feat: add provider and model routing tables, admin UI, and audit category for gateway

Implement the multi-provider routing system for the OpenAI gateway, including two new database tables (`gateway_providers`, `gateway_models`) ensured at init, a new admin page at `/admin/gateway` with full CRUD for providers and model routes, and a `"gateway"` audit category mapped to `"ai"`. The routing layer sits transparently on top of the existing single-provider default path: unmatched model names fall through unchanged, while matched routes forward to the configured provider with their own pricing economy, vision model, and context window. Cross-worker cache invalidation uses a shared `_ROUTING_CACHE` bumped via `"gateway_routing"` cache version.
2026-06-16 22:11:14 +00:00
parent c100b4b692
commit 7bc67662fa
21 changed files with 1401 additions and 20 deletions
@@ -112,6 +112,47 @@ def test_model_is_forced(local_db, monkeypatch):
    assert rt._client.calls[-1][1]["model"] == "deepseek-chat"


+def test_model_route_overrides_upstream(local_db, monkeypatch):
+    from devplacepy.services.openai_gateway import routing
+
+    monkeypatch.setattr(gwmod.httpx, "AsyncClient", FakeClient_openai_gateway)
+    routing.provider_store.set(
+        routing.ProviderIn(
+            name="gwroute",
+            base_url="https://routed.example/v1/chat/completions",
+            api_key="routed-key",
+        )
+    )
+    routing.model_store.set(
+        routing.ModelRouteIn(
+            source_model="custom-pro",
+            provider="gwroute",
+            target_model="vendor/pro",
+            kind="chat",
+            price_output_per_m=9.0,
+        )
+    )
+    try:
+        svc = GatewayService()
+        cfg = svc.effective_config()
+        cfg["gateway_vision_enabled"] = False
+        rt = svc.runtime()
+        run_async(
+            rt.handle_chat(
+                {"model": "custom-pro", "messages": [{"role": "user", "content": "hi"}]},
+                cfg,
+                ("guest", "test"),
+                "test",
+            )
+        )
+        url, body = rt._client.calls[-1]
+        assert str(url) == "https://routed.example/v1/chat/completions"
+        assert body["model"] == "vendor/pro"
+    finally:
+        routing.model_store.remove("custom-pro")
+        routing.provider_store.remove("gwroute")
+
+
 def test_vision_rewrites_image_to_text(local_db, monkeypatch):
    monkeypatch.setattr(gwmod.httpx, "AsyncClient", FakeClient_openai_gateway)
    svc = GatewayService()
@@ -0,0 +1,142 @@
+# retoor <retoor@molodetz.nl>
+
+from devplacepy.services.openai_gateway import routing as r
+from devplacepy.services.openai_gateway.usage import pricing_from_cfg
+
+
+def _cleanup(providers, models):
+    for name in providers:
+        r.provider_store.remove(name)
+    for source in models:
+        r.model_store.remove(source)
+
+
+def test_no_routes_pass_through(local_db):
+    assert r.chat_overlay("ghost-model-xyz", {}) is None
+    assert r.embed_overlay("ghost-embed-xyz", {}) is None
+
+
+def test_chat_route_overlay_and_economy(local_db):
+    r.provider_store.set(
+        r.ProviderIn(
+            name="utor",
+            base_url="https://up.example/v1/chat/completions",
+            api_key="k1",
+        )
+    )
+    r.model_store.set(
+        r.ModelRouteIn(
+            source_model="ut-chat",
+            provider="utor",
+            target_model="vendor/big",
+            kind="chat",
+            context_window=64000,
+            price_cache_hit_per_m=1.0,
+            price_cache_miss_per_m=2.0,
+            price_output_per_m=8.0,
+        )
+    )
+    try:
+        overlay = r.chat_overlay("ut-chat", {"gateway_model_context_map": "{}"})
+        assert overlay["gateway_force_model"] is True
+        assert overlay["gateway_model"] == "vendor/big"
+        assert overlay["gateway_upstream_url"] == "https://up.example/v1/chat/completions"
+        assert overlay["gateway_api_key"] == "k1"
+        assert overlay["gateway_price_output_per_m"] == 8.0
+        assert overlay["gateway_model_context_map"]["vendor/big"] == 64000
+        pricing = pricing_from_cfg(overlay)
+        assert pricing.chat_output_per_m == 8.0
+        assert pricing.chat_cache_miss_per_m == 2.0
+    finally:
+        _cleanup(["utor"], ["ut-chat"])
+
+
+def test_vision_merge_overlay(local_db):
+    r.provider_store.set(
+        r.ProviderIn(
+            name="utvis",
+            base_url="https://vis.example/v1/chat/completions",
+            api_key="vk",
+        )
+    )
+    r.model_store.set(
+        r.ModelRouteIn(
+            source_model="ut-vision",
+            provider="utvis",
+            target_model="vendor/vlm",
+            kind="chat",
+            vision_model="vendor/describe",
+            price_input_per_m=0.5,
+        )
+    )
+    try:
+        overlay = r.chat_overlay("ut-vision", {})
+        assert overlay["gateway_vision_enabled"] is True
+        assert overlay["gateway_vision_model"] == "vendor/describe"
+        assert overlay["gateway_vision_url"] == "https://vis.example/v1/chat/completions"
+        assert overlay["gateway_vision_key"] == "vk"
+        assert overlay["gateway_vision_price_input_per_m"] == 0.5
+    finally:
+        _cleanup(["utvis"], ["ut-vision"])
+
+
+def test_embed_route_and_kind_isolation(local_db):
+    r.provider_store.set(
+        r.ProviderIn(
+            name="utemb",
+            base_url="https://emb.example/v1/chat/completions",
+            api_key="ek",
+        )
+    )
+    r.model_store.set(
+        r.ModelRouteIn(
+            source_model="ut-embed",
+            provider="utemb",
+            target_model="vendor/embed",
+            kind="embed",
+            price_input_per_m=0.13,
+        )
+    )
+    try:
+        overlay = r.embed_overlay("ut-embed", {})
+        assert overlay["gateway_embed_model"] == "vendor/embed"
+        assert overlay["gateway_embed_url"] == "https://emb.example/v1/embeddings"
+        assert overlay["gateway_embed_key"] == "ek"
+        assert overlay["gateway_embed_price_input_per_m"] == 0.13
+        assert r.chat_overlay("ut-embed", {}) is None
+    finally:
+        _cleanup(["utemb"], ["ut-embed"])
+
+
+def test_inactive_route_ignored(local_db):
+    r.model_store.set(
+        r.ModelRouteIn(
+            source_model="ut-off",
+            target_model="vendor/x",
+            kind="chat",
+            is_active=False,
+        )
+    )
+    try:
+        assert r.chat_overlay("ut-off", {}) is None
+    finally:
+        _cleanup([], ["ut-off"])
+
+
+def test_blank_provider_uses_default_upstream(local_db):
+    r.model_store.set(
+        r.ModelRouteIn(
+            source_model="ut-default",
+            target_model="vendor/def",
+            kind="chat",
+            price_output_per_m=3.0,
+        )
+    )
+    try:
+        overlay = r.chat_overlay("ut-default", {})
+        assert overlay["gateway_model"] == "vendor/def"
+        assert "gateway_upstream_url" not in overlay
+        assert "gateway_api_key" not in overlay
+        assert overlay["gateway_price_output_per_m"] == 3.0
+    finally:
+        _cleanup([], ["ut-default"])