make sure when topup from untrusted mint to refund from primary mint

pin pnpm version

routstr/balance.py

@@ -337,21 +337,26 @@ async def refund_wallet_endpoint(
         )
 
     # --- MINT: balance is locked at zero, safe to create the refund token ---
+    # Proofs from untrusted mints are swapped to primary_mint on receive.
+    # Use primary_mint unless key.refund_mint_url is an explicitly trusted mint.
+    effective_refund_mint = (
+        key.refund_mint_url
+        if key.refund_mint_url and key.refund_mint_url in settings.cashu_mints
+        else settings.primary_mint
+    )
     try:
         if key.refund_address:
-            from .core.settings import settings as global_settings
-
             await send_to_lnurl(
                 remaining_balance,
                 key.refund_currency or "sat",
-                key.refund_mint_url or global_settings.primary_mint,
+                effective_refund_mint,
                 key.refund_address,
             )
             result = {"recipient": key.refund_address}
         else:
             refund_currency = key.refund_currency or "sat"
             token = await send_token(
-                remaining_balance, refund_currency, key.refund_mint_url
+                remaining_balance, refund_currency, effective_refund_mint
             )
             result = {"token": token}
 
@@ -379,15 +384,26 @@ async def refund_wallet_endpoint(
         # Minting failed — restore the debited balance
         await _restore_balance(session, key.hashed_key, pre_debit_balance, pre_debit_reserved)
         error_msg = str(e)
+        logger.error(
+            "refund_wallet_endpoint: mint/send failed",
+            extra={
+                "error": error_msg,
+                "error_type": type(e).__name__,
+                "hashed_key": key.hashed_key,
+                "remaining_balance": remaining_balance,
+                "refund_currency": key.refund_currency,
+                "refund_mint_url": key.refund_mint_url,
+                "has_refund_address": bool(key.refund_address),
+            },
+        )
         if (
             "mint" in error_msg.lower()
             or "connection" in error_msg.lower()
-            or isinstance(e, Exception)
-            and "ConnectError" in str(type(e))
+            or "ConnectError" in str(type(e))
         ):
-            raise HTTPException(status_code=503, detail="Mint service unavailable")
+            raise HTTPException(status_code=503, detail=f"Mint service unavailable: {error_msg}")
         else:
-            raise HTTPException(status_code=500, detail="Refund failed")
+            raise HTTPException(status_code=500, detail=f"Refund failed: {error_msg}")
 
     await _refund_cache_set(bearer_value, result)
 

routstr/upstream/base.py

@@ -42,6 +42,7 @@ from ..payment.models import (
 from ..payment.price import sats_usd_price
 from ..wallet import recieve_token, send_token
 from . import messages_dispatch
+from .count_tokens import count_tokens_locally
 from .litellm_routing import detect_litellm_prefix
 
 logger = get_logger(__name__)
@@ -2196,6 +2197,12 @@ class BaseUpstreamProvider:
         """
         path = self.normalize_request_path(path, model_obj)
 
+        if (
+            path.endswith("messages/count_tokens")
+            and not self.supports_anthropic_messages
+        ):
+            return count_tokens_locally(request_body, model_obj)
+
         if (
             path.endswith("messages")
             and not path.endswith("count_tokens")
@@ -3394,6 +3401,12 @@ class BaseUpstreamProvider:
 
         request_body = await request.body()
 
+        if (
+            path.endswith("messages/count_tokens")
+            and not self.supports_anthropic_messages
+        ):
+            return count_tokens_locally(request_body, model_obj)
+
         if (
             path.endswith("messages")
             and not path.endswith("count_tokens")

routstr/upstream/count_tokens.py

@@ -0,0 +1,108 @@
+"""Local handling of Anthropic ``/v1/messages/count_tokens`` for upstreams
+that do not natively expose the endpoint.
+
+Most non-Anthropic upstreams (OpenAI-compat, Gemini OpenAI-compat,
+OpenRouter chat-completions, generic providers) return 400/404 when asked
+to ``POST /messages/count_tokens``. Claude Code and other Anthropic SDK
+clients call this endpoint before each turn to size context windows and
+trigger compaction, so a failure breaks the whole chat.
+
+We answer locally. ``litellm.token_counter`` understands the Anthropic
+message shape and the per-model tokenizers, so we prefer it. If it raises
+(unknown model, encoding lookup failure, ...), we fall back to the
+project's own ``estimate_tokens`` heuristic, which is always defined and
+never raises.
+"""
+
+from __future__ import annotations
+
+import json
+from typing import Any
+
+import litellm
+from fastapi.responses import Response
+
+from ..core import get_logger
+from ..payment.helpers import estimate_tokens
+from ..payment.models import Model
+
+logger = get_logger(__name__)
+
+
+def _parse_request_body(request_body: bytes | None) -> dict[str, Any]:
+    if not request_body:
+        return {}
+    try:
+        parsed = json.loads(request_body)
+    except (ValueError, TypeError):
+        return {}
+    return parsed if isinstance(parsed, dict) else {}
+
+
+def _count_with_litellm(model: str, body: dict[str, Any]) -> int:
+    messages = body.get("messages")
+    if not isinstance(messages, list):
+        messages = []
+
+    system = body.get("system")
+    if isinstance(system, str) and system:
+        messages = [{"role": "system", "content": system}, *messages]
+    elif isinstance(system, list):
+        text = "".join(
+            block.get("text", "")
+            for block in system
+            if isinstance(block, dict) and block.get("type") == "text"
+        )
+        if text:
+            messages = [{"role": "system", "content": text}, *messages]
+
+    tools = body.get("tools") if isinstance(body.get("tools"), list) else None
+
+    return int(
+        litellm.token_counter(
+            model=model,
+            messages=messages,
+            tools=tools,
+        )
+    )
+
+
+def count_tokens_locally(
+    request_body: bytes | None,
+    model_obj: Model | None,
+) -> Response:
+    """Return an Anthropic-compatible count_tokens response without
+    touching the upstream. Always returns 200; never raises."""
+    body = _parse_request_body(request_body)
+
+    model_name = ""
+    if model_obj is not None:
+        model_name = model_obj.forwarded_model_id or model_obj.id or ""
+    if not model_name:
+        body_model = body.get("model")
+        if isinstance(body_model, str):
+            model_name = body_model
+
+    input_tokens: int
+    try:
+        input_tokens = _count_with_litellm(model_name, body)
+    except Exception as exc:
+        messages = body.get("messages")
+        fallback_messages = messages if isinstance(messages, list) else []
+        input_tokens = estimate_tokens(fallback_messages)
+        logger.debug(
+            "litellm token_counter failed; using local estimator",
+            extra={
+                "model": model_name,
+                "error": str(exc),
+                "error_type": type(exc).__name__,
+                "estimated_tokens": input_tokens,
+            },
+        )
+
+    payload = {"input_tokens": max(0, int(input_tokens))}
+    return Response(
+        content=json.dumps(payload).encode(),
+        status_code=200,
+        media_type="application/json",
+    )

routstr/wallet.py

@@ -56,17 +56,56 @@ 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 settings.primary_mint, unit)
-    proofs = get_proofs_per_mint_and_unit(
-        wallet, mint_url or settings.primary_mint, unit
+    effective_mint_url = mint_url or settings.primary_mint
+    wallet: Wallet = await get_wallet(effective_mint_url, unit)
+    proofs = get_proofs_per_mint_and_unit(wallet, effective_mint_url, unit)
+    proofs_for_mint = sum(p.amount for p in proofs)
+
+    # Fallback: proofs from untrusted source mints are swapped to primary_mint
+    # during receive, so the user's preferred refund_mint_url may have no proofs
+    # even though the global wallet has the balance.
+    if proofs_for_mint < amount and effective_mint_url != settings.primary_mint:
+        logger.info(
+            f"send: insufficient proofs at {effective_mint_url} "
+            f"(have {proofs_for_mint}, need {amount}), falling back to primary_mint={settings.primary_mint}"
+        )
+        effective_mint_url = settings.primary_mint
+        wallet = await get_wallet(effective_mint_url, unit)
+        proofs = get_proofs_per_mint_and_unit(wallet, effective_mint_url, unit)
+        proofs_for_mint = sum(p.amount for p in proofs)
+
+    all_mint_urls = list({k.mint_url for k in wallet.keysets.values()})
+    proof_summary = {
+        f"{k.mint_url}/{k.unit.name}": sum(p.amount for p in wallet.proofs if p.id == k.id)
+        for k in wallet.keysets.values()
+    }
+    # Show ALL proofs in DB by keyset_id, regardless of whether the loaded wallet
+    # knows about that keyset. This reveals proofs orphaned under stale keysets.
+    raw_proofs_by_keyset: dict[str, int] = {}
+    for p in wallet.proofs:
+        raw_proofs_by_keyset[p.id] = raw_proofs_by_keyset.get(p.id, 0) + p.amount
+    logger.info(
+        f"send: proof inventory | mint={effective_mint_url} unit={unit} amount={amount} "
+        f"primary_mint={settings.primary_mint} proofs_for_mint={proofs_for_mint} "
+        f"all_mints={all_mint_urls} by_keyset={proof_summary} "
+        f"raw_proofs_by_keyset_id={raw_proofs_by_keyset} "
+        f"total_wallet_proofs={sum(p.amount for p in wallet.proofs)}"
     )
 
+    # Reserve proofs only after serialization succeeds — if serialize_proofs or
+    # swap_to_send fails mid-way, proofs stay unreserved so dashboard balance
+    # doesn't go negative.
     send_proofs, _ = await wallet.select_to_send(
-        proofs, amount, set_reserved=True, include_fees=False
-    )
-    token = await wallet.serialize_proofs(
-        send_proofs, include_dleq=False, legacy=False, memo=None
+        proofs, amount, set_reserved=False, include_fees=False
     )
+    try:
+        token = await wallet.serialize_proofs(
+            send_proofs, include_dleq=False, legacy=False, memo=None
+        )
+    except Exception:
+        await wallet.set_reserved_for_send(send_proofs, reserved=False)
+        raise
+    await wallet.set_reserved_for_send(send_proofs, reserved=True)
     return amount, token
 
 
@@ -81,10 +120,11 @@ async def _calculate_swap_amount(
     token_mint_url: str,
     token_wallet: Wallet,
     primary_wallet: Wallet,
+    proofs: list,
 ) -> int:
     """
     Calculate the amount to mint on the primary mint after accounting for
-    potential swap fees (melt fees) on the foreign mint.
+    melt fees and NUT-02 input fees on the foreign mint.
     """
     if settings.primary_mint_unit == "sat":
         receive_amount = amount_msat // 1000
@@ -111,10 +151,11 @@ async def _calculate_swap_amount(
         dummy_melt_quote = await token_wallet.melt_quote(dummy_mint_quote.request)
 
         fee_reserve = dummy_melt_quote.fee_reserve
+        input_fees = token_wallet.get_fees_for_proofs(proofs)
         if token_unit == "sat":
-            fee_msat = fee_reserve * 1000
+            fee_msat = (fee_reserve + input_fees) * 1000
         else:
-            fee_msat = fee_reserve
+            fee_msat = fee_reserve + input_fees
 
         amount_msat_after_fee = amount_msat - fee_msat
 
@@ -124,13 +165,14 @@ async def _calculate_swap_amount(
             minted_amount = int(amount_msat_after_fee)
 
         if minted_amount <= 0:
-            raise ValueError(f"Fees ({fee_reserve} {token_unit}) exceed token amount")
+            raise ValueError(f"Fees ({fee_reserve + input_fees} {token_unit}) exceed token amount")
 
         logger.info(
             "swap_to_primary_mint: fee estimation result",
             extra={
                 "token_amount_sat": amount_msat // 1000,
                 "estimated_fee_sat": fee_msat // 1000,
+                "input_fees": input_fees,
                 "minted_amount": minted_amount,
                 "minted_unit": settings.primary_mint_unit,
             },
@@ -191,6 +233,7 @@ async def swap_to_primary_mint(
         token_obj.mint,
         token_wallet,
         primary_wallet,
+        token_obj.proofs,
     )
 
     mint_quote = await primary_wallet.request_mint(minted_amount)
@@ -200,13 +243,15 @@ async def swap_to_primary_mint(
     )
 
     melt_quote = await token_wallet.melt_quote(mint_quote.request)
-    total_needed = melt_quote.amount + melt_quote.fee_reserve
+    input_fees = token_wallet.get_fees_for_proofs(token_obj.proofs)
+    total_needed = melt_quote.amount + melt_quote.fee_reserve + input_fees
     logger.info(
         "swap_to_primary_mint: melt quote received",
         extra={
             "melt_quote_id": melt_quote.quote,
             "melt_amount": melt_quote.amount,
             "melt_fee_reserve": melt_quote.fee_reserve,
+            "input_fees": input_fees,
             "total_needed": total_needed,
             "token_amount": token_amount,
         },
@@ -219,6 +264,7 @@ async def swap_to_primary_mint(
                 "token_amount": token_amount,
                 "melt_amount": melt_quote.amount,
                 "melt_fee_reserve": melt_quote.fee_reserve,
+                "input_fees": input_fees,
                 "total_needed": total_needed,
                 "shortfall": total_needed - token_amount,
             },
@@ -226,7 +272,7 @@ async def swap_to_primary_mint(
         raise ValueError(
             f"Token amount ({token_amount} {token_obj.unit}) is insufficient to cover "
             f"melt fees. Needed: {total_needed} {token_obj.unit} "
-            f"(amount: {melt_quote.amount} + fee: {melt_quote.fee_reserve})"
+            f"(amount: {melt_quote.amount} + fee: {melt_quote.fee_reserve} + input_fees: {input_fees})"
         )
 
     try:
@@ -257,19 +303,66 @@ async def swap_to_primary_mint(
         extra={"minted_amount": minted_amount, "mint_quote_id": mint_quote.quote},
     )
 
+    await primary_wallet.load_proofs(reload=True)
+    pre_mint_balance = primary_wallet.available_balance.amount
     try:
         _ = await primary_wallet.mint(minted_amount, quote_id=mint_quote.quote)
     except Exception as e:
-        logger.error(
-            "swap_to_primary_mint: mint on primary failed after successful melt",
-            extra={
-                "error": str(e),
-                "error_type": type(e).__name__,
-                "minted_amount": minted_amount,
-                "mint_quote_id": mint_quote.quote,
-            },
-        )
-        raise
+        if "11003" in str(e) or "outputs already signed" in str(e).lower():
+            # Previous mint call signed outputs at the mint but failed before
+            # bump_secret_derivation ran locally. Recover orphaned proofs and
+            # advance the counter so the next request derives fresh secrets.
+            logger.warning(
+                "swap_to_primary_mint: outputs already signed — recovering orphaned proofs",
+                extra={"mint_quote_id": mint_quote.quote, "minted_amount": minted_amount},
+            )
+            try:
+                for keyset_id in primary_wallet.keysets:
+                    await primary_wallet.restore_tokens_for_keyset(keyset_id, to=1, batch=25)
+                await primary_wallet.load_proofs(reload=True)
+                post_recovery_balance = primary_wallet.available_balance.amount
+                balance_gained = post_recovery_balance - pre_mint_balance
+                logger.info(
+                    "swap_to_primary_mint: recovery scan completed",
+                    extra={
+                        "pre_mint_balance": pre_mint_balance,
+                        "post_recovery_balance": post_recovery_balance,
+                        "balance_gained": balance_gained,
+                        "expected": minted_amount,
+                    },
+                )
+                if balance_gained < minted_amount:
+                    # Recovery scan ran but did NOT restore the orphaned proofs
+                    # (mint reports them as spent — they're stuck). Refuse to
+                    # credit the API key balance for proofs we don't actually hold.
+                    raise ValueError(
+                        f"Swap recovery failed: mint signed outputs but proofs are "
+                        f"unrecoverable (mint reports them spent). "
+                        f"Expected {minted_amount}, recovered {balance_gained}. "
+                        f"Local wallet DB ('.wallet/') state is corrupted — "
+                        f"the counter for keyset is stuck at a bad index range."
+                    )
+            except ValueError:
+                raise
+            except Exception as recovery_err:
+                logger.error(
+                    "swap_to_primary_mint: recovery failed",
+                    extra={"error": str(recovery_err)},
+                )
+                raise ValueError(
+                    f"Mint on primary failed and recovery unsuccessful: {e}"
+                ) from e
+        else:
+            logger.error(
+                "swap_to_primary_mint: mint on primary failed after successful melt",
+                extra={
+                    "error": str(e),
+                    "error_type": type(e).__name__,
+                    "minted_amount": minted_amount,
+                    "mint_quote_id": mint_quote.quote,
+                },
+            )
+            raise
 
     logger.info(
         "swap_to_primary_mint: completed successfully",
@@ -443,7 +536,7 @@ async def fetch_all_balances(
                 "unit": unit,
                 "wallet_balance": proofs_balance,
                 "user_balance": user_balance,
-                "owner_balance": proofs_balance - user_balance,
+                "owner_balance": proofs_balance - user_balance if proofs_balance != 0 else 0,
             }
             return result
         except Exception as e:
@@ -491,7 +584,9 @@ async def fetch_all_balances(
             total_wallet_balance_sats += proofs_balance_sats
             total_user_balance_sats += user_balance_sats
 
-    owner_balance = total_wallet_balance_sats - total_user_balance_sats
+    owner_balance = 0
+    if total_wallet_balance_sats != 0:
+        owner_balance = total_wallet_balance_sats - total_user_balance_sats
 
     return (
         balance_details,

tests/unit/test_count_tokens_local.py

@@ -0,0 +1,177 @@
+"""Unit tests for the local count_tokens shim.
+
+The shim runs whenever an upstream that does not support Anthropic's
+``/v1/messages`` endpoint is asked for a token count. It must always
+return a 200 JSON ``{"input_tokens": N}`` response and must never raise.
+"""
+
+from __future__ import annotations
+
+import json
+from typing import Any
+from unittest.mock import patch
+
+from routstr.payment.models import Architecture, Model, Pricing
+from routstr.upstream import count_tokens as count_tokens_module
+from routstr.upstream.count_tokens import count_tokens_locally
+
+
+def _make_model(model_id: str = "anthropic/claude-3-5-sonnet") -> Model:
+    pricing = Pricing(prompt=0.000003, completion=0.000015)
+    architecture = Architecture(
+        modality="text",
+        input_modalities=["text"],
+        output_modalities=["text"],
+        tokenizer="cl100k_base",
+        instruct_type=None,
+    )
+    return Model(
+        id=model_id,
+        name=model_id,
+        created=0,
+        description="",
+        context_length=200_000,
+        architecture=architecture,
+        pricing=pricing,
+    )
+
+
+def _body(payload: dict[str, Any]) -> bytes:
+    return json.dumps(payload).encode()
+
+
+def _read_payload(response: Any) -> dict[str, Any]:
+    body = response.body if isinstance(response.body, bytes) else bytes(response.body)
+    return json.loads(body.decode())
+
+
+def test_returns_input_tokens_for_simple_messages() -> None:
+    model = _make_model()
+    request_body = _body(
+        {
+            "model": model.id,
+            "messages": [{"role": "user", "content": "hello world"}],
+        }
+    )
+
+    response = count_tokens_locally(request_body, model)
+
+    assert response.status_code == 200
+    assert response.media_type == "application/json"
+    payload = _read_payload(response)
+    assert "input_tokens" in payload
+    assert isinstance(payload["input_tokens"], int)
+    assert payload["input_tokens"] >= 0
+
+
+def test_falls_back_to_estimator_when_litellm_raises() -> None:
+    model = _make_model()
+    request_body = _body(
+        {
+            "model": model.id,
+            "messages": [{"role": "user", "content": "this is a longer message"}],
+        }
+    )
+
+    with patch.object(
+        count_tokens_module,
+        "_count_with_litellm",
+        side_effect=RuntimeError("boom"),
+    ):
+        response = count_tokens_locally(request_body, model)
+
+    assert response.status_code == 200
+    payload = _read_payload(response)
+    assert payload["input_tokens"] >= 1
+
+
+def test_handles_missing_model_object() -> None:
+    request_body = _body(
+        {
+            "model": "anthropic/claude-3-5-sonnet",
+            "messages": [{"role": "user", "content": "hi"}],
+        }
+    )
+
+    response = count_tokens_locally(request_body, None)
+
+    assert response.status_code == 200
+    payload = _read_payload(response)
+    assert payload["input_tokens"] >= 0
+
+
+def test_handles_empty_request_body() -> None:
+    response = count_tokens_locally(b"", _make_model())
+
+    assert response.status_code == 200
+    payload = _read_payload(response)
+    assert payload["input_tokens"] >= 0
+
+
+def test_handles_malformed_json() -> None:
+    response = count_tokens_locally(b"not-json", _make_model())
+
+    assert response.status_code == 200
+    payload = _read_payload(response)
+    assert payload["input_tokens"] >= 0
+
+
+def test_includes_system_prompt_in_count() -> None:
+    model = _make_model()
+    short = _body(
+        {
+            "model": model.id,
+            "messages": [{"role": "user", "content": "hi"}],
+        }
+    )
+    with_system = _body(
+        {
+            "model": model.id,
+            "system": "You are a helpful assistant with a long preamble " * 10,
+            "messages": [{"role": "user", "content": "hi"}],
+        }
+    )
+
+    short_count = _read_payload(count_tokens_locally(short, model))["input_tokens"]
+    long_count = _read_payload(count_tokens_locally(with_system, model))["input_tokens"]
+
+    assert long_count > short_count
+
+
+def test_supports_anthropic_system_block_list() -> None:
+    model = _make_model()
+    request_body = _body(
+        {
+            "model": model.id,
+            "system": [{"type": "text", "text": "be terse" * 50}],
+            "messages": [{"role": "user", "content": "ok"}],
+        }
+    )
+
+    response = count_tokens_locally(request_body, model)
+
+    payload = _read_payload(response)
+    assert payload["input_tokens"] > 0
+
+
+def test_uses_forwarded_model_id_when_present() -> None:
+    model = _make_model("anthropic/claude-3-5-sonnet")
+    model.forwarded_model_id = "claude-3-5-sonnet-20241022"
+    request_body = _body(
+        {
+            "model": "ignored",
+            "messages": [{"role": "user", "content": "hi"}],
+        }
+    )
+
+    captured: dict[str, Any] = {}
+
+    def _capture(model_name: str, body: dict[str, Any]) -> int:
+        captured["model"] = model_name
+        return 7
+
+    with patch.object(count_tokens_module, "_count_with_litellm", side_effect=_capture):
+        response = count_tokens_locally(request_body, model)
+
+    assert captured["model"] == "claude-3-5-sonnet-20241022"
+    assert _read_payload(response)["input_tokens"] == 7

tests/unit/test_messages_litellm_dispatch.py

@@ -908,7 +908,7 @@ async def test_forward_request_skips_litellm_when_provider_supports_messages() -
 
 
 @pytest.mark.asyncio
-async def test_forward_request_skips_litellm_for_count_tokens() -> None:
+async def test_forward_request_handles_count_tokens_locally() -> None:
     provider = _make_provider()
     key = _make_key()
     model = _make_model()
@@ -923,19 +923,25 @@ async def test_forward_request_skips_litellm_for_count_tokens() -> None:
         with patch.object(
             provider,
             "prepare_request_body",
-            side_effect=RuntimeError("stop here"),
+            side_effect=AssertionError("upstream should not be called"),
         ):
-            with pytest.raises(RuntimeError, match="stop here"):
-                await provider.forward_request(
-                    request=request,
-                    path="messages/count_tokens",
-                    headers={},
-                    request_body=_anthropic_request_body(),
-                    key=key,
-                    max_cost_for_model=10_000,
-                    session=session,
-                    model_obj=model,
-                )
+            response = await provider.forward_request(
+                request=request,
+                path="messages/count_tokens",
+                headers={},
+                request_body=_anthropic_request_body(),
+                key=key,
+                max_cost_for_model=10_000,
+                session=session,
+                model_obj=model,
+            )
+
+    assert response.status_code == 200
+    body = response.body if isinstance(response.body, bytes) else bytes(response.body)
+    payload = json.loads(body.decode())
+    assert "input_tokens" in payload
+    assert isinstance(payload["input_tokens"], int)
+    assert payload["input_tokens"] >= 0
 
 
 # ---------------------------------------------------------------------------
@@ -1010,7 +1016,7 @@ async def test_forward_x_cashu_request_skips_litellm_when_native_messages() -> N
 
 
 @pytest.mark.asyncio
-async def test_forward_x_cashu_request_skips_litellm_for_count_tokens() -> None:
+async def test_forward_x_cashu_request_handles_count_tokens_locally() -> None:
     provider = _make_provider()
     model = _make_model()
     request = _make_request()
@@ -1024,18 +1030,24 @@ async def test_forward_x_cashu_request_skips_litellm_for_count_tokens() -> None:
         with patch.object(
             provider,
             "prepare_request_body",
-            side_effect=RuntimeError("stop here"),
+            side_effect=AssertionError("upstream should not be called"),
         ):
-            with pytest.raises(RuntimeError, match="stop here"):
-                await provider.forward_x_cashu_request(
-                    request=request,
-                    path="v1/messages/count_tokens",
-                    headers={},
-                    amount=5_000,
-                    unit="sat",
-                    max_cost_for_model=10_000,
-                    model_obj=model,
-                )
+            response = await provider.forward_x_cashu_request(
+                request=request,
+                path="v1/messages/count_tokens",
+                headers={},
+                amount=5_000,
+                unit="sat",
+                max_cost_for_model=10_000,
+                model_obj=model,
+                mint="https://mint",
+                payment_token_hash="h",
+            )
+
+    assert response.status_code == 200
+    body = response.body if isinstance(response.body, bytes) else bytes(response.body)
+    payload = json.loads(body.decode())
+    assert "input_tokens" in payload
 
 
 # ---------------------------------------------------------------------------

tests/unit/test_wallet.py

@@ -123,6 +123,7 @@ async def test_swap_to_primary_mint_insufficient_for_fees() -> None:
     mock_token_wallet = Mock()
     mock_token_wallet.load_mint = AsyncMock()
     mock_token_wallet.load_proofs = AsyncMock()
+    mock_token_wallet.get_fees_for_proofs = Mock(return_value=0)
 
     mock_primary_wallet = Mock()
     mock_primary_wallet.load_mint = AsyncMock()
@@ -166,6 +167,7 @@ async def test_swap_to_primary_mint_melt_error_wrapped() -> None:
     mock_token_wallet = Mock()
     mock_token_wallet.load_mint = AsyncMock()
     mock_token_wallet.load_proofs = AsyncMock()
+    mock_token_wallet.get_fees_for_proofs = Mock(return_value=0)
 
     mock_primary_wallet = Mock()
     mock_primary_wallet.load_mint = AsyncMock()
@@ -261,6 +263,7 @@ async def test_swap_to_primary_mint_success() -> None:
     mock_token_wallet = Mock()
     mock_token_wallet.load_mint = AsyncMock()
     mock_token_wallet.load_proofs = AsyncMock()
+    mock_token_wallet.get_fees_for_proofs = Mock(return_value=0)
 
     mock_primary_wallet = Mock()
     mock_primary_wallet.load_mint = AsyncMock()