eggpool 0.4.3__tar.gz → 0.4.4__tar.gz

{eggpool-0.4.3 → eggpool-0.4.4}/CHANGELOG.md

@@ -9,6 +9,14 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 Nothing yet.
 
+## [0.4.4] - 2026-06-30
+
+### Fixed
+
+- **Provider-scoped transcodable protocol lookup**. `get_transcodable_protocols` and `count_eligible_accounts_for_protocol` on `ModelCatalogCache` now accept an optional `provider_id` filter so protocol inference and eligibility counting only consider accounts that belong to a specific provider. `_infer_upstream_protocol` in `proxy_request.py` and `RequestCoordinator._resolve_upstream_protocol` pass the selected account's `provider_id`, preventing cross-provider protocol leakage when the same model is served by providers with different protocol surfaces.
+- **MiniMax token-plan contract warning**. `check-config` now emits a warning when a MiniMax provider entry targets `api.minimax.io` with `openai` protocol but does not use the `/anthropic` path — token-plan keys hitting the OpenAI `/v1/chat/completions` surface can return upstream `insufficient balance (1008)` errors.
+- **`build_upstream_headers` now receives the upstream protocol**. The coordinator passes `context.upstream_protocol` to `build_upstream_headers` so protocol-specific auth header construction (e.g. `x-api-key` for Anthropic vs `Authorization: Bearer` for OpenAI) uses the correct upstream protocol instead of the client's protocol.
+
 ## [0.4.3] - 2026-06-30
 
 ### Added

{eggpool-0.4.3 → eggpool-0.4.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: eggpool
-Version: 0.4.3
+Version: 0.4.4
 Summary: A lightweight proxy that aggregates multiple LLM provider accounts behind one OpenAI-compatible endpoint
 Project-URL: Homepage, https://github.com/eggstack/eggpool
 Project-URL: Repository, https://github.com/eggstack/eggpool

{eggpool-0.4.3 → eggpool-0.4.4}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "eggpool"
-version = "0.4.3"
+version = "0.4.4"
 description = "A lightweight proxy that aggregates multiple LLM provider accounts behind one OpenAI-compatible endpoint"
 readme = "README.md"
 requires-python = ">=3.11"

{eggpool-0.4.3 → eggpool-0.4.4}/src/eggpool/api/proxy_request.py

@@ -147,6 +147,7 @@ def _infer_upstream_protocol(
     candidates = catalog.cache.get_transcodable_protocols(
         model_id,
         client_protocol=client_protocol,
+        provider_id=provider_id,
     )
     if not candidates:
         return None
@@ -155,6 +156,7 @@ def _infer_upstream_protocol(
         p: catalog.cache.count_eligible_accounts_for_protocol(
             model_id,
             p,
+            provider_id=provider_id,
         )
         for p in candidates
     }

{eggpool-0.4.3 → eggpool-0.4.4}/src/eggpool/app.py

@@ -67,6 +67,7 @@ from eggpool.providers.client_pool import ProviderClientPool
 from eggpool.providers.dns_cache import DnsNetworkBackend
 from eggpool.providers.outbound import OutboundClientManager, default_network_backend
 from eggpool.request.coordinator import RequestCoordinator
+from eggpool.routing.config import routing_stale_after_s
 from eggpool.routing.router import Router
 from eggpool.runtime_dispatch import DispatchOverheadRecorder
 from eggpool.stats import StatsService
@@ -784,7 +785,7 @@ async def _lifespan_runtime(app: FastAPI) -> AsyncGenerator[None]:
         registry,
         catalog,
         health_manager=health_manager,
-        stale_after_s=float(config.models.stale_after_s),
+        stale_after_s=routing_stale_after_s(config),
         local_quota_mode=config.routing.local_quota_mode,
         fairness_mode=config.routing.fairness_mode,
         fairness_epsilon=config.routing.fairness_epsilon,

{eggpool-0.4.3 → eggpool-0.4.4}/src/eggpool/catalog/cache.py

@@ -789,6 +789,7 @@ class ModelCatalogCache:
         model_id: str,
         *,
         client_protocol: str,
+        provider_id: str | None = None,
     ) -> set[str]:
         """Return the set of protocols a model can reach, minus the client protocol.
 
@@ -798,11 +799,13 @@ class ModelCatalogCache:
         supporting = self.get_supporting_accounts(model_id)
         protocols: set[str] = set()
         for account_name in supporting:
-            provider_id = self._account_providers.get(account_name)
-            if provider_id is None:
+            account_provider_id = self._account_providers.get(account_name)
+            if account_provider_id is None:
+                continue
+            if provider_id is not None and account_provider_id != provider_id:
                 continue
             if self._config is not None:
-                provider = self._config.providers.get(provider_id)
+                provider = self._config.providers.get(account_provider_id)
                 if provider is not None:
                     for proto in provider.protocols:
                         protocols.add(proto)
@@ -813,17 +816,21 @@ class ModelCatalogCache:
         self,
         model_id: str,
         protocol: str,
+        *,
+        provider_id: str | None = None,
     ) -> int:
         """Count enabled accounts whose provider supports *protocol* and
         has the model in its catalogue."""
         supporting = self.get_supporting_accounts(model_id)
         n = 0
         for account_name in supporting:
-            provider_id = self._account_providers.get(account_name)
-            if provider_id is None:
+            account_provider_id = self._account_providers.get(account_name)
+            if account_provider_id is None:
+                continue
+            if provider_id is not None and account_provider_id != provider_id:
                 continue
             if self._config is not None:
-                provider = self._config.providers.get(provider_id)
+                provider = self._config.providers.get(account_provider_id)
                 if provider is not None and protocol in provider.protocols:
                     n += 1
         return n

{eggpool-0.4.3 → eggpool-0.4.4}/src/eggpool/cli_full.py

@@ -11,6 +11,7 @@ import zipfile
 from contextlib import suppress
 from pathlib import Path
 from typing import TYPE_CHECKING, Any, cast
+from urllib.parse import urlsplit
 
 if TYPE_CHECKING:
     from collections.abc import Callable, Coroutine
@@ -599,6 +600,22 @@ def _check_stale_contracts(config: AppConfig, config_path: str) -> list[str]:
                 "in protocols; the field will be ignored"
             )
 
+        parsed_base = urlsplit(provider.base_url)
+        if (
+            parsed_base.hostname == "api.minimax.io"
+            and "openai" in provider.protocols
+            and parsed_base.path.rstrip("/") != "/anthropic"
+        ):
+            warnings.append(
+                f"[{provider.id}] api.minimax.io token-plan keys should use "
+                "the Anthropic-compatible MiniMax contract "
+                "(base_url='https://api.minimax.io/anthropic', "
+                "protocols=['anthropic'], anthropic_path='/v1/messages', "
+                "auth.header='x-api-key'); the OpenAI "
+                "/v1/chat/completions surface can return upstream "
+                "'insufficient balance (1008)' for token-plan keys"
+            )
+
         if endpoint is not None:
             try:
                 compose_provider_url(provider, endpoint.path)
@@ -3296,6 +3313,7 @@ def accounts_explain(
     from eggpool.catalog.cache import ModelCatalogCache
     from eggpool.models.config import AppConfig
     from eggpool.quota.estimation import QuotaEstimator
+    from eggpool.routing.config import routing_stale_after_s
     from eggpool.routing.router import Router
 
     config_path: str = ctx.obj["config_path"]
@@ -3342,6 +3360,11 @@ def accounts_explain(
             registry,
             catalog,  # type: ignore[reportArgumentType]
             quota_estimator=quota_estimator,
+            stale_after_s=routing_stale_after_s(config),
+            local_quota_mode=config.routing.local_quota_mode,
+            fairness_mode=config.routing.fairness_mode,
+            fairness_epsilon=config.routing.fairness_epsilon,
+            fairness_scope=config.routing.fairness_scope,
         )
         rows = await router.explain_account_eligibility(
             model_id=model_id,

{eggpool-0.4.3 → eggpool-0.4.4}/src/eggpool/request/coordinator.py

@@ -1858,7 +1858,11 @@ class RequestCoordinator:
             else None
         )
         if provider_cfg is not None:
-            auth_headers = build_upstream_headers(provider_cfg, selected.api_key)
+            auth_headers = build_upstream_headers(
+                provider_cfg,
+                selected.api_key,
+                protocol=context.upstream_protocol,
+            )
             sanitized.update(auth_headers)
             if logger.isEnabledFor(logging.DEBUG):
                 auth_shape = build_auth_headers(provider_cfg, selected.api_key)
@@ -2573,6 +2577,7 @@ class RequestCoordinator:
         candidates = self._catalog.cache.get_transcodable_protocols(
             context.model_id,
             client_protocol=context.protocol,
+            provider_id=context.provider_id,
         )
         if not candidates:
             return None
@@ -2582,6 +2587,7 @@ class RequestCoordinator:
             p: self._catalog.cache.count_eligible_accounts_for_protocol(
                 context.model_id,
                 p,
+                provider_id=context.provider_id,
             )
             for p in candidates
         }

eggpool-0.4.4/src/eggpool/routing/config.py

@@ -0,0 +1,21 @@
+"""Routing configuration helpers."""
+
+from __future__ import annotations
+
+from typing import TYPE_CHECKING
+
+if TYPE_CHECKING:
+    from eggpool.models.config import AppConfig
+
+
+def routing_stale_after_s(config: AppConfig) -> float | None:
+    """Return the per-account catalog stale gate for routing.
+
+    ``allow_stale_catalog`` means EggPool may continue serving from
+    previously known catalog support when refreshes are uncertain. In
+    that mode, routing must not silently de-pool healthy sibling
+    accounts just because their per-account refresh timestamp aged out.
+    """
+    if config.models.allow_stale_catalog:
+        return None
+    return float(config.models.stale_after_s)

{eggpool-0.4.3 → eggpool-0.4.4}/uv.lock

@@ -196,7 +196,7 @@ wheels = [
 
 [[package]]
 name = "eggpool"
-version = "0.4.3"
+version = "0.4.4"
 source = { editable = "." }
 dependencies = [
     { name = "aiosqlite" },