Merge pull request 'adapter: support OpenAI streaming (SSE) in /v1/chat/completions' (#1) from feat/adapter-streaming into main

OpenCode's @ai-sdk/openai-compatible sends stream:true and would render an
empty response because the adapter always returned a single non-streaming
chat.completion JSON body. Now when stream:true, emit OpenAI-compatible SSE
chat.completion.chunk events (role, content, [DONE]) so streaming clients
render text. Non-streaming path unchanged.

Adds a local EchoServer test that exercises the streaming route end-to-end.

lib/n8n_openai_adapter/router.ex

@@ -235,6 +235,7 @@ defmodule N8nOpenaiAdapter.Router do
 
     model = Map.get(data, "model")
     messages = Map.get(data, "messages", [])
+    stream? = Map.get(data, "stream", false)
     thread_id = Map.get(data, "thread_id")
     session_id = if is_binary(thread_id) and thread_id != "", do: thread_id, else: "default"
 
@@ -258,32 +259,106 @@ defmodule N8nOpenaiAdapter.Router do
               webhook ->
                 case call_n8n(webhook, session_id, chat_input) do
                   {:ok, output} ->
-                    resp = %{
-                      id:
-                        "chatcmpl-#{Base.encode16(:crypto.strong_rand_bytes(12), case: :lower)}",
-                      object: "chat.completion",
-                      created: System.system_time(:second),
-                      model: model,
-                      choices: [
-                        %{
-                          index: 0,
-                          message: %{role: "assistant", content: output},
-                          finish_reason: "stop"
-                        }
-                      ],
-                      usage: %{prompt_tokens: 0, completion_tokens: 0, total_tokens: 0}
-                    }
-
-                    json(conn, 200, resp)
+                    if stream? do
+                      stream_completion(conn, model, output)
+                    else
+                      json(conn, 200, completion_body(model, output))
+                    end
 
                   {:error, msg} ->
-                    error(conn, 502, msg)
+                    if stream? do
+                      stream_error(conn, msg)
+                    else
+                      error(conn, 502, msg)
+                    end
                 end
             end
         end
     end
   end
 
+  # Build a single, non-streaming chat.completion body.
+  defp completion_body(model, output) do
+    %{
+      id: "chatcmpl-#{Base.encode16(:crypto.strong_rand_bytes(12), case: :lower)}",
+      object: "chat.completion",
+      created: System.system_time(:second),
+      model: model,
+      choices: [
+        %{
+          index: 0,
+          message: %{role: "assistant", content: output},
+          finish_reason: "stop"
+        }
+      ],
+      usage: %{prompt_tokens: 0, completion_tokens: 0, total_tokens: 0}
+    }
+  end
+
+  # Emit an OpenAI-compatible SSE stream. The n8n reply arrives whole, so we
+  # deliver it as a single delta chunk (role chunk first, then content) and
+  # terminate with `data: [DONE]`. OpenAI SDKs that request `stream: true`
+  # consume this shape; without it they render an empty response.
+  defp stream_completion(conn, model, output) do
+    id = "chatcmpl-#{Base.encode16(:crypto.strong_rand_bytes(12), case: :lower)}"
+    created = System.system_time(:second)
+
+    conn =
+      conn
+      |> put_resp_content_type("text/event-stream")
+      |> put_resp_header("cache-control", "no-cache")
+      |> send_chunked(200)
+
+    chunks = [
+      %{
+        id: id,
+        object: "chat.completion.chunk",
+        created: created,
+        model: model,
+        choices: [%{index: 0, delta: %{role: "assistant"}, finish_reason: nil}]
+      },
+      %{
+        id: id,
+        object: "chat.completion.chunk",
+        created: created,
+        model: model,
+        choices: [%{index: 0, delta: %{content: output}, finish_reason: nil}]
+      },
+      %{
+        id: id,
+        object: "chat.completion.chunk",
+        created: created,
+        model: model,
+        choices: [%{index: 0, delta: %{}, finish_reason: "stop"}]
+      }
+    ]
+
+    {:ok, conn} =
+      Enum.reduce_while(chunks, {:ok, conn}, fn chunk, {:ok, conn} ->
+        case chunk(conn, "data: #{Jason.encode!(chunk)}\n\n") do
+          {:ok, conn} -> {:cont, {:ok, conn}}
+          {:error, _} -> {:halt, {:ok, conn}}
+        end
+      end)
+
+    {:ok, conn} = chunk(conn, "data: [DONE]\n\n")
+    conn
+  end
+
+  # Emit an SSE-formatted error so a streaming client sees the failure.
+  defp stream_error(conn, msg) do
+    conn =
+      conn
+      |> put_resp_content_type("text/event-stream")
+      |> put_resp_header("cache-control", "no-cache")
+      |> send_chunked(502)
+
+    payload = %{error: %{message: msg}}
+    {:ok, conn} = chunk(conn, "data: #{Jason.encode!(payload)}\n\n")
+    {:ok, conn} = chunk(conn, "data: [DONE]\n\n")
+    conn
+  end
+
   # Self-contained admin page: lists agents and lets you add/remove them via
   # the admin API. Uses the ADMIN_API_KEY (entered in the page) as the Bearer
   # token for the /admin/agents calls.

test/router_test.exs

@@ -6,11 +6,46 @@ defmodule N8nOpenaiAdapter.RouterTest do
   alias N8nOpenaiAdapter.AgentRegistry
   alias N8nOpenaiAdapter.Router
 
-  setup do
+  # A minimal local HTTP server that answers n8n-style chat webhooks with a
+  # fixed {output: ...} body, so we can exercise the adapter's n8n call path
+  # (including the SSE streaming route) without a real n8n instance.
+  defmodule EchoServer do
+    use Plug.Router
+
+    plug(:match)
+    plug(:dispatch)
+
+    post "/chat" do
+      conn
+      |> put_resp_content_type("application/json")
+      |> send_resp(200, Jason.encode!(%{"output" => "Echo reply"}))
+    end
+
+    match _ do
+      send_resp(conn, 404, "not found")
+    end
+  end
+
+  setup_all do
+    port = get_free_port()
+    {:ok, _} = Bandit.start_link(plug: EchoServer, port: port, scheme: :http)
+    {:ok, port: port}
+  end
+
+  defp get_free_port do
+    {:ok, socket} = :gen_tcp.listen(0, [:binary, packet: :raw, active: false, reuseaddr: true])
+    {:ok, port} = :inet.port(socket)
+    :gen_tcp.close(socket)
+    port
+  end
+
+  setup %{port: port} do
     # The app starts with an empty store (no AGENTS env seeding). Seed the two
     # test agents via the registry so the base tests have something to list.
     AgentRegistry.put("scholar-agent", "https://n8n.bueso.eu/webhook/scholar-id/chat")
     AgentRegistry.put("media-agent", "https://n8n.bueso.eu/webhook/media-id/chat")
+    # A third agent pointing at the local echo server, for the streaming path.
+    AgentRegistry.put("echo-agent", "http://127.0.0.1:#{port}/chat")
     :ok
   end
 
@@ -72,6 +107,57 @@ defmodule N8nOpenaiAdapter.RouterTest do
     assert conn.status == 400
   end
 
+  test "POST /v1/chat/completions with stream=true emits SSE JSON chunks + [DONE]" do
+    conn =
+      conn(
+        :post,
+        "/v1/chat/completions",
+        Jason.encode!(%{
+          "model" => "echo-agent",
+          "stream" => true,
+          "messages" => [%{"role" => "user", "content" => "hi"}]
+        })
+      )
+      |> put_req_header("authorization", "Bearer test-key")
+      |> put_req_header("content-type", "application/json")
+      |> Router.call(Router.init([]))
+
+    assert conn.status == 200, "expected 200, got #{conn.status}: #{conn.resp_body}"
+    body = conn.resp_body
+    assert {"content-type", ct} = List.keyfind(conn.resp_headers, "content-type", 0)
+    assert String.starts_with?(ct, "text/event-stream")
+    # Each SSE event is `data: <json>\n\n`; split and check the JSON is valid
+    events = String.split(body, "\n\n") |> Enum.reject(&(&1 == ""))
+
+    data_events = Enum.filter(events, &String.starts_with?(&1, "data: "))
+    assert Enum.any?(data_events, &String.ends_with?(&1, "[DONE]"))
+
+    # Parse a non-DONE chunk and require the OpenAI chunk shape
+    chunk_event = Enum.find(data_events, &(not String.ends_with?(&1, "[DONE]")))
+    chunk = Jason.decode!(String.replace_prefix(chunk_event, "data: ", ""))
+    assert chunk["object"] == "chat.completion.chunk"
+    assert is_list(chunk["choices"])
+    # The content delta chunk carries the reply text
+    content_chunks =
+      Enum.filter(data_events, fn ev ->
+        decoded =
+          case Jason.decode(String.replace_prefix(ev, "data: ", "")) do
+            {:ok, j} -> j
+            _ -> :error
+          end
+
+        case decoded do
+          %{"choices" => [choice]} ->
+            match?(%{"content" => c} when is_binary(c), Map.get(choice, "delta"))
+
+          _ ->
+            false
+        end
+      end)
+
+    assert length(content_chunks) == 1, "content_chunks=#{length(content_chunks)}"
+  end
+
   test "GET /admin serves the web admin page (no auth needed to load the form)" do
     conn = conn(:get, "/admin") |> Router.call(Router.init([]))