] = max(pricing.get("request", 0.0), 0.0)
+        if isinstance(sats_pricing, dict):
+            if float(sats_pricing.get("request", 0.0)) <= 0.0:
+                # Convert min_request_msat to sats for sats_pricing fields that are in sats
+                sats_min = max(1, int(settings.min_request_msat)) / 1000.0
+                sats_pricing["request"] = max(
+                    sats_pricing.get("request", 0.0), sats_min
+                )
+    except Exception:
+        pass
+
+    if isinstance(sats_pricing, dict):
+        effective_msats = compute_effective_max_cost_msats(sats_pricing)
+        sats_pricing["max_cost"] = effective_msats / 1000.0
+
+    return Model(
+        id=row.id,
+        name=row.name,
+        created=row.created,
+        description=row.description,
+        context_length=row.context_length,
+        architecture=Architecture.parse_obj(architecture),
+        pricing=Pricing.parse_obj(pricing),
+        sats_pricing=Pricing.parse_obj(sats_pricing) if sats_pricing else None,
+        per_request_limits=per_request_limits,
+        top_provider=TopProvider.parse_obj(top_provider) if top_provider else None,
+    )
+
+
+def _model_to_row_payload(model: Model) -> dict[str, str | int | None]:
+    return {
+        "id": model.id,
+        "name": model.name,
+        "created": model.created,
+        "description": model.description,
+        "context_length": model.context_length,
+        "architecture": json.dumps(model.architecture.dict()),
+        "pricing": json.dumps(model.pricing.dict()),
+        "sats_pricing": json.dumps(model.sats_pricing.dict())
+        if model.sats_pricing
+        else None,
+        "per_request_limits": json.dumps(model.per_request_limits)
+        if model.per_request_limits is not None
+        else None,
+        "top_provider": json.dumps(model.top_provider.dict())
+        if model.top_provider is not None
+        else None,
+    }
+
+
+async def list_models(session: AsyncSession | None = None) -> list[Model]:
+    if session is not None:
+        result = await session.exec(select(ModelRow))  # type: ignore
+        rows = result.all()
+        return [_row_to_model(r) for r in rows]
+    async with create_session() as s:
+        result = await s.exec(select(ModelRow))  # type: ignore
+        rows = result.all()
+        return [_row_to_model(r) for r in rows]
+
+
+async def get_model_by_id(
+    model_id: str, session: AsyncSession | None = None
+) -> Model | None:
+    if session is not None:
+        row = await session.get(ModelRow, model_id)
+        return _row_to_model(row) if row else None
+    async with create_session() as s:
+        row = await s.get(ModelRow, model_id)
+        return _row_to_model(row) if row else None
+
+
+async def ensure_models_bootstrapped() -> None:
+    async with create_session() as s:
+        existing = (await s.exec(select(ModelRow.id).limit(1))).all()  # type: ignore
+        if existing:
+            return
+
+        try:
+            models_path = Path(settings.models_path)
+        except Exception:
+            models_path = Path("models.json")
+
+        models_to_insert: list[dict] = []
+        if models_path.exists():
+            try:
+                with models_path.open("r") as f:
+                    data = json.load(f)
+                models_to_insert = data.get("models", [])
+                logger.info(
+                    f"Bootstrapping {len(models_to_insert)} models from {models_path}"
+                )
+            except Exception as e:
+                logger.error(f"Error loading models from {models_path}: {e}")
+
+        if not models_to_insert and is_openrouter_upstream():
+            logger.info("Bootstrapping models from OpenRouter API")
+            source_filter = None
+            try:
+                src = settings.source or None
+                source_filter = src if src and src.strip() else None
+            except Exception:
+                pass
+            models_to_insert = fetch_openrouter_models(source_filter=source_filter)
+        elif not models_to_insert:
+            logger.info(
+                "No models.json found and upstream is not OpenRouter; skipping bootstrap"
+            )
+
+        for m in models_to_insert:
+            try:
+                model = Model(**m)  # type: ignore
+            except Exception:
+                # Some OpenRouter models include extra fields; only map required ones
+                continue
+            exists = await s.get(ModelRow, model.id)
+            if exists:
+                continue
+            payload = _model_to_row_payload(model)
+            s.add(ModelRow(**payload))  # type: ignore
+        await s.commit()
 
 
 async def update_sats_pricing() -> None:
     while True:
         try:
+            try:
+                if not settings.enable_pricing_refresh:
+                    return
+            except Exception:
+                pass
             sats_to_usd = await sats_usd_ask_price()
-            for model in MODELS:
-                model.sats_pricing = Pricing(
-                    **{k: v / sats_to_usd for k, v in model.pricing.dict().items()}
-                )
-                mspp = model.sats_pricing.prompt
-                mspc = model.sats_pricing.completion
-                if (tp := model.top_provider) and (
-                    tp.context_length or tp.max_completion_tokens
-                ):
-                    if (cl := model.top_provider.context_length) and (
-                        mct := model.top_provider.max_completion_tokens
-                    ):
-                        model.sats_pricing.max_cost = (cl - mct) * mspp + mct * mspc
-                    elif cl := model.top_provider.context_length:
-                        model.sats_pricing.max_cost = cl * 0.8 * mspp + cl * 0.2 * mspc
-                    elif mct := model.top_provider.max_completion_tokens:
-                        model.sats_pricing.max_cost = mct * 4 * mspp + mct * mspc
-                    else:
-                        model.sats_pricing.max_cost = 1_000_000 * mspp + 32_000 * mspc
-                elif model.context_length:
-                    model.sats_pricing.max_cost = (
-                        model.sats_pricing.prompt * model.context_length * 0.8
-                    ) + (model.sats_pricing.completion * model.context_length * 0.2)
-                else:
-                    p = model.sats_pricing.prompt * 1_000_000
-                    c = model.sats_pricing.completion * 32_000
-                    r = model.sats_pricing.request * 100_000
-                    i = model.sats_pricing.image * 100
-                    w = model.sats_pricing.web_search * 1000
-                    ir = model.sats_pricing.internal_reasoning * 100
-                    model.sats_pricing.max_cost = p + c + r + i + w + ir
+            async with create_session() as s:
+                result = await s.exec(select(ModelRow))  # type: ignore
+                rows = result.all()
+                changed = 0
+                for row in rows:
+                    try:
+                        pricing = Pricing.parse_obj(json.loads(row.pricing))
+                        top_provider = (
+                            TopProvider.parse_obj(json.loads(row.top_provider))
+                            if row.top_provider
+                            else None
+                        )
+                        sats = Pricing.parse_obj(
+                            {k: v / sats_to_usd for k, v in pricing.dict().items()}
+                        )
+                        # Enforce minimum per-request charge floor in sats
+                        try:
+                            min_req_msat = max(
+                                1, int(getattr(settings, "min_request_msat", 1))
+                            )
+                        except Exception:
+                            min_req_msat = 1
+                        min_req_sats = float(min_req_msat) / 1000.0
+                        if sats.request <= 0.0:
+                            sats.request = min_req_sats
+                        mspp = sats.prompt
+                        mspc = sats.completion
+                        if top_provider and (
+                            top_provider.context_length
+                            or top_provider.max_completion_tokens
+                        ):
+                            if (cl := top_provider.context_length) and (
+                                mct := top_provider.max_completion_tokens
+                            ):
+                                max_prompt_cost = (cl - mct) * mspp
+                                max_completion_cost = mct * mspc
+                                sats.max_prompt_cost = max_prompt_cost
+                                sats.max_completion_cost = max_completion_cost
+                                sats.max_cost = max_prompt_cost + max_completion_cost
+                            elif cl := top_provider.context_length:
+                                max_prompt_cost = cl * 0.8 * mspp
+                                max_completion_cost = cl * 0.2 * mspc
+                                sats.max_prompt_cost = max_prompt_cost
+                                sats.max_completion_cost = max_completion_cost
+                                sats.max_cost = max_prompt_cost + max_completion_cost
+                            elif mct := top_provider.max_completion_tokens:
+                                max_prompt_cost = mct * 4 * mspp
+                                max_completion_cost = mct * mspc
+                                sats.max_prompt_cost = max_prompt_cost
+                                sats.max_completion_cost = max_completion_cost
+                                sats.max_cost = max_prompt_cost + max_completion_cost
+                            else:
+                                max_prompt_cost = 1_000_000 * mspp
+                                max_completion_cost = 32_000 * mspc
+                                sats.max_prompt_cost = max_prompt_cost
+                                sats.max_completion_cost = max_completion_cost
+                                sats.max_cost = max_prompt_cost + max_completion_cost
+                        elif row.context_length:
+                            max_prompt_cost = mspp * row.context_length * 0.8
+                            max_completion_cost = mspc * row.context_length * 0.2
+                            sats.max_prompt_cost = max_prompt_cost
+                            sats.max_completion_cost = max_completion_cost
+                            sats.max_cost = max_prompt_cost + max_completion_cost
+                        else:
+                            p = mspp * 1_000_000
+                            c = mspc * 32_000
+                            r = sats.request * 100_000
+                            i = sats.image * 100
+                            w = sats.web_search * 1000
+                            ir = sats.internal_reasoning * 100
+                            sats.max_prompt_cost = p
+                            sats.max_completion_cost = c
+                            sats.max_cost = p + c + r + i + w + ir
+
+                        # Ensure overall minimum per-request total cost floor
+                        if (sats.max_cost or 0.0) < min_req_sats:
+                            sats.max_cost = min_req_sats
+
+                        new_json = json.dumps(sats.dict())
+                        if row.sats_pricing != new_json:
+                            row.sats_pricing = new_json
+                            s.add(row)
+                            changed += 1
+                    except Exception as per_row_error:
+                        logger.error(
+                            "Failed to update pricing for model",
+                            extra={
+                                "model_id": row.id,
+                                "error": str(per_row_error),
+                                "error_type": type(per_row_error).__name__,
+                            },
+                        )
+                if changed:
+                    await s.commit()
         except asyncio.CancelledError:
             break
         except Exception as e:
             logger.error(f"Error updating sats pricing: {e}")
         try:
-            await asyncio.sleep(10)
+            interval = getattr(settings, "pricing_refresh_interval_seconds", 120)
+            jitter = max(0.0, float(interval) * 0.1)
+            await asyncio.sleep(interval + random.uniform(0, jitter))
+        except asyncio.CancelledError:
+            break
+
+
+async def sync_models_with_api(
+    source_filter: str | None = None, delete_removed: bool = False
+) -> dict[str, int]:
+    """Fetch models from OpenRouter and sync with database.
+
+    Args:
+        source_filter: Optional source filter (e.g., 'anthropic')
+        delete_removed: If True, delete models that no longer exist in API
+
+    Returns:
+        Dict with counts: inserted, updated, deleted
+    """
+    models = fetch_openrouter_models(source_filter=source_filter)
+    if not models:
+        return {"inserted": 0, "updated": 0, "deleted": 0}
+
+    async with create_session() as s:
+        result = await s.exec(select(ModelRow))  # type: ignore
+        existing_rows = {row.id: row for row in result.all()}
+
+        fetched_ids = set()
+        inserted = 0
+        updated = 0
+
+        for m in models:
+            try:
+                model = Model(**m)  # type: ignore
+            except Exception:
+                continue
+
+            fetched_ids.add(model.id)
+            payload = _model_to_row_payload(model)
+
+            if model.id not in existing_rows:
+                try:
+                    s.add(ModelRow(**payload))  # type: ignore
+                    inserted += 1
+                except Exception:
+                    pass
+            else:
+                existing_row = existing_rows[model.id]
+                changed = False
+                for key, value in payload.items():
+                    if getattr(existing_row, key) != value:
+                        setattr(existing_row, key, value)
+                        changed = True
+                if changed:
+                    s.add(existing_row)
+                    updated += 1
+
+        deleted = 0
+        if delete_removed:
+            for existing_id in existing_rows:
+                if existing_id not in fetched_ids:
+                    row_to_delete = existing_rows[existing_id]
+                    await s.delete(row_to_delete)
+                    deleted += 1
+
+        if inserted or updated or deleted:
+            await s.commit()
+
+        return {"inserted": inserted, "updated": updated, "deleted": deleted}
+
+
+async def refresh_models_periodically() -> None:
+    """Background task: periodically fetch OpenRouter models and sync with database.
+
+    - Respects optional SOURCE filter from settings
+    - Updates existing models with new information
+    - Optionally deletes models no longer in API (if settings.delete_removed_models)
+    - Sleeps according to settings.models_refresh_interval_seconds; disabled when 0
+    """
+    interval = getattr(settings, "models_refresh_interval_seconds", 0)
+    if not interval or interval <= 0:
+        return
+
+    # Only refresh from OpenRouter when upstream is OpenRouter
+    if not is_openrouter_upstream():
+        logger.info("Skipping models refresh: upstream_base_url is not OpenRouter")
+        return
+
+    while True:
+        try:
+            try:
+                if not settings.enable_models_refresh:
+                    return
+            except Exception:
+                pass
+            try:
+                src = settings.source or None
+                source_filter = src if src and src.strip() else None
+            except Exception:
+                source_filter = None
+
+            try:
+                delete_removed = getattr(settings, "delete_removed_models", False)
+            except Exception:
+                delete_removed = False
+
+            counts = await sync_models_with_api(
+                source_filter=source_filter, delete_removed=delete_removed
+            )
+
+            if counts["inserted"] or counts["updated"] or counts["deleted"]:
+                logger.info(
+                    "Models synced",
+                    extra={
+                        "inserted": counts["inserted"],
+                        "updated": counts["updated"],
+                        "deleted": counts["deleted"],
+                    },
+                )
+        except asyncio.CancelledError:
+            break
+        except Exception as e:
+            logger.error(
+                "Error during models refresh",
+                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
 
 
 @models_router.get("/v1/models")
 @models_router.get("/models", include_in_schema=False)
-async def models() -> dict:
-    return {"data": MODELS}
+async def models(session: AsyncSession = Depends(get_session)) -> dict:
+    items = await list_models(session)
+    return {"data": items}

routstr/payment/price.py

@@ -1,17 +1,15 @@
 import asyncio
-import os
 
 import httpx
 
 from ..core import get_logger
+from ..core.settings import settings
 
 logger = get_logger(__name__)
 
-# artifical spread to cover conversion fees
-EXCHANGE_FEE = float(os.environ.get("EXCHANGE_FEE", "1.005"))  # 0.5% default
-UPSTREAM_PROVIDER_FEE = float(
-    os.environ.get("UPSTREAM_PROVIDER_FEE", "1.05")
-)  # 5% default (e.g. openrouter charges 5% margin)
+
+def _fees() -> tuple[float, float]:
+    return settings.exchange_fee, settings.upstream_provider_fee
 
 
 async def kraken_btc_usd(client: httpx.AsyncClient) -> float | None:
@@ -95,7 +93,8 @@ async def btc_usd_ask_price() -> float:
                 raise ValueError("Unable to fetch BTC price from any exchange")
 
             min_price = min(valid_prices)
-            final_price = min_price / (EXCHANGE_FEE * UPSTREAM_PROVIDER_FEE)
+            exchange_fee, provider_fee = _fees()
+            final_price = min_price / (exchange_fee * provider_fee)
             return final_price
 
         except Exception as e:

routstr/payment/x_cashu.py

@@ -7,10 +7,11 @@ from fastapi import BackgroundTasks, HTTPException, Request
 from fastapi.responses import Response, StreamingResponse
 
 from ..core import get_logger
+from ..core.db import create_session
+from ..core.settings import settings
 from ..wallet import recieve_token, send_token
 from .cost_caculation import CostData, CostDataError, MaxCostData, calculate_cost
 from .helpers import (
-    UPSTREAM_BASE_URL,
     create_error_response,
     prepare_upstream_headers,
     prepare_upstream_params,
@@ -109,7 +110,7 @@ async def forward_to_upstream(
     if path.startswith("v1/"):
         path = path.replace("v1/", "")
 
-    url = f"{UPSTREAM_BASE_URL}/{path}"
+    url = f"{settings.upstream_base_url}/{path}"
 
     logger.debug(
         "Forwarding request to upstream",
@@ -553,43 +554,45 @@ async def get_cost(
         extra={"model": model, "has_usage": "usage" in response_data},
     )
 
-    match calculate_cost(response_data, max_cost_for_model):
-        case MaxCostData() as cost:
-            logger.debug(
-                "Using max cost pricing",
-                extra={"model": model, "max_cost_msats": cost.total_msats},
-            )
-            return cost
-        case CostData() as cost:
-            logger.debug(
-                "Using token-based pricing",
-                extra={
-                    "model": model,
-                    "total_cost_msats": cost.total_msats,
-                    "input_msats": cost.input_msats,
-                    "output_msats": cost.output_msats,
-                },
-            )
-            return cost
-        case CostDataError() as error:
-            logger.error(
-                "Cost calculation error",
-                extra={
-                    "model": model,
-                    "error_message": error.message,
-                    "error_code": error.code,
-                },
-            )
-            raise HTTPException(
-                status_code=400,
-                detail={
-                    "error": {
-                        "message": error.message,
-                        "type": "invalid_request_error",
-                        "code": error.code,
-                    }
-                },
-            )
+    async with create_session() as session:
+        match await calculate_cost(response_data, max_cost_for_model, session):
+            case MaxCostData() as cost:
+                logger.debug(
+                    "Using max cost pricing",
+                    extra={"model": model, "max_cost_msats": cost.total_msats},
+                )
+                return cost
+            case CostData() as cost:
+                logger.debug(
+                    "Using token-based pricing",
+                    extra={
+                        "model": model,
+                        "total_cost_msats": cost.total_msats,
+                        "input_msats": cost.input_msats,
+                        "output_msats": cost.output_msats,
+                    },
+                )
+                return cost
+            case CostDataError() as error:
+                logger.error(
+                    "Cost calculation error",
+                    extra={
+                        "model": model,
+                        "error_message": error.message,
+                        "error_code": error.code,
+                    },
+                )
+                raise HTTPException(
+                    status_code=400,
+                    detail={
+                        "error": {
+                            "message": error.message,
+                            "type": "invalid_request_error",
+                            "code": error.code,
+                        }
+                    },
+                )
+    return None
 
 
 async def send_refund(amount: int, unit: str, mint: str | None = None) -> str:

routstr/proxy.py

@@ -15,8 +15,9 @@ from .auth import (
 )
 from .core import get_logger
 from .core.db import ApiKey, AsyncSession, create_session, get_session
+from .core.settings import settings
 from .payment.helpers import (
-    UPSTREAM_BASE_URL,
+    calculate_discounted_max_cost,
     check_token_balance,
     create_error_response,
     get_max_cost_for_model,
@@ -29,6 +30,104 @@ logger = get_logger(__name__)
 proxy_router = APIRouter()
 
 
+def _extract_upstream_error_message(body_bytes: bytes) -> tuple[str, str | None]:
+    """Extract a human-friendly message and optional upstream error code from a response body."""
+    message: str = "Upstream request failed"
+    upstream_code: str | None = None
+    if not body_bytes:
+        return message, upstream_code
+    try:
+        data = json.loads(body_bytes)
+        if isinstance(data, dict):
+            err = data.get("error")
+            if isinstance(err, dict):
+                raw_msg = err.get("message") or err.get("detail") or err.get("error")
+                if isinstance(raw_msg, (str, int, float)):
+                    message = str(raw_msg)
+                upstream_code_raw = err.get("code") or err.get("type")
+                if isinstance(upstream_code_raw, (str, int, float)):
+                    upstream_code = str(upstream_code_raw)
+            elif "message" in data and isinstance(data["message"], (str, int, float)):
+                message = str(data["message"])  # type: ignore[arg-type]
+            elif "detail" in data and isinstance(data["detail"], (str, int, float)):
+                message = str(data["detail"])  # type: ignore[arg-type]
+    except Exception:
+        preview = body_bytes.decode("utf-8", errors="ignore").strip()
+        if preview:
+            message = preview[:500]
+    return message, upstream_code
+
+
+async def map_upstream_error_response(
+    request: Request,
+    path: str,
+    upstream_response: httpx.Response,
+) -> Response:
+    """Map upstream non-200 responses to standardized error responses.
+
+    - Known cases are mapped to friendly messages and appropriate status codes
+    - Unknown errors are converted to a generic 502
+    """
+    status_code = upstream_response.status_code
+    headers = dict(upstream_response.headers)
+    content_type = headers.get("content-type", "")
+    try:
+        body_bytes = await upstream_response.aread()
+    except Exception:
+        body_bytes = b""
+
+    message, upstream_code = _extract_upstream_error_message(body_bytes)
+    lowered_message = message.lower()
+    lowered_code = (upstream_code or "").lower()
+
+    error_type = "upstream_error"
+    mapped_status = 502
+
+    # Specific mappings
+    if status_code in (400, 422):
+        error_type = "invalid_request_error"
+        mapped_status = 400
+    elif status_code in (401, 403):
+        error_type = "upstream_auth_error"
+        mapped_status = 502
+    elif status_code == 404:
+        # Many providers return 404 for unknown models or routes
+        if path.endswith("chat/completions"):
+            error_type = "invalid_model"
+            mapped_status = 400
+            if not message or message == "Upstream request failed":
+                message = "Requested model is not available upstream"
+        elif "model" in lowered_message or "model" in lowered_code:
+            error_type = "invalid_model"
+            mapped_status = 400
+            if not message or message == "Upstream request failed":
+                message = "Requested model is not available upstream"
+        else:
+            error_type = "upstream_error"
+            mapped_status = 502
+    elif status_code == 429:
+        error_type = "rate_limit_exceeded"
+        mapped_status = 429
+    elif status_code >= 500:
+        error_type = "upstream_error"
+        mapped_status = 502
+
+    # Include upstream content type hint in logs for diagnostics
+    logger.debug(
+        "Mapped upstream error",
+        extra={
+            "path": path,
+            "upstream_status": status_code,
+            "mapped_status": mapped_status,
+            "error_type": error_type,
+            "upstream_content_type": content_type,
+            "message_preview": message[:200],
+        },
+    )
+
+    return create_error_response(error_type, message, mapped_status, request=request)
+
+
 async def handle_streaming_chat_completion(
     response: httpx.Response, key: ApiKey, max_cost_for_model: int
 ) -> StreamingResponse:
@@ -307,7 +406,7 @@ async def forward_to_upstream(
     if path.startswith("v1/"):
         path = path.replace("v1/", "")
 
-    url = f"{UPSTREAM_BASE_URL}/{path}"
+    url = f"{settings.upstream_base_url}/{path}"
 
     logger.info(
         "Forwarding request to upstream",
@@ -361,6 +460,17 @@ async def forward_to_upstream(
             },
         )
 
+        # Map and return errors immediately to provide clear messages
+        if response.status_code != 200:
+            try:
+                mapped_error = await map_upstream_error_response(
+                    request, path, response
+                )
+            finally:
+                await response.aclose()
+                await client.aclose()
+            return mapped_error
+
         # For chat completions, we need to handle token-based pricing
         if path.endswith("chat/completions"):
             # Check if client requested streaming
@@ -554,7 +664,10 @@ async def proxy(
             )
 
     model = request_body_dict.get("model", "unknown")
-    max_cost_for_model = get_max_cost_for_model(model=model)
+    _max_cost_for_model = await get_max_cost_for_model(model=model, session=session)
+    max_cost_for_model = await calculate_discounted_max_cost(
+        _max_cost_for_model, request_body_dict, session
+    )
     check_token_balance(headers, request_body_dict, max_cost_for_model)
 
     # Handle authentication
@@ -651,18 +764,10 @@ async def proxy(
                 "upstream_headers": response.headers
                 if hasattr(response, "headers")
                 else None,
-                "upstream_response": response.body
-                if hasattr(response, "body")
-                else None,
             },
         )
-        request_id = (
-            request.state.request_id if hasattr(request.state, "request_id") else None
-        )
-        raise HTTPException(
-            status_code=502,
-            detail=f"Upstream request failed, please contact support with request id: {request_id}",
-        )
+        # Return the mapped error response generated earlier rather than masking with 502
+        return response
 
     return response
 
@@ -756,7 +861,7 @@ async def forward_get_to_upstream(
     if path.startswith("v1/"):
         path = path.replace("v1/", "")
 
-    url = f"{UPSTREAM_BASE_URL}/{path}"
+    url = f"{settings.upstream_base_url}/{path}"
 
     logger.info(
         "Forwarding GET request to upstream",
@@ -782,6 +887,12 @@ async def forward_get_to_upstream(
                 "GET request forwarded successfully",
                 extra={"path": path, "status_code": response.status_code},
             )
+            if response.status_code != 200:
+                try:
+                    mapped = await map_upstream_error_response(request, path, response)
+                finally:
+                    await response.aclose()
+                return mapped
 
             return StreamingResponse(
                 response.aiter_bytes(),

routstr/wallet.py

@@ -1,6 +1,5 @@
 import asyncio
 import math
-import os
 from typing import TypedDict
 
 from cashu.core.base import Proof, Token
@@ -8,19 +7,14 @@ from cashu.wallet.helpers import deserialize_token_from_string
 from cashu.wallet.wallet import Wallet
 
 from .core import db, get_logger
+from .core.settings import settings
 from .payment.lnurl import raw_send_to_lnurl
 
 logger = get_logger(__name__)
 
 
-CASHU_MINTS = os.environ.get("CASHU_MINTS", "https://mint.minibits.cash/Bitcoin")
-TRUSTED_MINTS = CASHU_MINTS.split(",")
-PRIMARY_MINT_URL = TRUSTED_MINTS[0]
-RECEIVE_LN_ADDRESS = os.environ.get("RECEIVE_LN_ADDRESS", "")
-
-
 async def get_balance(unit: str) -> int:
-    wallet = await get_wallet(PRIMARY_MINT_URL, unit)
+    wallet = await get_wallet(settings.primary_mint, unit)
     return wallet.available_balance.amount
 
 
@@ -34,7 +28,7 @@ async def recieve_token(
     wallet = await get_wallet(token_obj.mint, token_obj.unit, load=False)
     wallet.keyset_id = token_obj.keysets[0]
 
-    if token_obj.mint not in TRUSTED_MINTS:
+    if token_obj.mint not in settings.cashu_mints:
         return await swap_to_primary_mint(token_obj, wallet)
 
     wallet.verify_proofs_dleq(token_obj.proofs)
@@ -44,8 +38,10 @@ async def recieve_token(
 
 async def send(amount: int, unit: str, mint_url: str | None = None) -> tuple[int, str]:
     """Internal send function - returns amount and serialized token"""
-    wallet: Wallet = await get_wallet(mint_url or PRIMARY_MINT_URL, unit)
-    proofs = get_proofs_per_mint_and_unit(wallet, mint_url or PRIMARY_MINT_URL, unit)
+    wallet: Wallet = await get_wallet(mint_url or settings.primary_mint, unit)
+    proofs = get_proofs_per_mint_and_unit(
+        wallet, mint_url or settings.primary_mint, unit
+    )
 
     send_proofs, _ = await wallet.select_to_send(
         proofs, amount, set_reserved=True, include_fees=False
@@ -86,7 +82,7 @@ async def swap_to_primary_mint(
         raise ValueError("Invalid unit")
     estimated_fee_sat = math.ceil(max(amount_msat // 1000 * 0.01, 2))
     amount_msat_after_fee = amount_msat - estimated_fee_sat * 1000
-    primary_wallet = await get_wallet(PRIMARY_MINT_URL, "sat")
+    primary_wallet = await get_wallet(settings.primary_mint, "sat")
 
     minted_amount = int(amount_msat_after_fee // 1000)
     mint_quote = await primary_wallet.request_mint(minted_amount)
@@ -100,7 +96,7 @@ async def swap_to_primary_mint(
     )
     _ = await primary_wallet.mint(minted_amount, quote_id=mint_quote.quote)
 
-    return int(minted_amount), "sat", PRIMARY_MINT_URL
+    return int(minted_amount), "sat", settings.primary_mint
 
 
 async def credit_balance(
@@ -156,9 +152,7 @@ async def get_wallet(mint_url: str, unit: str = "sat", load: bool = True) -> Wal
     global _wallets
     id = f"{mint_url}_{unit}"
     if id not in _wallets:
-        _wallets[id] = await Wallet.with_db(
-            mint_url, db=".wallet", load_all_keysets=True, unit=unit
-        )
+        _wallets[id] = await Wallet.with_db(mint_url, db=".wallet", unit=unit)
 
     if load:
         await _wallets[id].load_mint()
@@ -259,7 +253,7 @@ async def fetch_all_balances(
     async with db.create_session() as session:
         tasks = [
             fetch_balance(session, mint_url, unit)
-            for mint_url in TRUSTED_MINTS
+            for mint_url in settings.cashu_mints
             for unit in units
         ]
 
@@ -299,14 +293,14 @@ async def fetch_all_balances(
 
 
 async def periodic_payout() -> None:
-    if not RECEIVE_LN_ADDRESS:
+    if not settings.receive_ln_address:
         logger.error("RECEIVE_LN_ADDRESS is not set, skipping payout")
         return
     while True:
-        await asyncio.sleep(60 * 5)
+        await asyncio.sleep(60 * 15)
         try:
             async with db.create_session() as session:
-                for mint_url in TRUSTED_MINTS:
+                for mint_url in settings.cashu_mints:
                     for unit in ["sat", "msat"]:
                         wallet = await get_wallet(mint_url, unit)
                         proofs = get_proofs_per_mint_and_unit(
@@ -323,7 +317,11 @@ async def periodic_payout() -> None:
                         min_amount = 210 if unit == "sat" else 210000
                         if available_balance > min_amount:
                             amount_received = await raw_send_to_lnurl(
-                                wallet, proofs, RECEIVE_LN_ADDRESS, unit
+                                wallet,
+                                proofs,
+                                settings.receive_ln_address,
+                                unit,
+                                amount=available_balance,
                             )
                             logger.info(
                                 "Payout sent successfully",

scripts/crontab.example

@@ -1,7 +1,5 @@
-REPO_DIR=/home/user/proxy
-LOG_FILE=/home/user/proxy/update.log
-* * * * * /home/user/proxy/scripts/auto_update.sh >/dev/null 2>&1
+# Example crontab entries for Routstr tasks
 
-OUTPUT_FILE=/home/user/proxy/models.json
-BASE_URL=https://openrouter.ai/api/v1
-0 * * * * python3 /home/user/proxy/scripts/models_meta.py >/dev/null 2>&1
+# Update models.json daily at 03:15 (optional)
+# OUTPUT_FILE=/app/models.json SOURCE=openrouter
+15 3 * * * /usr/local/bin/python /app/scripts/models_meta.py >> /var/log/cron.log 2>&1

scripts/models_meta.py

@@ -42,13 +42,13 @@ class Model(TypedDict):
 
 
 OUTPUT_FILE = os.getenv("OUTPUT_FILE", "models.json")
-BASE_URL = os.getenv("BASE_URL", "https://openrouter.ai/api/v1")
 SOURCE = os.getenv("SOURCE")
 
 
 def fetch_openrouter_models(source_filter: str | None = None) -> list[Model]:
     """Fetches model information from OpenRouter API."""
-    with urlopen(f"{BASE_URL}/models") as response:
+    base_url = "https://openrouter.ai/api/v1"
+    with urlopen(f"{base_url}/models") as response:
         data = json.loads(response.read().decode("utf-8"))
 
         models_data: list[Model] = []

tests/integration/conftest.py

@@ -33,8 +33,8 @@ if use_local_services:
         "RECEIVE_LN_ADDRESS": "test@routstr.com",
         "REFUND_PROCESSING_INTERVAL": "3600",
         "NSEC": "nsec1testkey1234567890abcdef",
-        "COST_PER_REQUEST": "10",
-        "MODEL_BASED_PRICING": "true",
+        "FIXED_COST_PER_REQUEST": "10",
+        "FIXED_PRICING": "false",
         "MINIMUM_PAYOUT": "1000",
         "PAYOUT_INTERVAL": "86400",
         "NAME": "TestRoutstrNode",
@@ -55,14 +55,15 @@ else:
         "RECEIVE_LN_ADDRESS": "test@routstr.com",
         "REFUND_PROCESSING_INTERVAL": "3600",
         "NSEC": "nsec1testkey1234567890abcdef",
-        "COST_PER_REQUEST": "10",
-        "MODEL_BASED_PRICING": "true",
+        "FIXED_COST_PER_REQUEST": "10",
+        "FIXED_PRICING": "false",
         "MINIMUM_PAYOUT": "1000",
         "PAYOUT_INTERVAL": "86400",
     }
 
 # Set test environment variables before importing the app
 os.environ.update(test_env)
+os.environ.pop("ADMIN_PASSWORD", None)
 
 from routstr.core.db import ApiKey, get_session  # noqa: E402
 from routstr.core.main import app, lifespan  # noqa: E402
@@ -507,10 +508,15 @@ async def integration_app(
     else:
         # Use testmint with wallet patches for all integration tests
         mint_url = os.environ.get("CASHU_MINTS", "http://localhost:3338")
+        from routstr.core.settings import settings as _settings
+
+        # Passthrough discounted max cost to avoid dependence on MODELS in tests
+        def _passthrough_discount(max_cost_for_model: int, body: dict) -> int:
+            return max_cost_for_model
+
         with (
             patch("routstr.core.db.engine", integration_engine),
-            patch("routstr.wallet.TRUSTED_MINTS", [mint_url]),
-            patch("routstr.wallet.PRIMARY_MINT_URL", mint_url),
+            patch.object(_settings, "cashu_mints", [mint_url]),
             patch("routstr.auth.credit_balance", testmint_wallet.credit_balance),
             patch("routstr.wallet.credit_balance", testmint_wallet.credit_balance),
             patch("routstr.balance.credit_balance", testmint_wallet.credit_balance),
@@ -521,6 +527,10 @@ async def integration_app(
             patch("websockets.connect") as mock_websockets,
             patch("routstr.payment.price.btc_usd_ask_price", return_value=50000.0),
             patch("routstr.payment.price.sats_usd_ask_price", return_value=0.0005),
+            patch(
+                "routstr.payment.helpers.calculate_discounted_max_cost",
+                side_effect=_passthrough_discount,
+            ),
         ):
             # Configure the WebSocket mock for discovery service - fast failure for performance tests
             async def mock_websocket_connect(*args: Any, **kwargs: Any) -> None:

tests/integration/test_background_tasks.py

@@ -10,7 +10,7 @@ from unittest.mock import AsyncMock, patch
 import pytest
 
 from routstr.core.db import ApiKey
-from routstr.payment.models import MODELS, Model, Pricing, update_sats_pricing
+from routstr.payment.models import Model, Pricing, update_sats_pricing
 from routstr.wallet import periodic_payout
 
 
@@ -57,51 +57,49 @@ class TestPricingUpdateTask:
                 },
             )
 
-            # Add test model to MODELS list
-            original_models = MODELS.copy()
-            MODELS.clear()
-            MODELS.append(test_model)
+            # Compute sats pricing once using the same logic as the background task
+            # Run the pricing update logic once directly
+            sats_to_usd = mock_sats_usd
+            _pdict = {k: v / sats_to_usd for k, v in test_model.pricing.dict().items()}
+            test_model.sats_pricing = Pricing(
+                prompt=_pdict.get("prompt", 0.0),
+                completion=_pdict.get("completion", 0.0),
+                request=_pdict.get("request", 0.0),
+                image=_pdict.get("image", 0.0),
+                web_search=_pdict.get("web_search", 0.0),
+                internal_reasoning=_pdict.get("internal_reasoning", 0.0),
+                max_prompt_cost=_pdict.get("max_prompt_cost", 0.0),
+                max_completion_cost=_pdict.get("max_completion_cost", 0.0),
+                max_cost=_pdict.get("max_cost", 0.0),
+            )
+            mspp = test_model.sats_pricing.prompt
+            mspc = test_model.sats_pricing.completion
+            if (tp := test_model.top_provider) and (
+                tp.context_length or tp.max_completion_tokens
+            ):
+                if (cl := test_model.top_provider.context_length) and (
+                    mct := test_model.top_provider.max_completion_tokens
+                ):
+                    test_model.sats_pricing.max_cost = (cl - mct) * mspp + mct * mspc
 
-            try:
-                # Run the pricing update logic once directly
-                sats_to_usd = mock_sats_usd
-                for model in [test_model]:
-                    model.sats_pricing = Pricing(
-                        **{k: v / sats_to_usd for k, v in model.pricing.dict().items()}
-                    )
-                    mspp = model.sats_pricing.prompt
-                    mspc = model.sats_pricing.completion
-                    if (tp := model.top_provider) and (
-                        tp.context_length or tp.max_completion_tokens
-                    ):
-                        if (cl := model.top_provider.context_length) and (
-                            mct := model.top_provider.max_completion_tokens
-                        ):
-                            model.sats_pricing.max_cost = (cl - mct) * mspp + mct * mspc
+            # Verify sats pricing was calculated correctly
+            assert test_model.sats_pricing is not None
+            assert test_model.sats_pricing.prompt == pytest.approx(
+                0.001 / mock_sats_usd
+            )
+            assert test_model.sats_pricing.completion == pytest.approx(
+                0.002 / mock_sats_usd
+            )
 
-                # Verify sats pricing was calculated correctly
-                assert test_model.sats_pricing is not None
-                assert test_model.sats_pricing.prompt == pytest.approx(
-                    0.001 / mock_sats_usd
-                )
-                assert test_model.sats_pricing.completion == pytest.approx(
-                    0.002 / mock_sats_usd
-                )
+            # Verify max_cost calculation
+            # Logic uses (context_length - max_completion_tokens) * prompt + max_completion_tokens * completion
+            expected_max_cost = (
+                (4096 - 1024) * test_model.sats_pricing.prompt
+                + 1024 * test_model.sats_pricing.completion
+            )
+            assert test_model.sats_pricing.max_cost == pytest.approx(expected_max_cost)
 
-                # Verify max_cost calculation
-                # Logic uses (context_length - max_completion_tokens) * prompt + max_completion_tokens * completion
-                expected_max_cost = (
-                    (4096 - 1024) * test_model.sats_pricing.prompt
-                    + 1024 * test_model.sats_pricing.completion
-                )
-                assert test_model.sats_pricing.max_cost == pytest.approx(
-                    expected_max_cost
-                )
-
-            finally:
-                # Restore original models
-                MODELS.clear()
-                MODELS.extend(original_models)
+            # Nothing to clean up; no global state was modified
 
     async def test_handles_provider_api_failures(self) -> None:
         """Test that pricing update continues running even if price API fails"""
@@ -159,37 +157,39 @@ class TestPricingUpdateTask:
             ),
         )
 
-        original_models = MODELS.copy()
-        MODELS.clear()
-        MODELS.append(test_model)
+        # Initialize pricing once to ensure consistent state
+        with patch(
+            "routstr.payment.price.sats_usd_ask_price",
+            AsyncMock(return_value=0.00002),
+        ):
+            sats_to_usd = 0.00002
+            _pdict = {k: v / sats_to_usd for k, v in test_model.pricing.dict().items()}
+            test_model.sats_pricing = Pricing(
+                prompt=_pdict.get("prompt", 0.0),
+                completion=_pdict.get("completion", 0.0),
+                request=_pdict.get("request", 0.0),
+                image=_pdict.get("image", 0.0),
+                web_search=_pdict.get("web_search", 0.0),
+                internal_reasoning=_pdict.get("internal_reasoning", 0.0),
+                max_prompt_cost=_pdict.get("max_prompt_cost", 0.0),
+                max_completion_cost=_pdict.get("max_completion_cost", 0.0),
+                max_cost=_pdict.get("max_cost", 0.0),
+            )
 
-        try:
-            with patch(
-                "routstr.payment.price.sats_usd_ask_price",
-                AsyncMock(return_value=0.00002),
-            ):
-                # Initialize pricing once to ensure consistent state
-                sats_to_usd = 0.00002
-                test_model.sats_pricing = Pricing(
-                    **{k: v / sats_to_usd for k, v in test_model.pricing.dict().items()}
-                )
+            # Simulate concurrent access to the model
+            results = []
 
-                # Simulate concurrent access to the model
-                results = []
+            async def access_model() -> None:
+                await asyncio.sleep(0.05)  # Small delay
+                results.append(test_model.sats_pricing)
 
-                async def access_model() -> None:
-                    await asyncio.sleep(0.05)  # Small delay
-                    results.append(test_model.sats_pricing)
+            # Run multiple concurrent accesses - they should all see the consistent state
+            await asyncio.gather(*[access_model() for _ in range(10)])
 
-                # Run multiple concurrent accesses - they should all see the consistent state
-                await asyncio.gather(*[access_model() for _ in range(10)])
+            # All accesses should see consistent state
+            assert all(r is not None for r in results)
 
-                # All accesses should see consistent state
-                assert all(r is not None for r in results)
-
-        finally:
-            MODELS.clear()
-            MODELS.extend(original_models)
+        # No global state to restore
 
 
 @pytest.mark.asyncio

tests/integration/test_error_handling_edge_cases.py

@@ -628,14 +628,11 @@ class TestEdgeCaseCombinations:
         a single request (which costs 1000 msats). It then makes 5 concurrent requests
         to verify that all requests fail with 402 Payment Required errors.
 
-        Note: The test disables MODEL_BASED_PRICING to avoid model lookup errors
+        Note: The test enables fixed pricing to avoid model lookup errors
         since the test environment doesn't have models configured.
         """
-        # Disable MODEL_BASED_PRICING for this test to avoid model lookup issues
-        monkeypatch.setattr(
-            "routstr.payment.cost_caculation.MODEL_BASED_PRICING", False
-        )
-        monkeypatch.setattr("routstr.payment.helpers.MODEL_BASED_PRICING", False)
+        # Disable model-based pricing for this test to avoid model lookup issues
+        monkeypatch.setattr("routstr.core.settings.settings.fixed_pricing", True)
 
         # Create a new API key with very low balance
         # Generate a unique API key
@@ -645,7 +642,7 @@ class TestEdgeCaseCombinations:
         # Create the API key with only 500 msats (less than one request cost)
         new_key = ApiKey(
             hashed_key=api_key_hash,
-            balance=500,  # Less than COST_PER_REQUEST (1000 msats)
+            balance=500,  # Less than fixed cost per request (1000 msats)
             reserved_balance=0,
             total_spent=0,
             total_requests=0,

tests/integration/test_general_info_endpoints.py

@@ -271,36 +271,23 @@ async def test_models_endpoint_accept_headers(integration_client: AsyncClient) -
 async def test_admin_endpoint_unauthenticated(
     integration_client: AsyncClient, db_snapshot: Any
 ) -> None:
-    """Test GET /admin/ endpoint without authentication"""
-
-    # Capture initial database state
+    """Test GET /admin/ endpoint without authentication shows setup form"""
     await db_snapshot.capture()
 
     response = await integration_client.get("/admin/")
 
-    # Should return 200 with login form (not 401/403)
     assert response.status_code == 200
     assert "text/html" in response.headers["content-type"]
 
-    # Response should be HTML
     html_content = response.text
     assert "<!DOCTYPE html>" in html_content
     assert "<html>" in html_content
+    assert "<form" in html_content
+    assert 'type="password"' in html_content
+    assert "password" in html_content.lower()
+    assert "<script>" in html_content or "<script " in html_content
+    assert "Initial Admin Setup" in html_content or "setup" in html_content.lower()
 
-    # Either shows login form or message about setting ADMIN_PASSWORD
-    if "ADMIN_PASSWORD" in html_content:
-        # When ADMIN_PASSWORD is not set, it shows a message
-        assert "Please set a secure ADMIN_PASSWORD" in html_content
-    else:
-        # When ADMIN_PASSWORD is set, it shows a login form
-        assert "<form" in html_content
-        assert 'type="password"' in html_content
-        assert "password" in html_content.lower()
-        assert "login" in html_content.lower()
-        # Should have JavaScript for form handling
-        assert "<script>" in html_content or "<script " in html_content
-
-    # Verify no database state changes
     diff = await db_snapshot.diff()
     assert len(diff["api_keys"]["added"]) == 0
     assert len(diff["api_keys"]["removed"]) == 0

tests/unit/test_payment_helpers.py

@@ -1,46 +1,165 @@
 import os
-from unittest.mock import Mock, patch
+from unittest.mock import AsyncMock, Mock, patch
+
+import pytest
+from fastapi import HTTPException
 
 # Set required env vars before importing
 os.environ["UPSTREAM_BASE_URL"] = "http://test"
 os.environ["UPSTREAM_API_KEY"] = "test"
 
-from routstr.payment.helpers import get_max_cost_for_model  # noqa: E402
+from routstr.core.settings import settings  # noqa: E402
+from routstr.payment.helpers import (  # noqa: E402
+    calculate_discounted_max_cost,
+    check_token_balance,
+    get_max_cost_for_model,
+)
+from routstr.payment.models import Pricing  # noqa: E402
 
 
-def test_get_max_cost_for_model_known() -> None:
-    mock_model = Mock()
-    mock_model.id = "gpt-4"
-    mock_model.sats_pricing = Mock()
-    mock_model.sats_pricing.max_cost = 500
+async def test_get_max_cost_for_model_known() -> None:
+    # Mock DB session behavior
+    mock_session = AsyncMock()
+    # available ids
+    mock_exec_result = Mock()
+    mock_exec_result.all = Mock(return_value=[("gpt-4",)])
+    mock_session.exec.return_value = mock_exec_result
+    # row with sats_pricing
+    row = Mock()
+    row.sats_pricing = (
+        "{"  # minimal required fields for Pricing model
+        '"prompt": 0.0, "completion": 0.0, "request": 0.0, '
+        '"image": 0.0, "web_search": 0.0, "internal_reasoning": 0.0, '
+        '"max_cost": 500'
+        "}"
+    )
+    mock_session.get.return_value = row
 
-    with patch("routstr.payment.helpers.MODELS", [mock_model]):
-        with patch("routstr.payment.helpers.MODEL_BASED_PRICING", True):
-            cost = get_max_cost_for_model("gpt-4", tolerance_percentage=0)
+    with patch.object(settings, "fixed_pricing", False):
+        with patch.object(settings, "tolerance_percentage", 0):
+            cost = await get_max_cost_for_model("gpt-4", session=mock_session)
             assert cost == 500000  # 500 sats * 1000 = msats
 
 
-def test_get_max_cost_for_model_unknown() -> None:
-    with patch("routstr.payment.helpers.MODELS", []):
-        with patch("routstr.payment.helpers.COST_PER_REQUEST", 100):
-            cost = get_max_cost_for_model("unknown-model", tolerance_percentage=0)
-            assert cost == 100
+async def test_get_max_cost_for_model_unknown() -> None:
+    mock_session = AsyncMock()
+    mock_exec_result = Mock()
+    mock_exec_result.all = Mock(return_value=[])
+    mock_session.exec.return_value = mock_exec_result
+    mock_session.get.return_value = None
+
+    with patch.object(settings, "fixed_cost_per_request", 100):
+        with patch.object(settings, "tolerance_percentage", 0):
+            cost = await get_max_cost_for_model("unknown-model", session=mock_session)
+            assert cost == 100000
 
 
-def test_get_max_cost_for_model_disabled() -> None:
-    with patch("routstr.payment.helpers.MODEL_BASED_PRICING", False):
-        with patch("routstr.payment.helpers.COST_PER_REQUEST", 200):
-            cost = get_max_cost_for_model("any-model", tolerance_percentage=0)
-            assert cost == 200
+async def test_get_max_cost_for_model_disabled() -> None:
+    with patch.object(settings, "fixed_pricing", True):
+        with patch.object(settings, "fixed_cost_per_request", 200):
+            with patch.object(settings, "tolerance_percentage", 0):
+                cost = await get_max_cost_for_model("any-model", session=None)
+                assert cost == 200000
 
 
-def test_get_max_cost_for_model_tolerance() -> None:
-    mock_model = Mock()
-    mock_model.id = "gpt-4"
-    mock_model.sats_pricing = Mock()
-    mock_model.sats_pricing.max_cost = 500
+async def test_get_max_cost_for_model_tolerance() -> None:
+    mock_session = AsyncMock()
+    mock_exec_result = Mock()
+    mock_exec_result.all = Mock(return_value=[("gpt-4",)])
+    mock_session.exec.return_value = mock_exec_result
+    row = Mock()
+    row.sats_pricing = (
+        "{"  # minimal required fields for Pricing model
+        '"prompt": 0.0, "completion": 0.0, "request": 0.0, '
+        '"image": 0.0, "web_search": 0.0, "internal_reasoning": 0.0, '
+        '"max_cost": 500'
+        "}"
+    )
+    mock_session.get.return_value = row
 
-    with patch("routstr.payment.helpers.MODELS", [mock_model]):
-        with patch("routstr.payment.helpers.MODEL_BASED_PRICING", True):
-            cost = get_max_cost_for_model("gpt-4", tolerance_percentage=10)
+    with patch.object(settings, "fixed_pricing", False):
+        with patch.object(settings, "tolerance_percentage", 10):
+            cost = await get_max_cost_for_model("gpt-4", session=mock_session)
             assert cost == 450000  # 500 sats * 1000 * 0.9 = 450000
+
+
+async def test_calculate_discounted_max_cost_no_session() -> None:
+    with patch.object(settings, "fixed_pricing", False):
+        result = await calculate_discounted_max_cost(123, {}, session=None)
+        assert result == 123
+
+
+async def test_calculate_discounted_max_cost_clamped(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    mock_session = AsyncMock()
+    mock_exec_result = Mock()
+    mock_exec_result.all = Mock(return_value=[("gpt-4",)])
+    mock_session.exec.return_value = mock_exec_result
+
+    pricing = {
+        "prompt": 0.0,
+        "completion": 0.0,
+        "request": 0.0,
+        "image": 0.0,
+        "web_search": 0.0,
+        "internal_reasoning": 0.0,
+        "max_prompt_cost": 100.0,
+        "max_completion_cost": 200.0,
+        "max_cost": 300.0,
+    }
+
+    async def mock_get_model_cost_info(*args, **kwargs):  # type: ignore
+        return Pricing(**pricing)
+
+    monkeypatch.setattr(
+        "routstr.payment.helpers.get_model_cost_info", mock_get_model_cost_info
+    )
+
+    with patch.object(settings, "fixed_pricing", False):
+        with patch.object(settings, "tolerance_percentage", 0):
+            result = await calculate_discounted_max_cost(
+                320000,
+                {"model": "gpt-4", "max_tokens": 10, "messages": ["one"]},
+                session=mock_session,
+            )
+            assert 0 <= result <= 320000
+
+
+async def test_check_token_balance_with_fee_buffer(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    class Token:
+        unit = "msat"
+        amount = 320400
+
+    def mock_deserialize(_value: str) -> Token:
+        return Token()
+
+    monkeypatch.setattr(
+        "routstr.payment.helpers.deserialize_token_from_string", mock_deserialize
+    )
+    headers = {"x-cashu": "token"}
+    body = {"model": "gpt-4"}
+    check_token_balance(headers, body, max_cost_for_model=320450)
+
+
+def test_check_token_balance_insufficient(monkeypatch: pytest.MonkeyPatch) -> None:
+    class Token:
+        unit = "msat"
+        amount = 320200
+
+    def mock_deserialize(_value: str) -> Token:
+        return Token()
+
+    monkeypatch.setattr(
+        "routstr.payment.helpers.deserialize_token_from_string", mock_deserialize
+    )
+    headers = {"x-cashu": "token"}
+    body = {"model": "gpt-4"}
+    with pytest.raises(HTTPException) as exc:
+        check_token_balance(headers, body, max_cost_for_model=320450)
+    assert exc.value.status_code == 413
+    detail = exc.value.detail  # type: ignore[assignment]
+    assert isinstance(detail, dict)
+    assert detail.get("type") == "minimum_balance_required"

tests/unit/test_settings.py

@@ -0,0 +1,37 @@
+import os
+
+import pytest
+from sqlalchemy.ext.asyncio import create_async_engine
+from sqlmodel.ext.asyncio.session import AsyncSession
+
+from routstr.core.settings import SettingsService
+
+
+@pytest.mark.asyncio
+async def test_settings_seed_from_env_and_persist() -> None:
+    os.environ["UPSTREAM_BASE_URL"] = "https://api.test/v1"
+    os.environ.pop("ONION_URL", None)
+
+    engine = create_async_engine("sqlite+aiosqlite:///:memory:")
+    async with AsyncSession(engine, expire_on_commit=False) as session:
+        settings = await SettingsService.initialize(session)
+
+        assert settings.upstream_base_url == "https://api.test/v1"
+        # ONION_URL may be empty if not discoverable
+        assert isinstance(settings.onion_url, str)
+
+
+@pytest.mark.asyncio
+async def test_settings_db_precedence_over_env() -> None:
+    os.environ["UPSTREAM_BASE_URL"] = "https://api.env/v1"
+
+    engine = create_async_engine("sqlite+aiosqlite:///:memory:")
+    async with AsyncSession(engine, expire_on_commit=False) as session:
+        _ = await SettingsService.initialize(session)
+        updated = await SettingsService.update({"name": "DBName"}, session)
+        assert updated.name == "DBName"
+
+        # Change env and re-initialize; DB should still win
+        os.environ["NAME"] = "EnvName"
+        again = await SettingsService.initialize(session)
+        assert again.name == "DBName"

tests/unit/test_wallet.py

@@ -39,7 +39,9 @@ async def test_recieve_token_valid() -> None:
     mock_wallet = Mock()
     mock_wallet.split = AsyncMock()
 
-    with patch("routstr.wallet.TRUSTED_MINTS", ["http://mint:3338"]):
+    from routstr.core.settings import settings
+
+    with patch.object(settings, "cashu_mints", ["http://mint:3338"]):
         with patch("routstr.wallet.deserialize_token_from_string") as mock_deserialize:
             mock_token = Mock()
             mock_token.keysets = ["keyset1"]
@@ -82,7 +84,9 @@ async def test_credit_balance() -> None:
     mock_key.balance = 5000000
     mock_session = AsyncMock()
 
-    with patch("routstr.wallet.PRIMARY_MINT_URL", "http://mint:3338"):
+    from routstr.core.settings import settings
+
+    with patch.object(settings, "cashu_mints", ["http://mint:3338"]):
         with patch(
             "routstr.wallet.recieve_token",
             return_value=(1000, "sat", "http://mint:3338"),

uv.lock

@@ -1783,7 +1783,7 @@ wheels = [
 
 [[package]]
 name = "routstr"
-version = "0.1.2"
+version = "0.1.4"
 source = { editable = "." }
 dependencies = [
     { name = "aiosqlite" },
@@ -1793,6 +1793,7 @@ dependencies = [
     { name = "greenlet" },
     { name = "httpx", extra = ["socks"] },
     { name = "marshmallow" },
+    { name = "mdurl" },
     { name = "nostr" },
     { name = "python-json-logger" },
     { name = "secp256k1" },
@@ -1824,6 +1825,7 @@ requires-dist = [
     { name = "greenlet", specifier = ">=3.2.1" },
     { name = "httpx", extras = ["socks"], specifier = ">=0.25.2" },
     { name = "marshmallow", specifier = ">=3.13,<4.0" },
+    { name = "mdurl", specifier = "==0.1.2" },
     { name = "nostr", specifier = ">=0.0.2" },
     { name = "python-json-logger", specifier = ">=2.0.0" },
     { name = "secp256k1", git = "https://github.com/saschanaz/secp256k1-py?branch=upgrade060" },