Skip to content

Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.

fix(proxy): route Foundry Anthropic messages (#1878)

Vinay Gupta739f654bbd71
diff --git a/headroom/providers/proxy_routes.py b/headroom/providers/proxy_routes.py
index d215edf8c..ce6b5fe08 100644
--- a/headroom/providers/proxy_routes.py
+++ b/headroom/providers/proxy_routes.py
@@ -474,6 +474,13 @@ async def _handle_chatgpt_codex_images(
def register_provider_routes(app: FastAPI, proxy: Any) -> None:
"""Register provider-specific proxy endpoints."""
+ def normalize_request_path(request: Request, path: str) -> None:
+ request.scope["path"] = path
+ if "raw_path" in request.scope:
+ request.scope["raw_path"] = quote(path).encode("ascii")
+ if hasattr(request, "_url"):
+ delattr(request, "_url")
+
async def vertex_publisher_passthrough(request: Request, publisher: str, action: str):
return await proxy.handle_passthrough(
request,
@@ -486,6 +493,11 @@ def register_provider_routes(app: FastAPI, proxy: Any) -> None:
async def anthropic_messages(request: Request):
return await proxy.handle_anthropic_messages(request)
+ @app.post("/anthropic/v1/messages")
+ async def foundry_anthropic_messages(request: Request):
+ normalize_request_path(request, "/v1/messages")
+ return await proxy.handle_anthropic_messages(request, _api_target(proxy, "anthropic"))
+
# AWS Bedrock InvokeModel passthrough. Registered ONLY when an upstream is
# configured (`--bedrock-api-url` / BEDROCK_TARGET_API_URL): without it,
# `/model/{id}/invoke` keeps falling through to the catch-all (verbatim,
diff --git a/tests/test_provider_proxy_routes.py b/tests/test_provider_proxy_routes.py
index c80c03701..be588a3c1 100644
--- a/tests/test_provider_proxy_routes.py
+++ b/tests/test_provider_proxy_routes.py
@@ -173,6 +173,14 @@ def test_provider_passthrough_routes_forward_expected_targets(monkeypatch) -> No
"model": "claude-3-5-sonnet@20240620",
"force_stream": False,
}
+ assert client.post("/anthropic/v1/messages?beta=true").json() == {
+ "handler": "handle_anthropic_messages",
+ "path": "/v1/messages",
+ "upstream_base_url": "https://api.anthropic.test",
+ "provider": "anthropic",
+ "model": None,
+ "force_stream": False,
+ }
non_anthropic_raw = client.post(
"/projects/p/locations/us-central1/publishers/google/models/gemini-2.0-flash:rawPredict"
).json()
@@ -336,6 +344,11 @@ def test_provider_specific_routes_delegate_to_expected_proxy_handlers(monkeypatc
with TestClient(_app()) as client:
assert client.post("/v1/messages").json()["handler"] == "handle_anthropic_messages"
+ assert client.post("/anthropic/v1/messages").json() == {
+ "handler": "handle_anthropic_messages",
+ "path": "/v1/messages",
+ "args": ["https://api.anthropic.test"],
+ }
assert (
client.post("/v1/messages/batches").json()["handler"] == "handle_anthropic_batch_create"
)