diff --git a/docs-go/flows.md b/docs-go/flows.md
index 12d176acd..95a93fe33 100644
--- a/docs-go/flows.md
+++ b/docs-go/flows.md
@@ -156,16 +156,22 @@ first.
         // ...
       },
     )
-    genkit.StartFlowServer(":1234")
+  err := genkit.StartFlowServer(":1234", []string{})
+
+  // startProdServer always returns a non-nil error: the one returned by
+  // http.ListenAndServe.
   }
   ```
 
-  `StartFlowsServer` starts a `net/http` server that exposes each of the flows
-  you defined as HTTP endpoints
-  (for example, `http://localhost:3400/menuSuggestionFlow`).
-  You can optionally specify the address and port to listen on. If you don't,
-  the server listens on any address and the port specified by the PORT
-  environment variable; if that is empty, it uses the default of port 3400.
+  `StartFlowsServer` starts a `net/http` server that exposes your flows as HTTP
+  endpoints (for example, `http://localhost:3400/menuSuggestionFlow`). Both
+  parameters are optional:
+
+  - You can specify the address and port to listen on. If you don't,
+    the server listens on any address and the port specified by the PORT
+    environment variable; if that is empty, it uses the default of port 3400.
+  - You can specify which flows to serve. If you don't, `StartFlowsServer`
+    serves all of your defined flows.
 
   If you want to serve flows on the same host and port as other endpoints, you
   can call `NewFlowServeMux()` to get a handler for your Genkit flows, which you
diff --git a/docs-go/models.md b/docs-go/models.md
index 1d70b6a8e..9af6a52b3 100644
--- a/docs-go/models.md
+++ b/docs-go/models.md
@@ -24,13 +24,7 @@ models.
 
   ```go
   projectID := os.Getenv("GCLOUD_PROJECT")
-  err := vertexai.Init(context.Background(), vertexai.Config{
-    ProjectID: projectID,
-    Models: []string{
-      "gemini-1.5-pro",
-      "gemini-1.5-flash",
-    },
-  })
+  err := vertexai.Init(context.Background(), projectID, "us-central1")
   ```
 
 Note: Different plugins and models use different methods of
@@ -78,7 +72,7 @@ To just call the model:
   request := ai.GenerateRequest{Messages: []*ai.Message{
     {Content: []*ai.Part{ai.NewTextPart("Tell me a joke.")}},
   }}
-  response, err := ai.Generate(context.Background(), gemini15pro, &request, nil)
+  gemini15pro.Generate(context.Background(), &request, nil)
 
   responseText, err := response.Text()
   fmt.Println(responseText)
@@ -114,9 +108,8 @@ Genkit supports chunked streaming of model responses:
   request := ai.GenerateRequest{Messages: []*ai.Message{
     {Content: []*ai.Part{ai.NewTextPart("Tell a long story about robots and ninjas.")}},
   }}
-  response, err := ai.Generate(
+  response, err := gemini15pro.Generate(
     context.Background(),
-    gemini15pro,
     &request,
     func(ctx context.Context, grc *ai.GenerateResponseChunk) error {
       text, err := grc.Text()
@@ -147,7 +140,7 @@ If the model supports multimodal input, you can pass image prompts:
       ai.NewMediaPart("", "data:image/jpeg;base64,"+encodedImage),
     }},
   }}
-  response, err := ai.Generate(context.Background(), gemini15pro, &request, nil)
+  gemini15pro.Generate(context.Background(), &request, nil)
   ```
 
   <!-- TODO: gs:// wasn't working for me. HTTP? -->
@@ -186,7 +179,7 @@ it.
     },
     Tools: []*ai.ToolDefinition{myJoke},
   }
-  response, err := ai.Generate(context.Background(), gemini15pro, &request, nil)
+  gemini15pro.Generate(context.Background(), &request, nil)
   ```
 
 This will automatically call the tools in order to fulfill the user prompt.
@@ -232,7 +225,7 @@ chatbots.
   }
 
   request := ai.GenerateRequest{Messages: history}
-  response, err := ai.Generate(context.Background(), gemini15pro, &request, nil)
+  gemini15pro.Generate(context.Background(), &request, nil)
   ```
 
   When you get a response, add it to the history:
@@ -252,7 +245,7 @@ chatbots.
   })
 
   request := ai.GenerateRequest{Messages: history}
-  response, err := ai.Generate(context.Background(), gemini15pro, &request, nil)
+  gemini15pro.Generate(context.Background(), &request, nil)
   ```
 
 If the model you're using supports the system role, you can use the initial