feat(go): add OpenAI plugin

go/README.md

@@ -474,6 +474,7 @@ Genkit provides a unified interface across all major AI providers. Use whichever
 | **Google AI** | `googlegenai.GoogleAI` | Gemini 2.5 Flash, Gemini 2.5 Pro, and more |
 | **Vertex AI** | `vertexai.VertexAI` | Gemini models via Google Cloud |
 | **Anthropic** | `anthropic.Anthropic` | Claude 3.5, Claude 3 Opus, and more |
+| **OpenAI** | `openai.OpenAI` | GPT-5, GPT-5-mini, GPT-5-nano, GPT-4o and more |
 | **Ollama** | `ollama.Ollama` | Llama, Mistral, and other open models |
 | **OpenAI Compatible** | `compat_oai` | Any OpenAI-compatible API |
 
@@ -484,6 +485,9 @@ g := genkit.Init(ctx, genkit.WithPlugins(&googlegenai.GoogleAI{}))
 // Anthropic
 g := genkit.Init(ctx, genkit.WithPlugins(&anthropic.Anthropic{}))
 
+// OpenAI
+g := genkit.Init(ctx, genkit.WithPlugins(&openai.OpenAI{}))
+
 // Ollama (local models)
 g := genkit.Init(ctx, genkit.WithPlugins(&ollama.Ollama{
     ServerAddress: "http://localhost:11434",

go/go.mod

@@ -28,6 +28,7 @@ require (
 	github.com/jba/slog v0.2.0
 	github.com/lib/pq v1.10.9
 	github.com/mark3labs/mcp-go v0.29.0
+	github.com/openai/openai-go/v3 v3.16.0
 	github.com/pgvector/pgvector-go v0.3.0
 	github.com/stretchr/testify v1.10.0
 	github.com/weaviate/weaviate v1.30.0

go/go.sum

@@ -308,6 +308,8 @@ github.com/oklog/ulid v1.3.1 h1:EGfNDEx6MqHz8B3uNV6QAib1UR2Lm97sHi3ocA6ESJ4=
 github.com/oklog/ulid v1.3.1/go.mod h1:CirwcVhetQ6Lv90oh/F+FBtV6XMibvdAFo93nm5qn4U=
 github.com/openai/openai-go v1.8.2 h1:UqSkJ1vCOPUpz9Ka5tS0324EJFEuOvMc+lA/EarJWP8=
 github.com/openai/openai-go v1.8.2/go.mod h1:g461MYGXEXBVdV5SaR/5tNzNbSfwTBBefwc+LlDCK0Y=
+github.com/openai/openai-go/v3 v3.16.0 h1:VdqS+GFZgAvEOBcWNyvLVwPlYEIboW5xwiUCcLrVf8c=
+github.com/openai/openai-go/v3 v3.16.0/go.mod h1:cdufnVK14cWcT9qA1rRtrXx4FTRsgbDPW7Ia7SS5cZo=
 github.com/opentracing/opentracing-go v1.2.0 h1:uEJPy/1a5RIPAJ0Ov+OIO8OxWu77jEv+1B0VhjKrZUs=
 github.com/opentracing/opentracing-go v1.2.0/go.mod h1:GxEUsuufX4nBwe+T+Wl9TAgYrxe9dPLANfrWvHYVTgc=
 github.com/pelletier/go-toml v1.7.0/go.mod h1:vwGMzjaWMwyfHwgIBhI2YUM4fB6nL6lVAvS1LBMMhTE=

go/plugins/internal/models.go

@@ -46,4 +46,14 @@ var (
 		Media:       true,
 		Constrained: ai.ConstrainedSupportNone,
 	}
+
+	// Media describes model capabilities for models with media and text input and output
+	Media = ai.ModelSupports{
+		Multiturn:   false,
+		Tools:       false,
+		ToolChoice:  false,
+		SystemRole:  false,
+		Media:       true,
+		Constrained: ai.ConstrainedSupportNone,
+	}
 )

go/plugins/openai/generate.go

@@ -0,0 +1,135 @@
+// Copyright 2026 Google LLC
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package openai
+
+import (
+	"context"
+	"fmt"
+
+	"github.com/firebase/genkit/go/ai"
+	"github.com/openai/openai-go/v3"
+	"github.com/openai/openai-go/v3/responses"
+)
+
+// generate is the entry point function to request content generation to the OpenAI client
+func generate(ctx context.Context, client *openai.Client, model string, input *ai.ModelRequest, cb func(context.Context, *ai.ModelResponseChunk) error,
+) (*ai.ModelResponse, error) {
+	req, err := toOpenAIResponseParams(model, input)
+	if err != nil {
+		return nil, err
+	}
+
+	// stream mode
+	if cb != nil {
+		resp, err := generateStream(ctx, client, req, input, cb)
+		if err != nil {
+			return nil, err
+		}
+		return resp, nil
+
+	}
+
+	resp, err := generateComplete(ctx, client, req, input)
+	if err != nil {
+		return nil, err
+	}
+	return resp, nil
+}
+
+// generateStream starts a new streaming response
+func generateStream(ctx context.Context, client *openai.Client, req *responses.ResponseNewParams, input *ai.ModelRequest, cb func(context.Context, *ai.ModelResponseChunk) error) (*ai.ModelResponse, error) {
+	stream := client.Responses.NewStreaming(ctx, *req)
+	defer stream.Close()
+
+	var (
+		toolRefMap = make(map[string]string)
+		finalResp  *responses.Response
+	)
+
+	for stream.Next() {
+		evt := stream.Current()
+		chunk := &ai.ModelResponseChunk{}
+
+		switch v := evt.AsAny().(type) {
+		case responses.ResponseTextDeltaEvent:
+			chunk.Content = append(chunk.Content, ai.NewTextPart(v.Delta))
+
+		case responses.ResponseReasoningTextDeltaEvent:
+			chunk.Content = append(chunk.Content, ai.NewReasoningPart(v.Delta, nil))
+
+		case responses.ResponseFunctionCallArgumentsDeltaEvent:
+			name := toolRefMap[v.ItemID]
+			chunk.Content = append(chunk.Content, ai.NewToolRequestPart(&ai.ToolRequest{
+				Ref:   v.ItemID,
+				Name:  name,
+				Input: v.Delta,
+			}))
+
+		case responses.ResponseOutputItemAddedEvent:
+			switch item := v.Item.AsAny().(type) {
+			case responses.ResponseFunctionToolCall:
+				toolRefMap[item.CallID] = item.Name
+				chunk.Content = append(chunk.Content, ai.NewToolRequestPart(&ai.ToolRequest{
+					Ref:  item.CallID,
+					Name: item.Name,
+				}))
+			}
+
+		case responses.ResponseCompletedEvent:
+			finalResp = &v.Response
+		}
+
+		if len(chunk.Content) > 0 {
+			if err := cb(ctx, chunk); err != nil {
+				return nil, fmt.Errorf("callback error: %w", err)
+			}
+		}
+	}
+
+	if err := stream.Err(); err != nil {
+		return nil, fmt.Errorf("stream error: %w", err)
+	}
+
+	if finalResp != nil {
+		mResp, err := translateResponse(finalResp)
+		if err != nil {
+			return nil, err
+		}
+		mResp.Request = input
+		return mResp, nil
+	}
+
+	// prevent returning an error if stream does not provide [responses.ResponseCompletedEvent]
+	// user might already have received the chunks throughout the loop
+	return &ai.ModelResponse{
+		Request: input,
+		Message: &ai.Message{Role: ai.RoleModel},
+	}, nil
+}
+
+// generateComplete starts a new completion
+func generateComplete(ctx context.Context, client *openai.Client, req *responses.ResponseNewParams, input *ai.ModelRequest) (*ai.ModelResponse, error) {
+	resp, err := client.Responses.New(ctx, *req)
+	if err != nil {
+		return nil, err
+	}
+
+	modelResp, err := translateResponse(resp)
+	if err != nil {
+		return nil, err
+	}
+	modelResp.Request = input
+	return modelResp, nil
+}