Add GPT-5.6 to Azure and Vercel providers

commit: dbd0dfa9d4f1f2ad863881da93884a7b6be49dcb
author: patriceckhart · 3mo ago · 2026-07-10 05:56 UTC
parent: 1fa7669
signature: unsigned
refs: tag: v0.2.73

8 files · +207 -31

M docs/providers.md +7 -3


@@ -162,12 +162,16 @@ zot --provider cloudflare-workers-ai
162	162		```bash
163	163		export AZURE_OPENAI_API_KEY=...
164	164		export AZURE_OPENAI_BASE_URL=https://your-resource.openai.azure.com
165		-	export AZURE_OPENAI_API_VERSION=2024-02-01 # optional
	165	+	export AZURE_OPENAI_API_VERSION=v1 # optional, v1 is the default
166	166		zot --provider azure-openai-responses
167	167		```
168	168
169		-	If your Azure deployment names differ from zot model IDs, add model overrides
170		-	in `$ZOT_HOME/models.json`.
	169	+	The provider uses Azure's Responses API. If deployment names differ from zot
	170	+	model IDs, map them without changing the catalog:
	171	+
	172	+	```bash
	173	+	export AZURE_OPENAI_DEPLOYMENT_NAME_MAP='gpt-5.6-luna=luna-preview,gpt-5.6-sol=sol-preview'
	174	+	```
171	175
172	176		## Auth file
173	177

 ```bash
 export AZURE_OPENAI_API_KEY=...
 export AZURE_OPENAI_BASE_URL=https://your-resource.openai.azure.com
-export AZURE_OPENAI_API_VERSION=2024-02-01 # optional
+export AZURE_OPENAI_API_VERSION=v1 # optional, v1 is the default
 zot --provider azure-openai-responses
 ```
-If your Azure deployment names differ from zot model IDs, add model overrides
+The provider uses Azure's Responses API. If deployment names differ from zot
-in `$ZOT_HOME/models.json`.
+model IDs, map them without changing the catalog:
+```bash
+export AZURE_OPENAI_DEPLOYMENT_NAME_MAP='gpt-5.6-luna=luna-preview,gpt-5.6-sol=sol-preview'
+```
 ## Auth file

M packages/provider/azure_openai.go +90 -1


@@ -30,7 +30,10 @@ import (
30	30		"strings"
31	31		)
32	32
33		-	const defaultAzureAPIVersion = "2024-10-21"
	33	+	const (
	34	+	defaultAzureAPIVersion = "2024-10-21"
	35	+	defaultAzureResponsesAPIVersion = "v1"
	36	+	)
34	37
35	38		// azureRewriteTransport rewrites every outgoing OpenAI Chat Completions
36	39		// URL into the Azure deployment-scoped shape:

@@ -134,3 +137,89 @@ func NewAzureOpenAI(apiKey, baseURL string) Client {
134	137		http: httpClient,
135	138		}
136	139		}
	140	+
	141	+	// azureResponsesTransport applies Azure API-key authentication and API
	142	+	// versioning to requests using the OpenAI Responses wire format.
	143	+	type azureResponsesTransport struct {
	144	+	inner http.RoundTripper
	145	+	apiKey string
	146	+	apiVersion string
	147	+	}
	148	+
	149	+	func (t azureResponsesTransport) RoundTrip(req http.Request) (*http.Response, error) {
	150	+	clone := req.Clone(req.Context())
	151	+	clone.Header.Del("authorization")
	152	+	clone.Header.Del("chatgpt-account-id")
	153	+	clone.Header.Del("openai-beta")
	154	+	clone.Header.Del("originator")
	155	+	clone.Header.Set("api-key", t.apiKey)
	156	+	q := clone.URL.Query()
	157	+	q.Set("api-version", t.apiVersion)
	158	+	clone.URL.RawQuery = q.Encode()
	159	+	return t.inner.RoundTrip(clone)
	160	+	}
	161	+
	162	+	// NewAzureOpenAIResponses creates an Azure OpenAI Responses API client.
	163	+	// The configured model id is used as the deployment name unless overridden
	164	+	// through AZURE_OPENAI_DEPLOYMENT_NAME_MAP (model=deployment pairs).
	165	+	func NewAzureOpenAIResponses(apiKey, baseURL string) Client {
	166	+	if baseURL == "" {
	167	+	baseURL = os.Getenv("AZURE_OPENAI_BASE_URL")
	168	+	if baseURL == "" {
	169	+	if resource := os.Getenv("AZURE_OPENAI_RESOURCE_NAME"); resource != "" {
	170	+	baseURL = "https://" + resource + ".openai.azure.com"
	171	+	}
	172	+	}
	173	+	}
	174	+	if baseURL == "" {
	175	+	return &unimplementedClient{
	176	+	name: "azure-openai-responses",
	177	+	hint: "set AZURE_OPENAI_BASE_URL or AZURE_OPENAI_RESOURCE_NAME (or pass --base-url)",
	178	+	}
	179	+	}
	180	+
	181	+	endpoint := strings.TrimRight(baseURL, "/")
	182	+	switch {
	183	+	case strings.HasSuffix(endpoint, "/openai/v1/responses"):
	184	+	case strings.HasSuffix(endpoint, "/openai/v1"):
	185	+	endpoint += "/responses"
	186	+	case strings.HasSuffix(endpoint, "/openai"):
	187	+	endpoint += "/v1/responses"
	188	+	default:
	189	+	endpoint += "/openai/v1/responses"
	190	+	}
	191	+	apiVersion := os.Getenv("AZURE_OPENAI_API_VERSION")
	192	+	if apiVersion == "" {
	193	+	apiVersion = defaultAzureResponsesAPIVersion
	194	+	}
	195	+	httpClient := &http.Client{
	196	+	Transport: &azureResponsesTransport{
	197	+	inner: http.DefaultTransport,
	198	+	apiKey: apiKey,
	199	+	apiVersion: apiVersion,
	200	+	},
	201	+	Timeout: 0,
	202	+	}
	203	+	inner := &codexClient{
	204	+	token: apiKey,
	205	+	baseURL: endpoint,
	206	+	errorLabel: "azure openai",
	207	+	providerName: "azure-openai-responses",
	208	+	modelName: azureDeploymentName,
	209	+	disableCLIRouting: true,
	210	+	http: httpClient,
	211	+	}
	212	+	return &renamedClient{inner: inner, name: "azure-openai-responses"}
	213	+	}
	214	+
	215	+	func azureDeploymentName(model string) string {
	216	+	for _, pair := range strings.Split(os.Getenv("AZURE_OPENAI_DEPLOYMENT_NAME_MAP"), ",") {
	217	+	parts := strings.SplitN(strings.TrimSpace(pair), "=", 2)
	218	+	if len(parts) == 2 && strings.TrimSpace(parts[0]) == model {
	219	+	if deployment := strings.TrimSpace(parts[1]); deployment != "" {
	220	+	return deployment
	221	+	}
	222	+	}
	223	+	}
	224	+	return model
	225	+	}


@@ -30,7 +30,10 @@ import (
30		"strings"	30		"strings"
31		)	31		)
32			32
33	-	const defaultAzureAPIVersion = "2024-10-21"	33	+	const (
			34	+	defaultAzureAPIVersion = "2024-10-21"
			35	+	defaultAzureResponsesAPIVersion = "v1"
			36	+	)
34			37
35		// azureRewriteTransport rewrites every outgoing OpenAI Chat Completions	38		// azureRewriteTransport rewrites every outgoing OpenAI Chat Completions
36		// URL into the Azure deployment-scoped shape:	39		// URL into the Azure deployment-scoped shape:

@@ -134,3 +137,89 @@ func NewAzureOpenAI(apiKey, baseURL string) Client {
134		http: httpClient,	137		http: httpClient,
135		}	138		}
136		}	139		}
			140	+
			141	+	// azureResponsesTransport applies Azure API-key authentication and API
			142	+	// versioning to requests using the OpenAI Responses wire format.
			143	+	type azureResponsesTransport struct {
			144	+	inner http.RoundTripper
			145	+	apiKey string
			146	+	apiVersion string
			147	+	}
			148	+
			149	+	func (t azureResponsesTransport) RoundTrip(req http.Request) (*http.Response, error) {
			150	+	clone := req.Clone(req.Context())
			151	+	clone.Header.Del("authorization")
			152	+	clone.Header.Del("chatgpt-account-id")
			153	+	clone.Header.Del("openai-beta")
			154	+	clone.Header.Del("originator")
			155	+	clone.Header.Set("api-key", t.apiKey)
			156	+	q := clone.URL.Query()
			157	+	q.Set("api-version", t.apiVersion)
			158	+	clone.URL.RawQuery = q.Encode()
			159	+	return t.inner.RoundTrip(clone)
			160	+	}
			161	+
			162	+	// NewAzureOpenAIResponses creates an Azure OpenAI Responses API client.
			163	+	// The configured model id is used as the deployment name unless overridden
			164	+	// through AZURE_OPENAI_DEPLOYMENT_NAME_MAP (model=deployment pairs).
			165	+	func NewAzureOpenAIResponses(apiKey, baseURL string) Client {
			166	+	if baseURL == "" {
			167	+	baseURL = os.Getenv("AZURE_OPENAI_BASE_URL")
			168	+	if baseURL == "" {
			169	+	if resource := os.Getenv("AZURE_OPENAI_RESOURCE_NAME"); resource != "" {
			170	+	baseURL = "https://" + resource + ".openai.azure.com"
			171	+	}
			172	+	}
			173	+	}
			174	+	if baseURL == "" {
			175	+	return &unimplementedClient{
			176	+	name: "azure-openai-responses",
			177	+	hint: "set AZURE_OPENAI_BASE_URL or AZURE_OPENAI_RESOURCE_NAME (or pass --base-url)",
			178	+	}
			179	+	}
			180	+
			181	+	endpoint := strings.TrimRight(baseURL, "/")
			182	+	switch {
			183	+	case strings.HasSuffix(endpoint, "/openai/v1/responses"):
			184	+	case strings.HasSuffix(endpoint, "/openai/v1"):
			185	+	endpoint += "/responses"
			186	+	case strings.HasSuffix(endpoint, "/openai"):
			187	+	endpoint += "/v1/responses"
			188	+	default:
			189	+	endpoint += "/openai/v1/responses"
			190	+	}
			191	+	apiVersion := os.Getenv("AZURE_OPENAI_API_VERSION")
			192	+	if apiVersion == "" {
			193	+	apiVersion = defaultAzureResponsesAPIVersion
			194	+	}
			195	+	httpClient := &http.Client{
			196	+	Transport: &azureResponsesTransport{
			197	+	inner: http.DefaultTransport,
			198	+	apiKey: apiKey,
			199	+	apiVersion: apiVersion,
			200	+	},
			201	+	Timeout: 0,
			202	+	}
			203	+	inner := &codexClient{
			204	+	token: apiKey,
			205	+	baseURL: endpoint,
			206	+	errorLabel: "azure openai",
			207	+	providerName: "azure-openai-responses",
			208	+	modelName: azureDeploymentName,
			209	+	disableCLIRouting: true,
			210	+	http: httpClient,
			211	+	}
			212	+	return &renamedClient{inner: inner, name: "azure-openai-responses"}
			213	+	}
			214	+
			215	+	func azureDeploymentName(model string) string {
			216	+	for _, pair := range strings.Split(os.Getenv("AZURE_OPENAI_DEPLOYMENT_NAME_MAP"), ",") {
			217	+	parts := strings.SplitN(strings.TrimSpace(pair), "=", 2)
			218	+	if len(parts) == 2 && strings.TrimSpace(parts[0]) == model {
			219	+	if deployment := strings.TrimSpace(parts[1]); deployment != "" {
			220	+	return deployment
			221	+	}
			222	+	}
			223	+	}
			224	+	return model
			225	+	}

A packages/provider/azure_openai_test.go +69 -0

@@ -0,0 +1,69 @@
	1	+	package provider
	2	+
	3	+	import (
	4	+	"context"
	5	+	"encoding/json"
	6	+	"io"
	7	+	"net/http"
	8	+	"strings"
	9	+	"testing"
	10	+	)
	11	+
	12	+	func TestAzureOpenAIResponsesRequestShape(t *testing.T) {
	13	+	t.Setenv("AZURE_OPENAI_API_VERSION", "v1")
	14	+	t.Setenv("AZURE_OPENAI_DEPLOYMENT_NAME_MAP", "gpt-5.6-sol=sol-preview")
	15	+
	16	+	named := NewAzureOpenAIResponses("azure-key", "https://example.openai.azure.com").(*renamedClient)
	17	+	c := named.inner.(*codexClient)
	18	+	transport := c.http.Transport.(*azureResponsesTransport)
	19	+	var gotReq *http.Request
	20	+	var gotBody codexRequest
	21	+	transport.inner = roundTripFunc(func(r http.Request) (http.Response, error) {
	22	+	gotReq = r
	23	+	if err := json.NewDecoder(r.Body).Decode(&gotBody); err != nil {
	24	+	t.Fatal(err)
	25	+	}
	26	+	return &http.Response{
	27	+	StatusCode: http.StatusOK,
	28	+	Header: http.Header{"Content-Type": []string{"text/event-stream"}},
	29	+	Body: io.NopCloser(strings.NewReader("")),
	30	+	}, nil
	31	+	})
	32	+
	33	+	events, err := c.Stream(context.Background(), Request{
	34	+	Model: "gpt-5.6-sol",
	35	+	Reasoning: "max",
	36	+	Messages: []Message{{Role: RoleUser, Content: []Content{TextBlock{Text: "hi"}}}},
	37	+	})
	38	+	if err != nil {
	39	+	t.Fatal(err)
	40	+	}
	41	+	var startProvider string
	42	+	for event := range events {
	43	+	if start, ok := event.(EventStart); ok {
	44	+	startProvider = start.Provider
	45	+	}
	46	+	}
	47	+
	48	+	if startProvider != "azure-openai-responses" {
	49	+	t.Fatalf("start provider = %q", startProvider)
	50	+	}
	51	+	if gotReq == nil {
	52	+	t.Fatal("request was not sent")
	53	+	}
	54	+	if gotReq.URL.String() != "https://example.openai.azure.com/openai/v1/responses?api-version=v1" {
	55	+	t.Fatalf("url = %q", gotReq.URL.String())
	56	+	}
	57	+	if gotReq.Header.Get("api-key") != "azure-key" \|\| gotReq.Header.Get("authorization") != "" {
	58	+	t.Fatalf("headers = %#v", gotReq.Header)
	59	+	}
	60	+	if gotBody.Model != "sol-preview" {
	61	+	t.Fatalf("model = %q", gotBody.Model)
	62	+	}
	63	+	if gotBody.Reasoning == nil \|\| gotBody.Reasoning.Effort != "max" {
	64	+	t.Fatalf("reasoning = %+v", gotBody.Reasoning)
	65	+	}
	66	+	if gotBody.PromptCacheKey != "" {
	67	+	t.Fatalf("prompt_cache_key = %q", gotBody.PromptCacheKey)
	68	+	}
	69	+	}

@@ -0,0 +1,69 @@
	1	+	package provider
	2	+
	3	+	import (
	4	+	"context"
	5	+	"encoding/json"
	6	+	"io"
	7	+	"net/http"
	8	+	"strings"
	9	+	"testing"
	10	+	)
	11	+
	12	+	func TestAzureOpenAIResponsesRequestShape(t *testing.T) {
	13	+	t.Setenv("AZURE_OPENAI_API_VERSION", "v1")
	14	+	t.Setenv("AZURE_OPENAI_DEPLOYMENT_NAME_MAP", "gpt-5.6-sol=sol-preview")
	15	+
	16	+	named := NewAzureOpenAIResponses("azure-key", "https://example.openai.azure.com").(*renamedClient)
	17	+	c := named.inner.(*codexClient)
	18	+	transport := c.http.Transport.(*azureResponsesTransport)
	19	+	var gotReq *http.Request
	20	+	var gotBody codexRequest
	21	+	transport.inner = roundTripFunc(func(r http.Request) (http.Response, error) {
	22	+	gotReq = r
	23	+	if err := json.NewDecoder(r.Body).Decode(&gotBody); err != nil {
	24	+	t.Fatal(err)
	25	+	}
	26	+	return &http.Response{
	27	+	StatusCode: http.StatusOK,
	28	+	Header: http.Header{"Content-Type": []string{"text/event-stream"}},
	29	+	Body: io.NopCloser(strings.NewReader("")),
	30	+	}, nil
	31	+	})
	32	+
	33	+	events, err := c.Stream(context.Background(), Request{
	34	+	Model: "gpt-5.6-sol",
	35	+	Reasoning: "max",
	36	+	Messages: []Message{{Role: RoleUser, Content: []Content{TextBlock{Text: "hi"}}}},
	37	+	})
	38	+	if err != nil {
	39	+	t.Fatal(err)
	40	+	}
	41	+	var startProvider string
	42	+	for event := range events {
	43	+	if start, ok := event.(EventStart); ok {
	44	+	startProvider = start.Provider
	45	+	}
	46	+	}
	47	+
	48	+	if startProvider != "azure-openai-responses" {
	49	+	t.Fatalf("start provider = %q", startProvider)
	50	+	}
	51	+	if gotReq == nil {
	52	+	t.Fatal("request was not sent")
	53	+	}
	54	+	if gotReq.URL.String() != "https://example.openai.azure.com/openai/v1/responses?api-version=v1" {
	55	+	t.Fatalf("url = %q", gotReq.URL.String())
	56	+	}
	57	+	if gotReq.Header.Get("api-key") != "azure-key" \|\| gotReq.Header.Get("authorization") != "" {
	58	+	t.Fatalf("headers = %#v", gotReq.Header)
	59	+	}
	60	+	if gotBody.Model != "sol-preview" {
	61	+	t.Fatalf("model = %q", gotBody.Model)
	62	+	}
	63	+	if gotBody.Reasoning == nil \|\| gotBody.Reasoning.Effort != "max" {
	64	+	t.Fatalf("reasoning = %+v", gotBody.Reasoning)
	65	+	}
	66	+	if gotBody.PromptCacheKey != "" {
	67	+	t.Fatalf("prompt_cache_key = %q", gotBody.PromptCacheKey)
	68	+	}
	69	+	}

M packages/provider/catalog_builtin.go +6 -0


@@ -466,6 +466,9 @@ var builtinCatalog = []Model{
466	466		{Provider: "vercel-ai-gateway", ID: "openai/gpt-5.4-pro", DisplayName: "GPT 5.4 Pro", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 30, PriceOutput: 180, PriceCacheRead: 0, BaseURL: "https://ai-gateway.vercel.sh"},
467	467		{Provider: "vercel-ai-gateway", ID: "openai/gpt-5.5", DisplayName: "GPT 5.5", ContextWindow: 1000000, MaxOutput: 128000, Reasoning: true, PriceInput: 5, PriceOutput: 30, PriceCacheRead: 0.5, BaseURL: "https://ai-gateway.vercel.sh"},
468	468		{Provider: "vercel-ai-gateway", ID: "openai/gpt-5.5-pro", DisplayName: "GPT 5.5 Pro", ContextWindow: 1000000, MaxOutput: 128000, Reasoning: true, PriceInput: 30, PriceOutput: 180, PriceCacheRead: 0, BaseURL: "https://ai-gateway.vercel.sh"},
	469	+	{Provider: "vercel-ai-gateway", ID: "openai/gpt-5.6-luna", DisplayName: "GPT 5.6 Luna", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 1, PriceOutput: 6, PriceCacheRead: 0.1, PriceCacheWrite: 1.25, BaseURL: "https://ai-gateway.vercel.sh"},
	470	+	{Provider: "vercel-ai-gateway", ID: "openai/gpt-5.6-sol", DisplayName: "GPT 5.6 Sol", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 5, PriceOutput: 30, PriceCacheRead: 0.5, PriceCacheWrite: 6.25, BaseURL: "https://ai-gateway.vercel.sh"},
	471	+	{Provider: "vercel-ai-gateway", ID: "openai/gpt-5.6-terra", DisplayName: "GPT 5.6 Terra", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 2.5, PriceOutput: 15, PriceCacheRead: 0.25, PriceCacheWrite: 3.125, BaseURL: "https://ai-gateway.vercel.sh"},
469	472		{Provider: "vercel-ai-gateway", ID: "openai/gpt-oss-20b", DisplayName: "GPT OSS 20B", ContextWindow: 131072, MaxOutput: 8192, Reasoning: true, PriceInput: 0.05, PriceOutput: 0.2, PriceCacheRead: 0, BaseURL: "https://ai-gateway.vercel.sh"},
470	473		{Provider: "vercel-ai-gateway", ID: "openai/gpt-oss-safeguard-20b", DisplayName: "GPT OSS Safeguard 20B", ContextWindow: 131072, MaxOutput: 65536, Reasoning: true, PriceInput: 0.075, PriceOutput: 0.3, PriceCacheRead: 0.037, BaseURL: "https://ai-gateway.vercel.sh"},
471	474		{Provider: "vercel-ai-gateway", ID: "openai/o1", DisplayName: "o1", ContextWindow: 200000, MaxOutput: 100000, Reasoning: true, PriceInput: 15, PriceOutput: 60, PriceCacheRead: 7.5, BaseURL: "https://ai-gateway.vercel.sh"},

@@ -622,6 +625,9 @@ var builtinCatalog = []Model{
622	625		{Provider: "azure-openai-responses", ID: "gpt-5.4-pro", DisplayName: "GPT-5.4 Pro", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 30, PriceOutput: 180, PriceCacheRead: 0, BaseURL: ""},
623	626		{Provider: "azure-openai-responses", ID: "gpt-5.5", DisplayName: "GPT-5.5", ContextWindow: 272000, MaxOutput: 128000, Reasoning: true, PriceInput: 5, PriceOutput: 30, PriceCacheRead: 0.5, BaseURL: ""},
624	627		{Provider: "azure-openai-responses", ID: "gpt-5.5-pro", DisplayName: "GPT-5.5 Pro", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 30, PriceOutput: 180, PriceCacheRead: 0, BaseURL: ""},
	628	+	{Provider: "azure-openai-responses", ID: "gpt-5.6-luna", DisplayName: "GPT-5.6 Luna", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 1, PriceOutput: 6, PriceCacheRead: 0.1, PriceCacheWrite: 1.25, BaseURL: ""},
	629	+	{Provider: "azure-openai-responses", ID: "gpt-5.6-sol", DisplayName: "GPT-5.6 Sol", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 5, PriceOutput: 30, PriceCacheRead: 0.5, PriceCacheWrite: 6.25, BaseURL: ""},
	630	+	{Provider: "azure-openai-responses", ID: "gpt-5.6-terra", DisplayName: "GPT-5.6 Terra", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 2.5, PriceOutput: 15, PriceCacheRead: 0.25, PriceCacheWrite: 3.125, BaseURL: ""},
625	631		{Provider: "azure-openai-responses", ID: "o1", DisplayName: "o1", ContextWindow: 200000, MaxOutput: 100000, Reasoning: true, PriceInput: 15, PriceOutput: 60, PriceCacheRead: 7.5, BaseURL: ""},
626	632		{Provider: "azure-openai-responses", ID: "o1-pro", DisplayName: "o1-pro", ContextWindow: 200000, MaxOutput: 100000, Reasoning: true, PriceInput: 150, PriceOutput: 600, PriceCacheRead: 0, BaseURL: ""},
627	633		{Provider: "azure-openai-responses", ID: "o3", DisplayName: "o3", ContextWindow: 200000, MaxOutput: 100000, Reasoning: true, PriceInput: 2, PriceOutput: 8, PriceCacheRead: 0.5, BaseURL: ""},


@@ -466,6 +466,9 @@ var builtinCatalog = []Model{
466	{Provider: "vercel-ai-gateway", ID: "openai/gpt-5.4-pro", DisplayName: "GPT 5.4 Pro", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 30, PriceOutput: 180, PriceCacheRead: 0, BaseURL: "https://ai-gateway.vercel.sh"},	466		{Provider: "vercel-ai-gateway", ID: "openai/gpt-5.4-pro", DisplayName: "GPT 5.4 Pro", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 30, PriceOutput: 180, PriceCacheRead: 0, BaseURL: "https://ai-gateway.vercel.sh"},
467	{Provider: "vercel-ai-gateway", ID: "openai/gpt-5.5", DisplayName: "GPT 5.5", ContextWindow: 1000000, MaxOutput: 128000, Reasoning: true, PriceInput: 5, PriceOutput: 30, PriceCacheRead: 0.5, BaseURL: "https://ai-gateway.vercel.sh"},	467		{Provider: "vercel-ai-gateway", ID: "openai/gpt-5.5", DisplayName: "GPT 5.5", ContextWindow: 1000000, MaxOutput: 128000, Reasoning: true, PriceInput: 5, PriceOutput: 30, PriceCacheRead: 0.5, BaseURL: "https://ai-gateway.vercel.sh"},
468	{Provider: "vercel-ai-gateway", ID: "openai/gpt-5.5-pro", DisplayName: "GPT 5.5 Pro", ContextWindow: 1000000, MaxOutput: 128000, Reasoning: true, PriceInput: 30, PriceOutput: 180, PriceCacheRead: 0, BaseURL: "https://ai-gateway.vercel.sh"},	468		{Provider: "vercel-ai-gateway", ID: "openai/gpt-5.5-pro", DisplayName: "GPT 5.5 Pro", ContextWindow: 1000000, MaxOutput: 128000, Reasoning: true, PriceInput: 30, PriceOutput: 180, PriceCacheRead: 0, BaseURL: "https://ai-gateway.vercel.sh"},
		469	+	{Provider: "vercel-ai-gateway", ID: "openai/gpt-5.6-luna", DisplayName: "GPT 5.6 Luna", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 1, PriceOutput: 6, PriceCacheRead: 0.1, PriceCacheWrite: 1.25, BaseURL: "https://ai-gateway.vercel.sh"},
		470	+	{Provider: "vercel-ai-gateway", ID: "openai/gpt-5.6-sol", DisplayName: "GPT 5.6 Sol", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 5, PriceOutput: 30, PriceCacheRead: 0.5, PriceCacheWrite: 6.25, BaseURL: "https://ai-gateway.vercel.sh"},
		471	+	{Provider: "vercel-ai-gateway", ID: "openai/gpt-5.6-terra", DisplayName: "GPT 5.6 Terra", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 2.5, PriceOutput: 15, PriceCacheRead: 0.25, PriceCacheWrite: 3.125, BaseURL: "https://ai-gateway.vercel.sh"},
469	{Provider: "vercel-ai-gateway", ID: "openai/gpt-oss-20b", DisplayName: "GPT OSS 20B", ContextWindow: 131072, MaxOutput: 8192, Reasoning: true, PriceInput: 0.05, PriceOutput: 0.2, PriceCacheRead: 0, BaseURL: "https://ai-gateway.vercel.sh"},	472		{Provider: "vercel-ai-gateway", ID: "openai/gpt-oss-20b", DisplayName: "GPT OSS 20B", ContextWindow: 131072, MaxOutput: 8192, Reasoning: true, PriceInput: 0.05, PriceOutput: 0.2, PriceCacheRead: 0, BaseURL: "https://ai-gateway.vercel.sh"},
470	{Provider: "vercel-ai-gateway", ID: "openai/gpt-oss-safeguard-20b", DisplayName: "GPT OSS Safeguard 20B", ContextWindow: 131072, MaxOutput: 65536, Reasoning: true, PriceInput: 0.075, PriceOutput: 0.3, PriceCacheRead: 0.037, BaseURL: "https://ai-gateway.vercel.sh"},	473		{Provider: "vercel-ai-gateway", ID: "openai/gpt-oss-safeguard-20b", DisplayName: "GPT OSS Safeguard 20B", ContextWindow: 131072, MaxOutput: 65536, Reasoning: true, PriceInput: 0.075, PriceOutput: 0.3, PriceCacheRead: 0.037, BaseURL: "https://ai-gateway.vercel.sh"},
471	{Provider: "vercel-ai-gateway", ID: "openai/o1", DisplayName: "o1", ContextWindow: 200000, MaxOutput: 100000, Reasoning: true, PriceInput: 15, PriceOutput: 60, PriceCacheRead: 7.5, BaseURL: "https://ai-gateway.vercel.sh"},	474		{Provider: "vercel-ai-gateway", ID: "openai/o1", DisplayName: "o1", ContextWindow: 200000, MaxOutput: 100000, Reasoning: true, PriceInput: 15, PriceOutput: 60, PriceCacheRead: 7.5, BaseURL: "https://ai-gateway.vercel.sh"},

@@ -622,6 +625,9 @@ var builtinCatalog = []Model{
622	{Provider: "azure-openai-responses", ID: "gpt-5.4-pro", DisplayName: "GPT-5.4 Pro", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 30, PriceOutput: 180, PriceCacheRead: 0, BaseURL: ""},	625		{Provider: "azure-openai-responses", ID: "gpt-5.4-pro", DisplayName: "GPT-5.4 Pro", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 30, PriceOutput: 180, PriceCacheRead: 0, BaseURL: ""},
623	{Provider: "azure-openai-responses", ID: "gpt-5.5", DisplayName: "GPT-5.5", ContextWindow: 272000, MaxOutput: 128000, Reasoning: true, PriceInput: 5, PriceOutput: 30, PriceCacheRead: 0.5, BaseURL: ""},	626		{Provider: "azure-openai-responses", ID: "gpt-5.5", DisplayName: "GPT-5.5", ContextWindow: 272000, MaxOutput: 128000, Reasoning: true, PriceInput: 5, PriceOutput: 30, PriceCacheRead: 0.5, BaseURL: ""},
624	{Provider: "azure-openai-responses", ID: "gpt-5.5-pro", DisplayName: "GPT-5.5 Pro", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 30, PriceOutput: 180, PriceCacheRead: 0, BaseURL: ""},	627		{Provider: "azure-openai-responses", ID: "gpt-5.5-pro", DisplayName: "GPT-5.5 Pro", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 30, PriceOutput: 180, PriceCacheRead: 0, BaseURL: ""},
		628	+	{Provider: "azure-openai-responses", ID: "gpt-5.6-luna", DisplayName: "GPT-5.6 Luna", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 1, PriceOutput: 6, PriceCacheRead: 0.1, PriceCacheWrite: 1.25, BaseURL: ""},
		629	+	{Provider: "azure-openai-responses", ID: "gpt-5.6-sol", DisplayName: "GPT-5.6 Sol", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 5, PriceOutput: 30, PriceCacheRead: 0.5, PriceCacheWrite: 6.25, BaseURL: ""},
		630	+	{Provider: "azure-openai-responses", ID: "gpt-5.6-terra", DisplayName: "GPT-5.6 Terra", ContextWindow: 1050000, MaxOutput: 128000, Reasoning: true, PriceInput: 2.5, PriceOutput: 15, PriceCacheRead: 0.25, PriceCacheWrite: 3.125, BaseURL: ""},
625	{Provider: "azure-openai-responses", ID: "o1", DisplayName: "o1", ContextWindow: 200000, MaxOutput: 100000, Reasoning: true, PriceInput: 15, PriceOutput: 60, PriceCacheRead: 7.5, BaseURL: ""},	631		{Provider: "azure-openai-responses", ID: "o1", DisplayName: "o1", ContextWindow: 200000, MaxOutput: 100000, Reasoning: true, PriceInput: 15, PriceOutput: 60, PriceCacheRead: 7.5, BaseURL: ""},
626	{Provider: "azure-openai-responses", ID: "o1-pro", DisplayName: "o1-pro", ContextWindow: 200000, MaxOutput: 100000, Reasoning: true, PriceInput: 150, PriceOutput: 600, PriceCacheRead: 0, BaseURL: ""},	632		{Provider: "azure-openai-responses", ID: "o1-pro", DisplayName: "o1-pro", ContextWindow: 200000, MaxOutput: 100000, Reasoning: true, PriceInput: 150, PriceOutput: 600, PriceCacheRead: 0, BaseURL: ""},
627	{Provider: "azure-openai-responses", ID: "o3", DisplayName: "o3", ContextWindow: 200000, MaxOutput: 100000, Reasoning: true, PriceInput: 2, PriceOutput: 8, PriceCacheRead: 0.5, BaseURL: ""},	633		{Provider: "azure-openai-responses", ID: "o3", DisplayName: "o3", ContextWindow: 200000, MaxOutput: 100000, Reasoning: true, PriceInput: 2, PriceOutput: 8, PriceCacheRead: 0.5, BaseURL: ""},

M packages/provider/extra_providers.go +0 -10


@@ -253,16 +253,6 @@ func NewGoogleVertex(apiKey, baseURL string) Client {
253	253		return NewVertex(apiKey, baseURL)
254	254		}
255	255
256		-	// NewAzureOpenAIResponses delegates to the real Azure OpenAI client.
257		-	// Despite the provider id mentioning "responses", zot uses Azure's
258		-	// Chat Completions endpoint (older but functionally complete for our
259		-	// agent loop) to avoid duplicating the full openai-responses wire
260		-	// client. Models register under provider id `azure-openai-responses`
261		-	// so user catalogs keep working unchanged.
262		-	func NewAzureOpenAIResponses(apiKey, baseURL string) Client {
263		-	return NewAzureOpenAI(apiKey, baseURL)
264		-	}
265		-
266	256		// NewMistral returns a Mistral client using their OpenAI-compatible Chat
267	257		// Completions endpoint at https://api.mistral.ai/v1. Mistral also offers a
268	258		// bespoke "Conversations" API, but the OpenAI-compat endpoint supports


@@ -253,16 +253,6 @@ func NewGoogleVertex(apiKey, baseURL string) Client {
253		return NewVertex(apiKey, baseURL)	253	return NewVertex(apiKey, baseURL)
254		}	254	}
255			255
256	-	// NewAzureOpenAIResponses delegates to the real Azure OpenAI client.
257	-	// Despite the provider id mentioning "responses", zot uses Azure's
258	-	// Chat Completions endpoint (older but functionally complete for our
259	-	// agent loop) to avoid duplicating the full openai-responses wire
260	-	// client. Models register under provider id `azure-openai-responses`
261	-	// so user catalogs keep working unchanged.
262	-	func NewAzureOpenAIResponses(apiKey, baseURL string) Client {
263	-	return NewAzureOpenAI(apiKey, baseURL)
264	-	}
265	-
266		// NewMistral returns a Mistral client using their OpenAI-compatible Chat	256	// NewMistral returns a Mistral client using their OpenAI-compatible Chat
267		// Completions endpoint at https://api.mistral.ai/v1. Mistral also offers a	257	// Completions endpoint at https://api.mistral.ai/v1. Mistral also offers a
268		// bespoke "Conversations" API, but the OpenAI-compat endpoint supports	258	// bespoke "Conversations" API, but the OpenAI-compat endpoint supports

M packages/provider/openai_codex.go +23 -12


@@ -38,11 +38,14 @@ import (
38	38		const codexDefaultBaseURL = "https://chatgpt.com/backend-api/codex/responses"
39	39
40	40		type codexClient struct {
41		-	token string
42		-	accountID string
43		-	baseURL string
44		-	errorLabel string
45		-	http *http.Client
	41	+	token string
	42	+	accountID string
	43	+	baseURL string
	44	+	errorLabel string
	45	+	providerName string
	46	+	modelName func(string) string
	47	+	disableCLIRouting bool
	48	+	http *http.Client
46	49		}
47	50
48	51		// NewOpenAICodex creates a client that talks to ChatGPT's Codex endpoint

@@ -53,11 +56,12 @@ func NewOpenAICodex(token, accountID, baseURL string) Client {
53	56		baseURL = codexDefaultBaseURL
54	57		}
55	58		return &codexClient{
56		-	token: token,
57		-	accountID: accountID,
58		-	baseURL: strings.TrimRight(baseURL, "/"),
59		-	errorLabel: "codex",
60		-	http: &http.Client{Timeout: 0},
	59	+	token: token,
	60	+	accountID: accountID,
	61	+	baseURL: strings.TrimRight(baseURL, "/"),
	62	+	errorLabel: "codex",
	63	+	providerName: "openai-codex",
	64	+	http: &http.Client{Timeout: 0},
61	65		}
62	66		}
63	67

@@ -345,8 +349,11 @@ func (c *codexClient) Stream(ctx context.Context, req Request) (<-chan Event, er
345	349		if err != nil {
346	350		return nil, err
347	351		}
	352	+	if c.modelName != nil {
	353	+	wire.Model = c.modelName(wire.Model)
	354	+	}
348	355		var codexCLISessionID string
349		-	if usesCodexCLIRouting(wire.Model) {
	356	+	if !c.disableCLIRouting && usesCodexCLIRouting(wire.Model) {
350	357		codexCLISessionID = newCodexSessionID()
351	358		wire.PromptCacheKey = codexCLISessionID
352	359		}

@@ -402,7 +409,11 @@ func (c codexClient) runStream(ctx context.Context, resp http.Response, req Re
402	409		if model.ID == "" {
403	410		model, _ = FindModel("openai", req.Model)
404	411		}
405		-	out <- EventStart{Model: req.Model, Provider: "openai-codex"}
	412	+	providerName := c.providerName
	413	+	if providerName == "" {
	414	+	providerName = "openai-codex"
	415	+	}
	416	+	out <- EventStart{Model: req.Model, Provider: providerName}
406	417
407	418		raw := make(chan sseEvent, 16)
408	419		go readSSE(resp.Body, raw)

 const codexDefaultBaseURL = "https://chatgpt.com/backend-api/codex/responses"
 type codexClient struct {
-	token      string
+	token             string
-	accountID  string
+	accountID         string
-	baseURL    string
+	baseURL           string
-	errorLabel string
+	errorLabel        string
-	http       *http.Client
+	providerName      string
+	modelName         func(string) string
+	disableCLIRouting bool
+	http              *http.Client
 }
 // NewOpenAICodex creates a client that talks to ChatGPT's Codex endpoint
 		baseURL = codexDefaultBaseURL
 	}
 	return &codexClient{
-		token:      token,
+		token:        token,
-		accountID:  accountID,
+		accountID:    accountID,
-		baseURL:    strings.TrimRight(baseURL, "/"),
+		baseURL:      strings.TrimRight(baseURL, "/"),
-		errorLabel: "codex",
+		errorLabel:   "codex",
-		http:       &http.Client{Timeout: 0},
+		providerName: "openai-codex",
+		http:         &http.Client{Timeout: 0},
 	}
 }
 	if err != nil {
 		return nil, err
 	}
+	if c.modelName != nil {
+		wire.Model = c.modelName(wire.Model)
+	}
 	var codexCLISessionID string
-	if usesCodexCLIRouting(wire.Model) {
+	if !c.disableCLIRouting && usesCodexCLIRouting(wire.Model) {
 		codexCLISessionID = newCodexSessionID()
 		wire.PromptCacheKey = codexCLISessionID
 	}
 	if model.ID == "" {
 		model, _ = FindModel("openai", req.Model)
 	}
-	out <- EventStart{Model: req.Model, Provider: "openai-codex"}
+	providerName := c.providerName
+	if providerName == "" {
+		providerName = "openai-codex"
+	}
+	out <- EventStart{Model: req.Model, Provider: providerName}
 	raw := make(chan sseEvent, 16)
 	go readSSE(resp.Body, raw)

M packages/provider/openai_responses.go +6 -5


@@ -63,11 +63,12 @@ func NewOpenAIResponsesNamed(apiKey, baseURL, name string) Client {
63	63		Timeout: 0,
64	64		}
65	65		inner := &codexClient{
66		-	token: apiKey,
67		-	accountID: "", // unused; transport strips the header
68		-	baseURL: strings.TrimRight(baseURL, "/"),
69		-	errorLabel: "openai",
70		-	http: httpClient,
	66	+	token: apiKey,
	67	+	accountID: "", // unused; transport strips the header
	68	+	baseURL: strings.TrimRight(baseURL, "/"),
	69	+	errorLabel: "openai",
	70	+	providerName: name,
	71	+	http: httpClient,
71	72		}
72	73		return &renamedClient{inner: inner, name: name}
73	74		}


@@ -63,11 +63,12 @@ func NewOpenAIResponsesNamed(apiKey, baseURL, name string) Client {
63		Timeout: 0,	63		Timeout: 0,
64		}	64		}
65		inner := &codexClient{	65		inner := &codexClient{
66	-	token: apiKey,	66	+	token: apiKey,
67	-	accountID: "", // unused; transport strips the header	67	+	accountID: "", // unused; transport strips the header
68	-	baseURL: strings.TrimRight(baseURL, "/"),	68	+	baseURL: strings.TrimRight(baseURL, "/"),
69	-	errorLabel: "openai",	69	+	errorLabel: "openai",
70	-	http: httpClient,	70	+	providerName: name,
			71	+	http: httpClient,
71		}	72		}
72		return &renamedClient{inner: inner, name: name}	73		return &renamedClient{inner: inner, name: name}
73		}	74		}

M packages/provider/provider_test.go +6 -0


@@ -414,6 +414,12 @@ func TestGPT56CatalogEntries(t *testing.T) {
414	414		{"openai-codex", "gpt-5.6-luna", 372000, 1, 6, 0.1, 1.25},
415	415		{"openai-codex", "gpt-5.6-sol", 372000, 5, 30, 0.5, 6.25},
416	416		{"openai-codex", "gpt-5.6-terra", 372000, 2.5, 15, 0.25, 3.125},
	417	+	{"azure-openai-responses", "gpt-5.6-luna", 1050000, 1, 6, 0.1, 1.25},
	418	+	{"azure-openai-responses", "gpt-5.6-sol", 1050000, 5, 30, 0.5, 6.25},
	419	+	{"azure-openai-responses", "gpt-5.6-terra", 1050000, 2.5, 15, 0.25, 3.125},
	420	+	{"vercel-ai-gateway", "openai/gpt-5.6-luna", 1050000, 1, 6, 0.1, 1.25},
	421	+	{"vercel-ai-gateway", "openai/gpt-5.6-sol", 1050000, 5, 30, 0.5, 6.25},
	422	+	{"vercel-ai-gateway", "openai/gpt-5.6-terra", 1050000, 2.5, 15, 0.25, 3.125},
417	423		}
418	424		for _, tc := range cases {
419	425		m, err := FindModel(tc.provider, tc.id)


@@ -414,6 +414,12 @@ func TestGPT56CatalogEntries(t *testing.T) {
414	{"openai-codex", "gpt-5.6-luna", 372000, 1, 6, 0.1, 1.25},	414		{"openai-codex", "gpt-5.6-luna", 372000, 1, 6, 0.1, 1.25},
415	{"openai-codex", "gpt-5.6-sol", 372000, 5, 30, 0.5, 6.25},	415		{"openai-codex", "gpt-5.6-sol", 372000, 5, 30, 0.5, 6.25},
416	{"openai-codex", "gpt-5.6-terra", 372000, 2.5, 15, 0.25, 3.125},	416		{"openai-codex", "gpt-5.6-terra", 372000, 2.5, 15, 0.25, 3.125},
		417	+	{"azure-openai-responses", "gpt-5.6-luna", 1050000, 1, 6, 0.1, 1.25},
		418	+	{"azure-openai-responses", "gpt-5.6-sol", 1050000, 5, 30, 0.5, 6.25},
		419	+	{"azure-openai-responses", "gpt-5.6-terra", 1050000, 2.5, 15, 0.25, 3.125},
		420	+	{"vercel-ai-gateway", "openai/gpt-5.6-luna", 1050000, 1, 6, 0.1, 1.25},
		421	+	{"vercel-ai-gateway", "openai/gpt-5.6-sol", 1050000, 5, 30, 0.5, 6.25},
		422	+	{"vercel-ai-gateway", "openai/gpt-5.6-terra", 1050000, 2.5, 15, 0.25, 3.125},
417	}	423		}
418	for _, tc := range cases {	424		for _, tc := range cases {
419	m, err := FindModel(tc.provider, tc.id)	425		m, err := FindModel(tc.provider, tc.id)