fix provider model pricing

Allow multiple providers per base URL

docs/api/endpoints.md

@@ -434,6 +434,42 @@ Authorization: Bearer sk-...
 }
 ```
 
+### Create Child Key
+
+Creates one or more child API keys that share the parent's balance. Each child key creation costs a fixed amount (configurable).
+
+```http
+POST /v1/balance/child-key
+Authorization: Bearer sk-...
+```
+
+**Request Body:**
+
+```json
+{
+  "count": 1
+}
+```
+
+**Parameters:**
+
+| Parameter | Type | Required | Default | Description |
+|-----------|------|----------|---------|-------------|
+| `count` | integer | Yes | - | Number of child keys to create (1-50) |
+
+**Response:**
+
+```json
+{
+  "api_keys": ["sk-abc...", "sk-def..."],
+  "count": 2,
+  "cost_msats": 2000,
+  "cost_sats": 2,
+  "parent_balance": 98000,
+  "parent_balance_sats": 98
+}
+```
+
 ## Provider Discovery
 
 ## Admin Settings

migrations/versions/c2d3e4f5a6b7_upstream_provider_base_url_api_key_unique.py

@@ -0,0 +1,118 @@
+"""make upstream provider base_url + api_key unique
+
+Revision ID: c2d3e4f5a6b7
+Revises: a86e5348850b
+Create Date: 2026-01-25 00:00:00.000000
+"""
+
+from __future__ import annotations
+
+import sqlalchemy as sa
+from alembic import op
+
+revision = "c2d3e4f5a6b7"
+down_revision = "a86e5348850b"
+branch_labels = None
+depends_on = None
+
+
+def _recreate_table_sqlite(add_base_url_unique: bool) -> None:
+    conn = op.get_bind()
+    existing_tables = {
+        row[0]
+        for row in conn.exec_driver_sql(
+            "SELECT name FROM sqlite_master WHERE type='table'"
+        ).fetchall()
+    }
+    if "upstream_providers_old" in existing_tables:
+        if "upstream_providers" in existing_tables:
+            op.drop_table("upstream_providers_old")
+        else:
+            op.execute(
+                "ALTER TABLE upstream_providers_old RENAME TO upstream_providers"
+            )
+            existing_tables.add("upstream_providers")
+    if "upstream_providers" not in existing_tables:
+        return
+
+    constraints = [
+        sa.UniqueConstraint(
+            "base_url",
+            "api_key",
+            name="uq_upstream_providers_base_url_api_key",
+        )
+    ]
+    if add_base_url_unique:
+        constraints.append(
+            sa.UniqueConstraint("base_url", name="uq_upstream_providers_base_url")
+        )
+
+    op.execute("ALTER TABLE upstream_providers RENAME TO upstream_providers_old")
+    op.create_table(
+        "upstream_providers",
+        sa.Column(
+            "id", sa.Integer(), primary_key=True, nullable=False, autoincrement=True
+        ),
+        sa.Column("provider_type", sa.String(), nullable=False),
+        sa.Column("base_url", sa.String(), nullable=False),
+        sa.Column("api_key", sa.String(), nullable=False),
+        sa.Column("api_version", sa.String(), nullable=True),
+        sa.Column("enabled", sa.Boolean(), nullable=False),
+        sa.Column("provider_fee", sa.Float(), nullable=False, server_default="1.01"),
+        *constraints,
+    )
+    op.execute(
+        "INSERT INTO upstream_providers (id, provider_type, base_url, api_key, api_version, enabled, provider_fee) "
+        "SELECT id, provider_type, base_url, api_key, api_version, enabled, provider_fee "
+        "FROM upstream_providers_old"
+    )
+    op.drop_table("upstream_providers_old")
+
+
+def upgrade() -> None:
+    conn = op.get_bind()
+    if conn.dialect.name == "sqlite":
+        _recreate_table_sqlite(add_base_url_unique=False)
+        return
+
+    inspector = sa.inspect(conn)
+    for constraint in inspector.get_unique_constraints("upstream_providers"):
+        name = constraint.get("name")
+        if constraint.get("column_names") == ["base_url"] and name:
+            op.drop_constraint(
+                name,
+                "upstream_providers",
+                type_="unique",
+            )
+    index_names = {idx["name"] for idx in inspector.get_indexes("upstream_providers")}
+    if "ix_upstream_providers_base_url" in index_names:
+        op.drop_index("ix_upstream_providers_base_url", table_name="upstream_providers")
+    op.create_unique_constraint(
+        "uq_upstream_providers_base_url_api_key",
+        "upstream_providers",
+        ["base_url", "api_key"],
+    )
+
+
+def downgrade() -> None:
+    conn = op.get_bind()
+    if conn.dialect.name == "sqlite":
+        _recreate_table_sqlite(add_base_url_unique=True)
+        return
+
+    op.drop_constraint(
+        "uq_upstream_providers_base_url_api_key",
+        "upstream_providers",
+        type_="unique",
+    )
+    op.create_unique_constraint(
+        "uq_upstream_providers_base_url",
+        "upstream_providers",
+        ["base_url"],
+    )
+    op.create_index(
+        "ix_upstream_providers_base_url",
+        "upstream_providers",
+        ["base_url"],
+        unique=True,
+    )

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "routstr"
-version = "0.2.2"
+version = "0.3.0"
 description = "Payment proxy for your LLM endpoint using cashu and nostr."
 readme = "README.md"
 requires-python = ">=3.11"

routstr/balance.py

@@ -251,12 +251,24 @@ async def donate(token: str, ref: str | None = None) -> str:
         return "Invalid token."
 
 
+class ChildKeyRequest(BaseModel):
+    count: int
+
+
 @router.post("/child-key")
 async def create_child_key(
+    payload: ChildKeyRequest,
     key: ApiKey = Depends(get_key_from_header),
     session: AsyncSession = Depends(get_session),
 ) -> dict:
-    """Creates a child API key that uses the parent's balance."""
+    """Creates one or more child API keys that use the parent's balance."""
+    # Log incoming request for debugging
+    logger.debug(f"Child key creation request: count={payload.count}")
+
+    count = payload.count
+    if count < 1 or count > 50:
+        raise HTTPException(status_code=400, detail="Count must be between 1 and 50.")
+
     # Check if this is already a child key
     if key.parent_key_hash:
         raise HTTPException(
@@ -264,38 +276,48 @@ async def create_child_key(
             detail="Cannot create a child key for another child key.",
         )
 
-    cost = settings.child_key_cost
+    cost_per_key = settings.child_key_cost
+    total_cost = cost_per_key * count
 
-    if key.total_balance < cost:
+    if key.total_balance < total_cost:
         raise HTTPException(
             status_code=402,
-            detail=f"Insufficient balance to create child key. {cost} mSats required.",
+            detail=f"Insufficient balance to create {count} child keys. {total_cost} mSats required.",
         )
 
     # Deduct cost from parent
-    key.balance -= cost
-    key.total_spent += cost
+    key.balance -= total_cost
+    key.total_spent += total_cost
     session.add(key)
 
-    # Generate new key
+    # Generate new keys
     import secrets
 
-    new_key_raw = secrets.token_hex(32)
-    new_key_hash = new_key_raw  # We use the raw key as the hash for sk- keys
+    new_keys = []
+    for _ in range(count):
+        new_key_raw = secrets.token_hex(32)
+        new_key_hash = new_key_raw  # We use the raw key as the hash for sk- keys
+
+        child_key = ApiKey(
+            hashed_key=new_key_hash,
+            balance=0,
+            parent_key_hash=key.hashed_key,
+        )
+        session.add(child_key)
+        new_keys.append("sk-" + new_key_hash)
 
-    child_key = ApiKey(
-        hashed_key=new_key_hash,
-        balance=0,
-        parent_key_hash=key.hashed_key,
-    )
-    session.add(child_key)
     await session.commit()
 
-    return {
-        "api_key": "sk-" + new_key_hash,
-        "cost_msats": cost,
+    response_data = {
+        "api_keys": new_keys,
+        "count": count,
+        "cost_msats": total_cost,
+        "cost_sats": total_cost // 1000,
         "parent_balance": key.balance,
+        "parent_balance_sats": key.balance // 1000,
     }
+    logger.debug(f"Child key creation response: {response_data}")
+    return response_data
 
 
 @router.api_route(

routstr/core/admin.py

@@ -2516,7 +2516,7 @@ async def get_provider_model(provider_id: int, model_id: str) -> dict[str, objec
                 status_code=404, detail="Model not found for this provider"
             )
         return _row_to_model(
-            row, apply_provider_fee=True, provider_fee=provider.provider_fee
+            row, apply_provider_fee=False, provider_fee=provider.provider_fee
         ).dict()  # type: ignore
 
 
@@ -2598,12 +2598,14 @@ async def create_upstream_provider(
     async with create_session() as session:
         result = await session.exec(
             select(UpstreamProviderRow).where(
-                UpstreamProviderRow.base_url == payload.base_url
+                UpstreamProviderRow.base_url == payload.base_url,
+                UpstreamProviderRow.api_key == payload.api_key,
             )
         )
         if result.first():
             raise HTTPException(
-                status_code=409, detail="Provider with this base URL already exists"
+                status_code=409,
+                detail="Provider with this base URL and API key already exists",
             )
 
         provider = UpstreamProviderRow(
@@ -2727,7 +2729,10 @@ async def get_provider_models(provider_id: int) -> dict[str, object]:
             raise HTTPException(status_code=404, detail="Provider not found")
 
         db_models = await list_models(
-            session=session, upstream_id=provider_id, include_disabled=True
+            session=session,
+            upstream_id=provider_id,
+            include_disabled=True,
+            apply_fees=False,
         )
 
         upstream_models = []
@@ -2735,10 +2740,7 @@ async def get_provider_models(provider_id: int) -> dict[str, object]:
         if upstream_instance:
             try:
                 raw_models = await upstream_instance.fetch_models()
-                upstream_models = [
-                    upstream_instance._apply_provider_fee_to_model(m)
-                    for m in raw_models
-                ]
+                upstream_models = raw_models
             except Exception as e:
                 logger.error(
                     f"Failed to fetch models from {provider.provider_type}: {e}"

routstr/core/db.py

@@ -5,6 +5,7 @@ from typing import AsyncGenerator
 
 from alembic import command
 from alembic.config import Config
+from sqlalchemy import UniqueConstraint
 from sqlalchemy.ext.asyncio.engine import create_async_engine
 from sqlmodel import Field, Relationship, SQLModel, func, select, update
 from sqlmodel.ext.asyncio.session import AsyncSession
@@ -111,11 +112,14 @@ class LightningInvoice(SQLModel, table=True):  # type: ignore
 
 class UpstreamProviderRow(SQLModel, table=True):  # type: ignore
     __tablename__ = "upstream_providers"
+    __table_args__ = (
+        UniqueConstraint("base_url", "api_key", name="uq_upstream_providers_base_url_api_key"),
+    )
     id: int | None = Field(default=None, primary_key=True)
     provider_type: str = Field(
         description="Provider type: custom, openai, anthropic, azure, openrouter, etc."
     )
-    base_url: str = Field(unique=True, description="Base URL of the upstream API")
+    base_url: str = Field(description="Base URL of the upstream API")
     api_key: str = Field(description="API key for the upstream provider")
     api_version: str | None = Field(
         default=None, description="API version for Azure OpenAI"

routstr/core/main.py

@@ -30,9 +30,9 @@ setup_logging()
 logger = get_logger(__name__)
 
 if os.getenv("VERSION_SUFFIX") is not None:
-    __version__ = f"0.2.2-{os.getenv('VERSION_SUFFIX')}"
+    __version__ = f"0.3.0-{os.getenv('VERSION_SUFFIX')}"
 else:
-    __version__ = "0.2.2"
+    __version__ = "0.3.0"
 
 
 @asynccontextmanager
@@ -188,6 +188,7 @@ async def info() -> dict:
         "mints": global_settings.cashu_mints,
         "http_url": global_settings.http_url,
         "onion_url": global_settings.onion_url,
+        "child_key_cost_msats": global_settings.child_key_cost,
     }
 
 

routstr/payment/cost_calculation.py

@@ -15,6 +15,7 @@ class CostData(BaseModel):
     input_msats: int
     output_msats: int
     total_msats: int
+    total_usd: float = 0.0
 
 
 class MaxCostData(CostData):
@@ -61,6 +62,7 @@ async def calculate_cost(  # todo: can be sync
             input_msats=0,
             output_msats=0,
             total_msats=0,
+            total_usd=0.0,
         )
 
     usage_data = response_data["usage"]
@@ -101,6 +103,7 @@ async def calculate_cost(  # todo: can be sync
                 input_msats=-1,  # Cost field doesn't break down by token type
                 output_msats=-1,
                 total_msats=cost_in_msats,
+                total_usd=usd_cost,
             )
         except Exception as e:
             logger.warning(
@@ -210,6 +213,7 @@ async def calculate_cost(  # todo: can be sync
 
     output_msats = round(output_tokens / 1000 * MSATS_PER_1K_OUTPUT_TOKENS, 3)
     token_based_cost = math.ceil(input_msats + output_msats)
+    total_usd = (token_based_cost / 1000.0) * sats_usd_price()
 
     logger.info(
         "Calculated token-based cost",
@@ -219,6 +223,7 @@ async def calculate_cost(  # todo: can be sync
             "input_cost_msats": input_msats,
             "output_cost_msats": output_msats,
             "total_cost_msats": token_based_cost,
+            "total_usd": total_usd,
             "model": response_data.get("model", "unknown"),
         },
     )
@@ -228,4 +233,5 @@ async def calculate_cost(  # todo: can be sync
         input_msats=int(input_msats),
         output_msats=int(output_msats),
         total_msats=token_based_cost,
+        total_usd=total_usd,
     )

routstr/payment/models.py

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

routstr/upstream/base.py

@@ -451,164 +451,111 @@ class BaseUpstreamProvider:
         async def stream_with_cost(
             max_cost_for_model: int,
         ) -> AsyncGenerator[bytes, None]:
-            stored_chunks: list[bytes] = []
             usage_finalized: bool = False
             last_model_seen: str | None = None
+            usage_chunk_data: dict | None = None
+            done_seen: bool = False
 
-            async def finalize_without_usage() -> bytes | None:
+            async def finalize_db_only() -> None:
                 nonlocal usage_finalized
                 if usage_finalized:
-                    return None
+                    return
                 async with create_session() as new_session:
                     fresh_key = await new_session.get(key.__class__, key.hashed_key)
                     if not fresh_key:
-                        logger.warning(
-                            "Key not found when finalizing streaming payment",
-                            extra={"key_hash": key.hashed_key[:8] + "..."},
-                        )
-                        usage_finalized = True
-                        return None
+                        return
                     try:
-                        fallback: dict = {
-                            "model": last_model_seen or "unknown",
-                            "usage": None,
-                        }
-                        cost_data = await adjust_payment_for_tokens(
-                            fresh_key, fallback, new_session, max_cost_for_model
+                        await adjust_payment_for_tokens(
+                            fresh_key,
+                            {"model": last_model_seen or "unknown", "usage": None},
+                            new_session,
+                            max_cost_for_model,
                         )
                         usage_finalized = True
-                        logger.info(
-                            "Finalized streaming payment without explicit usage",
-                            extra={
-                                "key_hash": key.hashed_key[:8] + "...",
-                                "cost_data": cost_data,
-                                "balance_after_adjustment": fresh_key.balance,
-                            },
-                        )
-                        return f"data: {json.dumps({'cost': cost_data})}\n\n".encode()
-                    except Exception as cost_error:
-                        logger.error(
-                            "Error finalizing payment without usage",
-                            extra={
-                                "error": str(cost_error),
-                                "error_type": type(cost_error).__name__,
-                                "key_hash": key.hashed_key[:8] + "...",
-                            },
-                        )
-                        usage_finalized = True
-                        return None
-
-            try:
-                async for chunk in response.aiter_bytes():
-                    stored_chunks.append(chunk)
-                    try:
-                        for part in re.split(b"data: ", chunk):
-                            if not part or part.strip() in (b"[DONE]", b""):
-                                continue
-                            try:
-                                obj = json.loads(part)
-                                if isinstance(obj, dict) and obj.get("model"):
-                                    last_model_seen = str(obj.get("model"))
-                            except json.JSONDecodeError:
-                                pass
                     except Exception:
                         pass
 
-                    yield chunk
+            try:
+                async for chunk in response.aiter_bytes():
+                    # Split chunk into SSE events
+                    parts = re.split(b"data: ", chunk)
+                    for i, part in enumerate(parts):
+                        if not part:
+                            continue
 
-                logger.debug(
-                    "Streaming completed, analyzing usage data",
-                    extra={
-                        "key_hash": key.hashed_key[:8] + "...",
-                        "chunks_count": len(stored_chunks),
-                    },
-                )
+                        stripped_part = part.strip()
+                        if not stripped_part:
+                            continue
 
-                for i in range(len(stored_chunks) - 1, -1, -1):
-                    chunk = stored_chunks[i]
-                    if not chunk:
-                        continue
-                    try:
-                        events = re.split(b"data: ", chunk)
-                        for event_data in events:
-                            if not event_data or event_data.strip() in (b"[DONE]", b""):
-                                continue
-                            try:
-                                data = json.loads(event_data)
-                                if isinstance(data, dict) and data.get("model"):
-                                    last_model_seen = str(data.get("model"))
-                                if isinstance(data, dict) and isinstance(
-                                    data.get("usage"), dict
-                                ):
-                                    async with create_session() as new_session:
-                                        fresh_key = await new_session.get(
-                                            key.__class__, key.hashed_key
-                                        )
-                                        if fresh_key:
-                                            try:
-                                                cost_data = (
-                                                    await adjust_payment_for_tokens(
-                                                        fresh_key,
-                                                        data,
-                                                        new_session,
-                                                        max_cost_for_model,
-                                                    )
-                                                )
-                                                usage_finalized = True
-                                                logger.info(
-                                                    "Payment adjustment completed for streaming",
-                                                    extra={
-                                                        "key_hash": key.hashed_key[:8]
-                                                        + "...",
-                                                        "cost_data": cost_data,
-                                                        "model": last_model_seen,
-                                                        "balance_after_adjustment": fresh_key.balance,
-                                                    },
-                                                )
-                                                yield f"data: {json.dumps({'cost': cost_data})}\n\n".encode()
-                                            except Exception as cost_error:
-                                                logger.error(
-                                                    "Error adjusting payment for streaming tokens",
-                                                    extra={
-                                                        "error": str(cost_error),
-                                                        "error_type": type(
-                                                            cost_error
-                                                        ).__name__,
-                                                        "key_hash": key.hashed_key[:8]
-                                                        + "...",
-                                                    },
-                                                )
-                                    break
-                            except json.JSONDecodeError:
-                                continue
-                    except Exception as e:
-                        logger.error(
-                            "Error processing streaming response chunk",
-                            extra={
-                                "error": str(e),
-                                "error_type": type(e).__name__,
-                                "key_hash": key.hashed_key[:8] + "...",
-                            },
+                        if stripped_part == b"[DONE]":
+                            done_seen = True
+                            continue
+
+                        try:
+                            obj = json.loads(part)
+                            if isinstance(obj, dict):
+                                if obj.get("model"):
+                                    last_model_seen = str(obj.get("model"))
+
+                                if isinstance(obj.get("usage"), dict):
+                                    # Hold this chunk back to merge cost later
+                                    usage_chunk_data = obj
+                                    continue
+                        except json.JSONDecodeError:
+                            pass
+
+                        prefix = (
+                            b"data: " if (i > 0 or chunk.startswith(b"data: ")) else b""
                         )
+                        yield prefix + part
+
+                # Stream finished, process usage if found
+                if usage_chunk_data:
+                    async with create_session() as session:
+                        fresh_key = await session.get(key.__class__, key.hashed_key)
+                        if fresh_key:
+                            try:
+                                cost_data = await adjust_payment_for_tokens(
+                                    fresh_key,
+                                    usage_chunk_data,
+                                    session,
+                                    max_cost_for_model,
+                                )
+                                # Merge cost into usage
+                                usage_chunk_data["usage"]["cost"] = cost_data.get(
+                                    "total_usd", 0.0
+                                )
+                                # Keep detailed cost in metadata
+                                usage_chunk_data["metadata"] = usage_chunk_data.get(
+                                    "metadata", {}
+                                )
+                                usage_chunk_data["metadata"]["routstr"] = {
+                                    "cost": cost_data
+                                }
+                                yield f"data: {json.dumps(usage_chunk_data)}\n\n".encode()
+                                usage_finalized = True
+                            except Exception:
+                                # Fallback: yield original usage chunk if adjustment fails
+                                yield f"data: {json.dumps(usage_chunk_data)}\n\n".encode()
 
                 if not usage_finalized:
-                    maybe_cost_event = await finalize_without_usage()
-                    if maybe_cost_event is not None:
-                        yield maybe_cost_event
+                    await finalize_db_only()
+
+                if done_seen:
+                    yield b"data: [DONE]\n\n"
 
             except Exception as stream_error:
                 logger.warning(
-                    "Streaming interrupted; finalizing without usage",
+                    "Streaming interrupted; finalizing in background",
                     extra={
                         "error": str(stream_error),
-                        "error_type": type(stream_error).__name__,
                         "key_hash": key.hashed_key[:8] + "...",
                     },
                 )
                 raise
             finally:
                 if not usage_finalized:
-                    await finalize_without_usage()
+                    await finalize_db_only()
 
         # Remove inaccurate encoding headers from upstream response
         response_headers = dict(response.headers)
@@ -648,6 +595,7 @@ class BaseUpstreamProvider:
             },
         )
 
+        content: bytes | None = None
         try:
             content = await response.aread()
             response_json = json.loads(content)
@@ -664,6 +612,14 @@ class BaseUpstreamProvider:
             cost_data = await adjust_payment_for_tokens(
                 key, response_json, session, deducted_max_cost
             )
+
+            # Merge cost into usage for OpenCode
+            if "usage" in response_json:
+                response_json["usage"]["cost"] = cost_data.get("total_usd", 0.0)
+
+            # Keep detailed cost
+            response_json["metadata"] = response_json.get("metadata", {})
+            response_json["metadata"]["routstr"] = {"cost": cost_data}
             response_json["cost"] = cost_data
 
             logger.info(
@@ -749,180 +705,135 @@ class BaseUpstreamProvider:
         async def stream_with_responses_cost(
             max_cost_for_model: int,
         ) -> AsyncGenerator[bytes, None]:
-            stored_chunks: list[bytes] = []
             usage_finalized: bool = False
             last_model_seen: str | None = None
             reasoning_tokens: int = 0
+            usage_chunk_data: dict | None = None
+            done_seen: bool = False
 
-            async def finalize_without_usage() -> bytes | None:
+            async def finalize_db_only() -> None:
                 nonlocal usage_finalized
                 if usage_finalized:
-                    return None
+                    return
                 async with create_session() as new_session:
                     fresh_key = await new_session.get(key.__class__, key.hashed_key)
                     if not fresh_key:
-                        logger.warning(
-                            "Key not found when finalizing Responses API streaming payment",
-                            extra={"key_hash": key.hashed_key[:8] + "..."},
-                        )
-                        usage_finalized = True
-                        return None
+                        return
                     try:
-                        fallback: dict = {
-                            "model": last_model_seen or "unknown",
-                            "usage": None,
-                        }
-                        cost_data = await adjust_payment_for_tokens(
-                            fresh_key, fallback, new_session, max_cost_for_model
+                        await adjust_payment_for_tokens(
+                            fresh_key,
+                            {"model": last_model_seen or "unknown", "usage": None},
+                            new_session,
+                            max_cost_for_model,
                         )
                         usage_finalized = True
-                        logger.info(
-                            "Finalized Responses API streaming payment without explicit usage",
-                            extra={
-                                "key_hash": key.hashed_key[:8] + "...",
-                                "cost_data": cost_data,
-                                "balance_after_adjustment": fresh_key.balance,
-                            },
-                        )
-                        return f"data: {json.dumps({'cost': cost_data})}\\n\\n".encode()
-                    except Exception as cost_error:
-                        logger.error(
-                            "Error finalizing Responses API payment without usage",
-                            extra={
-                                "error": str(cost_error),
-                                "error_type": type(cost_error).__name__,
-                                "key_hash": key.hashed_key[:8] + "...",
-                            },
-                        )
-                        usage_finalized = True
-                        return None
-
-            try:
-                async for chunk in response.aiter_bytes():
-                    stored_chunks.append(chunk)
-                    try:
-                        for part in re.split(b"data: ", chunk):
-                            if not part or part.strip() in (b"[DONE]", b""):
-                                continue
-                            try:
-                                obj = json.loads(part)
-                                if isinstance(obj, dict):
-                                    if obj.get("model"):
-                                        last_model_seen = str(obj.get("model"))
-
-                                    # Track reasoning tokens for Responses API
-                                    if usage := obj.get("usage", {}):
-                                        if (
-                                            isinstance(usage, dict)
-                                            and "reasoning_tokens" in usage
-                                        ):
-                                            reasoning_tokens += usage.get(
-                                                "reasoning_tokens", 0
-                                            )
-                            except json.JSONDecodeError:
-                                pass
                     except Exception:
                         pass
 
-                    yield chunk
+            try:
+                async for chunk in response.aiter_bytes():
+                    # Split chunk into SSE events
+                    parts = re.split(b"data: ", chunk)
+                    for i, part in enumerate(parts):
+                        if not part:
+                            continue
 
-                logger.debug(
-                    "Responses API streaming completed, analyzing usage data",
-                    extra={
-                        "key_hash": key.hashed_key[:8] + "...",
-                        "chunks_count": len(stored_chunks),
-                        "reasoning_tokens": reasoning_tokens,
-                    },
-                )
+                        stripped_part = part.strip()
+                        if not stripped_part:
+                            continue
 
-                # Process final usage data
-                for i in range(len(stored_chunks) - 1, -1, -1):
-                    chunk = stored_chunks[i]
-                    if not chunk:
-                        continue
-                    try:
-                        events = re.split(b"data: ", chunk)
-                        for event_data in events:
-                            if not event_data or event_data.strip() in (b"[DONE]", b""):
-                                continue
-                            try:
-                                data = json.loads(event_data)
-                                if isinstance(data, dict) and data.get("model"):
-                                    last_model_seen = str(data.get("model"))
-                                if isinstance(data, dict) and isinstance(
-                                    data.get("usage"), dict
-                                ):
-                                    # Include reasoning tokens in usage calculation
-                                    async with create_session() as new_session:
-                                        fresh_key = await new_session.get(
-                                            key.__class__, key.hashed_key
+                        if stripped_part == b"[DONE]":
+                            done_seen = True
+                            continue
+
+                        try:
+                            obj = json.loads(part)
+                            if isinstance(obj, dict):
+                                if obj.get("model"):
+                                    last_model_seen = str(obj.get("model"))
+
+                                # Track reasoning tokens for Responses API
+                                if usage := obj.get("usage", {}):
+                                    if (
+                                        isinstance(usage, dict)
+                                        and "reasoning_tokens" in usage
+                                    ):
+                                        reasoning_tokens += usage.get(
+                                            "reasoning_tokens", 0
                                         )
-                                        if fresh_key:
-                                            try:
-                                                cost_data = (
-                                                    await adjust_payment_for_tokens(
-                                                        fresh_key,
-                                                        data,
-                                                        new_session,
-                                                        max_cost_for_model,
-                                                    )
-                                                )
-                                                usage_finalized = True
-                                                logger.info(
-                                                    "Payment adjustment completed for Responses API streaming",
-                                                    extra={
-                                                        "key_hash": key.hashed_key[:8]
-                                                        + "...",
-                                                        "cost_data": cost_data,
-                                                        "model": last_model_seen,
-                                                        "reasoning_tokens": reasoning_tokens,
-                                                        "balance_after_adjustment": fresh_key.balance,
-                                                    },
-                                                )
-                                                yield f"data: {json.dumps({'cost': cost_data})}\\n\\n".encode()
-                                            except Exception as cost_error:
-                                                logger.error(
-                                                    "Error adjusting payment for Responses API streaming tokens",
-                                                    extra={
-                                                        "error": str(cost_error),
-                                                        "error_type": type(
-                                                            cost_error
-                                                        ).__name__,
-                                                        "key_hash": key.hashed_key[:8]
-                                                        + "...",
-                                                    },
-                                                )
-                                    break
-                            except json.JSONDecodeError:
-                                continue
-                    except Exception as e:
-                        logger.error(
-                            "Error processing Responses API streaming response chunk",
-                            extra={
-                                "error": str(e),
-                                "error_type": type(e).__name__,
-                                "key_hash": key.hashed_key[:8] + "...",
-                            },
+
+                                # Responses API usage is in response.completed/incomplete events
+                                chunk_type = obj.get("type", "")
+                                if chunk_type in (
+                                    "response.completed",
+                                    "response.incomplete",
+                                ):
+                                    usage_chunk_data = obj
+                                    continue
+                        except json.JSONDecodeError:
+                            pass
+
+                        prefix = (
+                            b"data: " if (i > 0 or chunk.startswith(b"data: ")) else b""
                         )
+                        yield prefix + part
+
+                # Stream finished, process usage if found
+                if usage_chunk_data:
+                    async with create_session() as session:
+                        fresh_key = await session.get(key.__class__, key.hashed_key)
+                        if fresh_key:
+                            try:
+                                cost_data = await adjust_payment_for_tokens(
+                                    fresh_key,
+                                    usage_chunk_data,
+                                    session,
+                                    max_cost_for_model,
+                                )
+                                # Merge cost into usage chunk
+                                if (
+                                    "response" in usage_chunk_data
+                                    and "usage" in usage_chunk_data["response"]
+                                ):
+                                    usage_chunk_data["response"]["usage"]["cost"] = (
+                                        cost_data.get("total_usd", 0.0)
+                                    )
+                                elif "usage" in usage_chunk_data:
+                                    usage_chunk_data["usage"]["cost"] = cost_data.get(
+                                        "total_usd", 0.0
+                                    )
+
+                                # Keep detailed cost in metadata
+                                usage_chunk_data["metadata"] = usage_chunk_data.get(
+                                    "metadata", {}
+                                )
+                                usage_chunk_data["metadata"]["routstr"] = {
+                                    "cost": cost_data
+                                }
+                                yield f"data: {json.dumps(usage_chunk_data)}\n\n".encode()
+                                usage_finalized = True
+                            except Exception:
+                                # Fallback: yield original usage chunk if adjustment fails
+                                yield f"data: {json.dumps(usage_chunk_data)}\n\n".encode()
 
                 if not usage_finalized:
-                    maybe_cost_event = await finalize_without_usage()
-                    if maybe_cost_event is not None:
-                        yield maybe_cost_event
+                    await finalize_db_only()
+
+                if done_seen:
+                    yield b"data: [DONE]\n\n"
 
             except Exception as stream_error:
                 logger.warning(
-                    "Responses API streaming interrupted; finalizing without usage",
+                    "Responses API streaming interrupted; finalizing in background",
                     extra={
                         "error": str(stream_error),
-                        "error_type": type(stream_error).__name__,
                         "key_hash": key.hashed_key[:8] + "...",
                     },
                 )
                 raise
             finally:
                 if not usage_finalized:
-                    await finalize_without_usage()
+                    await finalize_db_only()
 
         # Remove inaccurate encoding headers from upstream response
         response_headers = dict(response.headers)
@@ -962,6 +873,7 @@ class BaseUpstreamProvider:
             },
         )
 
+        content: bytes | None = None
         try:
             content = await response.aread()
             response_json = json.loads(content)
@@ -981,6 +893,14 @@ class BaseUpstreamProvider:
             cost_data = await adjust_payment_for_tokens(
                 key, response_json, session, deducted_max_cost
             )
+
+            # Merge cost into usage for OpenCode
+            if "usage" in response_json:
+                response_json["usage"]["cost"] = cost_data.get("total_usd", 0.0)
+
+            # Keep detailed cost
+            response_json["metadata"] = response_json.get("metadata", {})
+            response_json["metadata"]["routstr"] = {"cost": cost_data}
             response_json["cost"] = cost_data
 
             logger.info(

routstr/upstream/helpers.py

@@ -236,7 +236,7 @@ async def _seed_providers_from_settings(
     from . import upstream_provider_classes
 
     providers_to_add: list[UpstreamProviderRow] = []
-    seeded_base_urls: set[str] = set()
+    seeded_provider_keys: set[tuple[str, str]] = set()
 
     provider_classes_by_type = {
         cls.provider_type: cls
@@ -261,7 +261,8 @@ async def _seed_providers_from_settings(
                 base_url = provider_class.default_base_url  # type: ignore[attr-defined]
                 result = await session.exec(
                     select(UpstreamProviderRow).where(
-                        UpstreamProviderRow.base_url == base_url
+                        UpstreamProviderRow.base_url == base_url,
+                        UpstreamProviderRow.api_key == api_key,
                     )
                 )
                 if not result.first():
@@ -273,13 +274,15 @@ async def _seed_providers_from_settings(
                             enabled=True,
                         )
                     )
-                    seeded_base_urls.add(base_url)
+                    seeded_provider_keys.add((base_url, api_key))
 
     ollama_base_url = os.environ.get("OLLAMA_BASE_URL")
     if ollama_base_url:
+        ollama_api_key = os.environ.get("OLLAMA_API_KEY", "")
         result = await session.exec(
             select(UpstreamProviderRow).where(
-                UpstreamProviderRow.base_url == ollama_base_url
+                UpstreamProviderRow.base_url == ollama_base_url,
+                UpstreamProviderRow.api_key == ollama_api_key,
             )
         )
         if not result.first():
@@ -287,18 +290,20 @@ async def _seed_providers_from_settings(
                 UpstreamProviderRow(
                     provider_type="ollama",
                     base_url=ollama_base_url,
-                    api_key=os.environ.get("OLLAMA_API_KEY", ""),
+                    api_key=ollama_api_key,
                     enabled=True,
                 )
             )
-            seeded_base_urls.add(ollama_base_url)
+            seeded_provider_keys.add((ollama_base_url, ollama_api_key))
 
     if settings.chat_completions_api_version and settings.upstream_base_url:
         base_url = settings.upstream_base_url
-        if base_url not in seeded_base_urls:
+        api_key = settings.upstream_api_key
+        if (base_url, api_key) not in seeded_provider_keys:
             result = await session.exec(
                 select(UpstreamProviderRow).where(
-                    UpstreamProviderRow.base_url == base_url
+                    UpstreamProviderRow.base_url == base_url,
+                    UpstreamProviderRow.api_key == api_key,
                 )
             )
             if not result.first():
@@ -306,19 +311,21 @@ async def _seed_providers_from_settings(
                     UpstreamProviderRow(
                         provider_type="azure",
                         base_url=base_url,
-                        api_key=settings.upstream_api_key,
+                        api_key=api_key,
                         api_version=settings.chat_completions_api_version,
                         enabled=True,
                     )
                 )
-                seeded_base_urls.add(base_url)
+                seeded_provider_keys.add((base_url, api_key))
 
     if settings.upstream_base_url and settings.upstream_api_key:
         base_url = settings.upstream_base_url
-        if base_url not in seeded_base_urls:
+        api_key = settings.upstream_api_key
+        if (base_url, api_key) not in seeded_provider_keys:
             result = await session.exec(
                 select(UpstreamProviderRow).where(
-                    UpstreamProviderRow.base_url == base_url
+                    UpstreamProviderRow.base_url == base_url,
+                    UpstreamProviderRow.api_key == api_key,
                 )
             )
             if not result.first():
@@ -326,11 +333,11 @@ async def _seed_providers_from_settings(
                     UpstreamProviderRow(
                         provider_type="custom",
                         base_url=base_url,
-                        api_key=settings.upstream_api_key,
+                        api_key=api_key,
                         enabled=True,
                     )
                 )
-                seeded_base_urls.add(base_url)
+                seeded_provider_keys.add((base_url, api_key))
 
     for provider in providers_to_add:
         session.add(provider)

routstr/upstream/ppqai.py

@@ -210,8 +210,8 @@ class PPQAIUpstreamProvider(BaseUpstreamProvider):
 
             async with create_session() as session:
                 statement = select(UpstreamProviderRow).where(
-                    UpstreamProviderRow.base_url == self.base_url
-                    and UpstreamProviderRow.api_key == self.api_key
+                    UpstreamProviderRow.base_url == self.base_url,
+                    UpstreamProviderRow.api_key == self.api_key,
                 )
                 result = await session.exec(statement)
                 provider = result.first()

tests/integration/test_child_keys.py

@@ -6,7 +6,7 @@ from fastapi import HTTPException
 from sqlmodel.ext.asyncio.session import AsyncSession
 
 from routstr.auth import adjust_payment_for_tokens, pay_for_request
-from routstr.balance import create_child_key
+from routstr.balance import ChildKeyRequest, create_child_key
 from routstr.core.db import ApiKey
 from routstr.core.settings import settings
 
@@ -27,13 +27,15 @@ async def test_child_key_flow(integration_session: AsyncSession) -> None:
     settings.child_key_cost = 1000  # 1 sat
 
     # 2. Call create_child_key
-    result = await create_child_key(parent_key, integration_session)
+    result = await create_child_key(
+        ChildKeyRequest(count=1), parent_key, integration_session
+    )
 
-    assert "api_key" in result
+    assert "api_keys" in result
     assert result["cost_msats"] == 1000
     assert result["parent_balance"] == 9000
 
-    child_key_raw = result["api_key"][3:]  # remove sk-
+    child_key_raw = result["api_keys"][0][3:]  # remove sk-
 
     # 3. Verify child key exists in DB
     child_key_db = await integration_session.get(ApiKey, child_key_raw)
@@ -111,7 +113,9 @@ async def test_child_key_insufficient_balance(
     settings.child_key_cost = 1000
 
     with pytest.raises(HTTPException) as exc:
-        await create_child_key(parent_key, integration_session)
+        await create_child_key(
+            ChildKeyRequest(count=1), parent_key, integration_session
+        )
     assert exc.value.status_code == 402
 
 
@@ -132,6 +136,6 @@ async def test_child_key_cannot_create_child(integration_session: AsyncSession)
     await integration_session.refresh(child_key)
 
     with pytest.raises(HTTPException) as exc:
-        await create_child_key(child_key, integration_session)
+        await create_child_key(ChildKeyRequest(count=1), child_key, integration_session)
     assert exc.value.status_code == 400
     assert "Cannot create a child key for another child key" in str(exc.value.detail)

ui/components/AddProviderModelDialog.tsx

@@ -248,7 +248,9 @@ export function AddProviderModelDialog({
     const pricing = model.pricing as Record<string, number>;
     const topProvider = model.top_provider as Record<string, unknown> | null;
 
-    form.setValue('id', model.id);
+    if (!isOverride) {
+      form.setValue('id', model.id);
+    }
     form.setValue('name', model.name);
     form.setValue('description', model.description || '');
     form.setValue('context_length', model.context_length);
@@ -425,7 +427,7 @@ export function AddProviderModelDialog({
           </DialogTitle>
           <DialogDescription>{description}</DialogDescription>
         </DialogHeader>
-        {!isEdit && !isOverride && (
+        {!isEdit && (
           <div className='bg-muted/30 rounded-md border p-3'>
             <div className='mb-2 text-sm font-medium'>Presets</div>
             <div className='grid gap-2 sm:grid-cols-3 sm:items-start'>
@@ -488,8 +490,9 @@ export function AddProviderModelDialog({
               </Popover>
             </div>
             <div className='text-muted-foreground mt-1 text-xs'>
-              Prefill fields from a preset model definition, then adjust as
-              needed.
+              {isOverride
+                ? 'Apply pricing and settings from a preset model (keeping the model ID unchanged).'
+                : 'Prefill fields from a preset model definition, then adjust as needed.'}
             </div>
           </div>
         )}
@@ -805,45 +808,6 @@ export function AddProviderModelDialog({
                     </FormItem>
                   )}
                 />
-                <FormField
-                  control={form.control}
-                  name='max_prompt_cost'
-                  render={({ field }) => (
-                    <FormItem>
-                      <FormLabel>Max Prompt Cost</FormLabel>
-                      <FormControl>
-                        <Input type='number' step='0.0001' {...field} />
-                      </FormControl>
-                      <FormMessage />
-                    </FormItem>
-                  )}
-                />
-                <FormField
-                  control={form.control}
-                  name='max_completion_cost'
-                  render={({ field }) => (
-                    <FormItem>
-                      <FormLabel>Max Completion Cost</FormLabel>
-                      <FormControl>
-                        <Input type='number' step='0.0001' {...field} />
-                      </FormControl>
-                      <FormMessage />
-                    </FormItem>
-                  )}
-                />
-                <FormField
-                  control={form.control}
-                  name='max_cost'
-                  render={({ field }) => (
-                    <FormItem>
-                      <FormLabel>Max Total Cost</FormLabel>
-                      <FormControl>
-                        <Input type='number' step='0.0001' {...field} />
-                      </FormControl>
-                      <FormMessage />
-                    </FormItem>
-                  )}
-                />
               </div>
             </div>
 

ui/components/child-key-creator.tsx

@@ -0,0 +1,286 @@
+'use client';
+
+import { useState } from 'react';
+import { WalletService } from '@/lib/api/services/wallet';
+import { Button } from '@/components/ui/button';
+import {
+  Card,
+  CardContent,
+  CardDescription,
+  CardHeader,
+  CardTitle,
+} from '@/components/ui/card';
+import { Alert, AlertDescription, AlertTitle } from '@/components/ui/alert';
+import { Input } from '@/components/ui/input';
+import { Key, Copy, Check, Loader2 } from 'lucide-react';
+import { toast } from 'sonner';
+
+interface ChildKeyCreatorProps {
+  baseUrl?: string;
+  apiKey?: string;
+  onApiKeyChange?: (apiKey: string) => void;
+  costPerKeyMsats?: number;
+}
+
+export function ChildKeyCreator({
+  baseUrl,
+  apiKey: propApiKey,
+  onApiKeyChange,
+  costPerKeyMsats,
+}: ChildKeyCreatorProps) {
+  const [internalApiKey, setInternalApiKey] = useState('');
+  const [loading, setLoading] = useState(false);
+  const [count, setCount] = useState(1);
+  const [newKeys, setNewKeys] = useState<string[]>([]);
+  const [resultInfo, setResultInfo] = useState<{
+    cost_msats: number;
+    parent_balance: number;
+  } | null>(null);
+  const [copiedKey, setCopiedKey] = useState<string | null>(null);
+
+  const activeApiKey = propApiKey ?? internalApiKey;
+
+  const handleApiKeyChange = (val: string) => {
+    setInternalApiKey(val);
+    onApiKeyChange?.(val);
+  };
+
+  const handleCreateKey = async () => {
+    if (!activeApiKey && baseUrl) {
+      toast.error('Please provide a Parent API key first');
+      return;
+    }
+
+    const requestedCount = Math.max(1, Math.min(50, Number(count)));
+
+    setLoading(true);
+    try {
+      const result = await WalletService.createChildKey(
+        baseUrl,
+        activeApiKey,
+        requestedCount
+      );
+
+      console.log('Created child keys:', result);
+
+      if (result.api_keys && result.api_keys.length > 0) {
+        setNewKeys(result.api_keys);
+      } else {
+        throw new Error('No API keys returned from server');
+      }
+
+      setResultInfo({
+        cost_msats: result.cost_msats,
+        parent_balance: result.parent_balance,
+      });
+
+      toast.success(
+        `${requestedCount} child API key${
+          requestedCount > 1 ? 's' : ''
+        } created successfully`
+      );
+    } catch (error) {
+      console.error('Failed to create child key:', error);
+      toast.error(
+        error instanceof Error ? error.message : 'Failed to create child key'
+      );
+    } finally {
+      setLoading(false);
+    }
+  };
+
+  const copyToClipboard = (key: string) => {
+    navigator.clipboard.writeText(key);
+    setCopiedKey(key);
+    toast.success('API key copied to clipboard');
+    setTimeout(() => setCopiedKey(null), 2000);
+  };
+
+  const copyAllToClipboard = () => {
+    navigator.clipboard.writeText(newKeys.join('\n'));
+    toast.success('All API keys copied to clipboard');
+  };
+
+  return (
+    <div className='space-y-6'>
+      <Card>
+        <CardHeader>
+          <div className='flex items-center justify-between'>
+            <div className='space-y-1'>
+              <CardTitle>Create Child API Key</CardTitle>
+              <CardDescription>
+                Generate secondary API keys that share your account balance.
+              </CardDescription>
+            </div>
+            {costPerKeyMsats !== undefined && (
+              <div className='text-right'>
+                <p className='text-muted-foreground text-[0.65rem] tracking-wide uppercase'>
+                  Unit Cost
+                </p>
+                <p className='text-primary text-sm font-bold'>
+                  {costPerKeyMsats / 1000} sats
+                </p>
+              </div>
+            )}
+          </div>
+        </CardHeader>
+        <CardContent>
+          <div className='space-y-4'>
+            {baseUrl && (
+              <div className='space-y-2'>
+                <label className='text-muted-foreground text-[0.7rem] tracking-wider uppercase'>
+                  Parent API Key
+                </label>
+                <Input
+                  value={activeApiKey}
+                  onChange={(e) => handleApiKeyChange(e.target.value)}
+                  placeholder='sk-...'
+                  className='font-mono text-sm'
+                />
+              </div>
+            )}
+
+            <div className='flex flex-col gap-4 sm:flex-row sm:items-end sm:justify-between'>
+              <div className='flex-1 space-y-2'>
+                <div className='flex items-center justify-between'>
+                  <label className='text-muted-foreground text-[0.7rem] tracking-wider uppercase'>
+                    Number of keys
+                  </label>
+                  {costPerKeyMsats && (
+                    <span className='text-muted-foreground text-[10px]'>
+                      Cost: {costPerKeyMsats * count} mSats
+                    </span>
+                  )}
+                </div>
+                <Input
+                  type='number'
+                  min={1}
+                  max={50}
+                  value={count}
+                  onChange={(e) => {
+                    const val = parseInt(e.target.value);
+                    if (!isNaN(val)) {
+                      setCount(Math.max(1, Math.min(50, val)));
+                    } else {
+                      setCount(1);
+                    }
+                  }}
+                  className='w-full sm:w-24'
+                />
+              </div>
+              <Button
+                onClick={handleCreateKey}
+                disabled={loading || (!!baseUrl && !activeApiKey)}
+                className='w-full sm:w-auto'
+              >
+                {loading ? (
+                  <>
+                    <Loader2 className='mr-2 h-4 w-4 animate-spin' />
+                    Creating...
+                  </>
+                ) : (
+                  <>
+                    <Key className='mr-2 h-4 w-4' />
+                    Generate {count > 1 ? `${count} Keys` : 'Key'}
+                  </>
+                )}
+              </Button>
+            </div>
+
+            <p className='text-muted-foreground text-xs'>
+              Each key creation has a small one-time fee.
+            </p>
+
+            {newKeys.length > 0 && (
+              <div className='mt-6 space-y-4'>
+                <Alert className='border-green-200 bg-green-50 dark:border-green-900/20 dark:bg-green-900/10'>
+                  <AlertTitle className='text-green-800 dark:text-green-400'>
+                    {newKeys.length} New API Key{newKeys.length > 1 ? 's' : ''}{' '}
+                    Generated
+                  </AlertTitle>
+                  <AlertDescription className='text-green-700 dark:text-green-500'>
+                    Copy {newKeys.length > 1 ? 'these keys' : 'this key'} now.
+                    You won&apos;t be able to see them again.
+                    {resultInfo && (
+                      <div className='mt-2 font-medium opacity-80'>
+                        Total Cost: {resultInfo.cost_msats / 1000} sats | New
+                        Balance: {resultInfo.parent_balance / 1000} sats
+                      </div>
+                    )}
+                  </AlertDescription>
+                </Alert>
+
+                <div className='space-y-2'>
+                  <div className='flex items-center justify-between'>
+                    <span className='text-muted-foreground text-xs font-medium uppercase'>
+                      Generated Keys ({newKeys.length})
+                    </span>
+                    {newKeys.length > 1 && (
+                      <Button
+                        variant='ghost'
+                        size='sm'
+                        className='h-7 text-[10px] uppercase'
+                        onClick={copyAllToClipboard}
+                      >
+                        <Copy className='mr-1 h-3 w-3' />
+                        Copy All
+                      </Button>
+                    )}
+                  </div>
+                  <div className='grid gap-2'>
+                    {newKeys.map((key, index) => (
+                      <div
+                        key={index}
+                        className='group relative flex items-center gap-2'
+                      >
+                        <code className='bg-muted/50 flex-1 rounded border p-2.5 font-mono text-[10px] break-all sm:text-xs'>
+                          {key}
+                        </code>
+                        <Button
+                          size='icon'
+                          variant='ghost'
+                          className='h-8 w-8 shrink-0'
+                          onClick={() => copyToClipboard(key)}
+                        >
+                          {copiedKey === key ? (
+                            <Check className='h-3.5 w-3.5 text-green-500' />
+                          ) : (
+                            <Copy className='h-3.5 w-3.5 opacity-50 group-hover:opacity-100' />
+                          )}
+                        </Button>
+                      </div>
+                    ))}
+                  </div>
+                </div>
+
+                {newKeys.length > 3 && (
+                  <div className='space-y-2'>
+                    <label className='text-muted-foreground text-[0.7rem] tracking-wider uppercase'>
+                      Bulk Export (All Keys)
+                    </label>
+                    <div className='relative'>
+                      <textarea
+                        readOnly
+                        value={newKeys.join('\n')}
+                        rows={Math.min(newKeys.length, 6)}
+                        className='bg-muted/30 w-full rounded-md border p-3 font-mono text-[10px] focus:outline-none'
+                      />
+                      <Button
+                        size='sm'
+                        variant='secondary'
+                        className='absolute right-2 bottom-2 h-7 text-[10px]'
+                        onClick={copyAllToClipboard}
+                      >
+                        Copy Bulk
+                      </Button>
+                    </div>
+                  </div>
+                )}
+              </div>
+            )}
+          </div>
+        </CardContent>
+      </Card>
+    </div>
+  );
+}

ui/components/landing/cheat-sheet.tsx

@@ -14,6 +14,7 @@ import { ConfigurationService } from '@/lib/api/services/configuration';
 import { CashuPaymentWorkflow } from './cashu-payment-workflow';
 import { LightningPaymentWorkflow } from './lightning-payment-workflow';
 import { ApiKeyManager } from './api-key-manager';
+import { ChildKeyCreator } from '@/components/child-key-creator';
 
 type NodeInfo = {
   name: string;
@@ -23,6 +24,7 @@ type NodeInfo = {
   mints: string[];
   http_url?: string | null;
   onion_url?: string | null;
+  child_key_cost_msats?: number;
 };
 
 type WalletSnapshot = {
@@ -362,10 +364,11 @@ export function CheatSheet(): JSX.Element {
         </section>
 
         <Tabs defaultValue='cashu' className='w-full'>
-          <TabsList className='grid w-full grid-cols-3'>
+          <TabsList className='grid w-full grid-cols-4'>
             <TabsTrigger value='cashu'>Cashu Payments</TabsTrigger>
             <TabsTrigger value='lightning'>Lightning Payments</TabsTrigger>
             <TabsTrigger value='manage'>Manage Keys</TabsTrigger>
+            <TabsTrigger value='child-keys'>Child Keys</TabsTrigger>
           </TabsList>
 
           <TabsContent value='cashu' className='space-y-4'>
@@ -422,6 +425,15 @@ export function CheatSheet(): JSX.Element {
               </Card>
             )}
           </TabsContent>
+
+          <TabsContent value='child-keys' className='space-y-4'>
+            <ChildKeyCreator
+              baseUrl={normalizedBaseUrl}
+              apiKey={apiKeyInput}
+              onApiKeyChange={handleApiKeyChanged}
+              costPerKeyMsats={nodeInfo?.child_key_cost_msats}
+            />
+          </TabsContent>
         </Tabs>
       </main>
     </div>

ui/lib/api/services/admin.ts

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

ui/lib/api/services/wallet.ts

@@ -29,6 +29,28 @@ export type RedeemTokenResponse = z.infer<typeof RedeemTokenResponseSchema>;
 export type SendTokenRequest = z.infer<typeof SendTokenRequestSchema>;
 export type SendTokenResponse = z.infer<typeof SendTokenResponseSchema>;
 
+export interface BalanceDetail {
+  mint_url: string;
+  unit: string;
+  wallet_balance: number;
+  user_balance: number;
+  owner_balance: number;
+  error?: string;
+}
+
+export interface WithdrawResponse {
+  token: string;
+}
+
+export interface CreateChildKeyResponse {
+  api_keys: string[];
+  count: number;
+  cost_msats: number;
+  cost_sats: number;
+  parent_balance: number;
+  parent_balance_sats: number;
+}
+
 export class WalletService {
   static async redeemToken(token: string): Promise<RedeemTokenResponse> {
     try {
@@ -108,17 +130,38 @@ export class WalletService {
       throw error;
     }
   }
-}
 
-export interface BalanceDetail {
-  mint_url: string;
-  unit: string;
-  wallet_balance: number;
-  user_balance: number;
-  owner_balance: number;
-  error?: string;
-}
+  static async createChildKey(
+    baseUrl?: string,
+    apiKey?: string,
+    count: number = 1
+  ): Promise<CreateChildKeyResponse> {
+    try {
+      if (baseUrl && apiKey) {
+        const response = await fetch(`${baseUrl}/v1/balance/child-key`, {
+          method: 'POST',
+          headers: {
+            'Content-Type': 'application/json',
+            Authorization: `Bearer ${apiKey}`,
+          },
+          body: JSON.stringify({ count }),
+        });
 
-export interface WithdrawResponse {
-  token: string;
+        if (!response.ok) {
+          const errorText = await response.text();
+          throw new Error(errorText || 'Failed to create child key');
+        }
+
+        return (await response.json()) as CreateChildKeyResponse;
+      }
+
+      return await apiClient.post<CreateChildKeyResponse>(
+        '/v1/balance/child-key',
+        { count }
+      );
+    } catch (error) {
+      console.error('Error creating child key:', error);
+      throw error;
+    }
+  }
 }

uv.lock

@@ -1878,7 +1878,7 @@ wheels = [
 
 [[package]]
 name = "routstr"
-version = "0.2.2"
+version = "0.3.0"
 source = { editable = "." }
 dependencies = [
     { name = "aiosqlite" },