From 153ee1502c18e4d01ece9192657ce2e4dcaa5a42 Mon Sep 17 00:00:00 2001
From: "Jakub A. W" <jakubwasek@gmail.com>
Date: Sat, 8 Aug 2026 20:19:31 +0200
Subject: [PATCH 1/7] feat(providers): add Chutes AI support

---
 config/config.example.yaml               |   5 +
 docs/advanced/configuration.mdx          |   1 +
 docs/providers/overview.mdx              |   5 +
 internal/providers/chutes/chutes.go      |  97 +++++++++++++
 internal/providers/chutes/chutes_test.go | 167 +++++++++++++++++++++++
 internal/providers/chutes/models.go      | 134 ++++++++++++++++++
 internal/providers/config_test.go        |  23 ++++
 run/providers.go                         |   2 +
 run/providers_test.go                    |   2 +-
 9 files changed, 435 insertions(+), 1 deletion(-)
 create mode 100644 internal/providers/chutes/chutes.go
 create mode 100644 internal/providers/chutes/chutes_test.go
 create mode 100644 internal/providers/chutes/models.go

diff --git a/config/config.example.yaml b/config/config.example.yaml
index dc687d5a6..d43412956 100644
--- a/config/config.example.yaml
+++ b/config/config.example.yaml
@@ -380,6 +380,11 @@ providers:
     # models:
     #   - id: "accounts/fireworks/models/gpt-oss-120b"
 
+  chutes:
+    type: chutes
+    api_key: "${CHUTES_API_KEY}"
+    # base_url defaults to "https://llm.chutes.ai/v1"
+
   meta:
     type: meta
     api_key: "..."
diff --git a/docs/advanced/configuration.mdx b/docs/advanced/configuration.mdx
index 339ccc57a..b5c35aa47 100644
--- a/docs/advanced/configuration.mdx
+++ b/docs/advanced/configuration.mdx
@@ -409,6 +409,7 @@ export GEMINI_API_KEY="..."          # Registers "gemini" provider
 export DEEPSEEK_API_KEY="..."        # Registers "deepseek" provider
 export XAI_API_KEY="..."             # Registers "xai" provider
 export GROQ_API_KEY="gsk_..."        # Registers "groq" provider
+export CHUTES_API_KEY="cpk_..."      # Registers "chutes" provider
 export OPENROUTER_API_KEY="sk-or-..." # Registers "openrouter" provider
 export KILO_API_KEY="..."            # Registers "kilo" provider
 export ZAI_API_KEY="..."             # Registers "zai" provider
diff --git a/docs/providers/overview.mdx b/docs/providers/overview.mdx
index 49e62a391..50e32aad6 100644
--- a/docs/providers/overview.mdx
+++ b/docs/providers/overview.mdx
@@ -48,6 +48,7 @@ support, not every individual model capability exposed by an upstream provider.
 | DeepSeek | `DEEPSEEK_API_KEY` | `deepseek-v4-pro` | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | [DeepSeek](/providers/deepseek) |
 | Groq | `GROQ_API_KEY` | `llama-3.3-70b-versatile` | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ | — |
 | Fireworks AI | `FIREWORKS_API_KEY` (`FIREWORKS_BASE_URL` optional) | `accounts/fireworks/models/gpt-oss-120b` | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | — |
+| Chutes AI | `CHUTES_API_KEY` (`CHUTES_BASE_URL` optional) | `Qwen/Qwen3-32B-TEE` | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | — |
 | Meta (Muse Spark) | `META_API_KEY` (`META_BASE_URL` optional) | `muse-spark-1.1` | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | — |
 | OpenRouter | `OPENROUTER_API_KEY` | `google/gemini-2.5-flash` | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | — |
 | Kilo AI | `KILO_API_KEY` (`KILO_BASE_URL` optional) | `anthropic/claude-sonnet-4.5` | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | — |
@@ -90,6 +91,10 @@ support, not every individual model capability exposed by an upstream provider.
 - **Fireworks AI** — model IDs are account-scoped paths such as
   `accounts/fireworks/models/gpt-oss-120b`; use them verbatim in requests and
   in `FIREWORKS_MODELS`.
+- **Chutes AI** — defaults to `https://llm.chutes.ai/v1` and discovers its
+  current model IDs, context limits, capabilities, and pricing from the live
+  catalog. GoModel translates `/v1/responses` requests to chat completions;
+  Chutes' shared LLM endpoint does not expose embeddings.
 - **Meta (Muse Spark)** — the Meta Model API is OpenAI-compatible; set
   `META_API_KEY` and route to `muse-spark-1.1`. Muse Spark models are not in
   the upstream model catalog yet, so declare `context_window` and `pricing`
diff --git a/internal/providers/chutes/chutes.go b/internal/providers/chutes/chutes.go
new file mode 100644
index 000000000..dc172c80b
--- /dev/null
+++ b/internal/providers/chutes/chutes.go
@@ -0,0 +1,97 @@
+// Package chutes provides Chutes AI integration for the LLM gateway.
+package chutes
+
+import (
+	"context"
+	"io"
+	"net/http"
+
+	"github.com/enterpilot/gomodel/internal/core"
+	"github.com/enterpilot/gomodel/internal/llmclient"
+	"github.com/enterpilot/gomodel/internal/providers"
+	"github.com/enterpilot/gomodel/internal/providers/openai"
+)
+
+const defaultBaseURL = "https://llm.chutes.ai/v1"
+
+// Registration provides factory registration for the Chutes AI provider.
+var Registration = providers.Registration{
+	Type: "chutes",
+	New:  New,
+	Discovery: providers.DiscoveryConfig{
+		DefaultBaseURL: defaultBaseURL,
+	},
+}
+
+// Provider implements Chutes' OpenAI-compatible chat surface. Chutes does not
+// expose native Responses or embeddings endpoints, so Responses requests are
+// translated through chat completions and embeddings fail locally.
+type Provider struct {
+	compat *openai.CompatibleProvider
+}
+
+var _ core.Provider = (*Provider)(nil)
+var _ core.PassthroughProvider = (*Provider)(nil)
+
+// New creates a new Chutes AI provider.
+func New(cfg providers.ProviderConfig, opts providers.ProviderOptions) core.Provider {
+	return &Provider{compat: openai.NewCompatibleProvider(cfg.APIKey, opts, compatibleConfig(
+		providers.ResolveBaseURL(cfg.BaseURL, defaultBaseURL),
+	))}
+}
+
+// NewWithHTTPClient creates a new Chutes AI provider with a custom HTTP client.
+// If httpClient is nil, http.DefaultClient is used.
+func NewWithHTTPClient(apiKey string, baseURL string, httpClient *http.Client, hooks llmclient.Hooks) *Provider {
+	return &Provider{compat: openai.NewCompatibleProviderWithHTTPClient(apiKey, httpClient, hooks, compatibleConfig(
+		providers.ResolveBaseURL(baseURL, defaultBaseURL),
+	))}
+}
+
+func compatibleConfig(baseURL string) openai.CompatibleProviderConfig {
+	return openai.CompatibleProviderConfig{
+		ProviderName: "chutes",
+		BaseURL:      baseURL,
+		SetHeaders:   setHeaders,
+	}
+}
+
+func setHeaders(req *http.Request, apiKey string) {
+	providers.SetAuthHeaders(req, apiKey, providers.AuthHeaderConfig{AuthScheme: "Bearer "})
+}
+
+// SetBaseURL changes the Chutes API base URL.
+func (p *Provider) SetBaseURL(baseURL string) {
+	p.compat.SetBaseURL(baseURL)
+}
+
+// ChatCompletion sends a chat completion request to Chutes.
+func (p *Provider) ChatCompletion(ctx context.Context, req *core.ChatRequest) (*core.ChatResponse, error) {
+	return p.compat.ChatCompletion(ctx, req)
+}
+
+// StreamChatCompletion sends a streaming chat completion request to Chutes.
+func (p *Provider) StreamChatCompletion(ctx context.Context, req *core.ChatRequest) (io.ReadCloser, error) {
+	return p.compat.StreamChatCompletion(ctx, req)
+}
+
+// Responses translates an OpenAI Responses request through Chutes chat completions.
+func (p *Provider) Responses(ctx context.Context, req *core.ResponsesRequest) (*core.ResponsesResponse, error) {
+	return providers.ResponsesViaChat(ctx, p, req)
+}
+
+// StreamResponses translates a streaming Responses request through Chutes chat completions.
+func (p *Provider) StreamResponses(ctx context.Context, req *core.ResponsesRequest) (io.ReadCloser, error) {
+	return providers.StreamResponsesViaChat(ctx, p, req, "chutes")
+}
+
+// Embeddings returns an error because the shared Chutes LLM endpoint does not
+// expose an OpenAI-compatible embeddings route.
+func (p *Provider) Embeddings(_ context.Context, _ *core.EmbeddingRequest) (*core.EmbeddingResponse, error) {
+	return nil, core.NewInvalidRequestError("chutes does not support embeddings", nil)
+}
+
+// Passthrough forwards an opaque request to Chutes.
+func (p *Provider) Passthrough(ctx context.Context, req *core.PassthroughRequest) (*core.PassthroughResponse, error) {
+	return p.compat.Passthrough(ctx, req)
+}
diff --git a/internal/providers/chutes/chutes_test.go b/internal/providers/chutes/chutes_test.go
new file mode 100644
index 000000000..63376f4ac
--- /dev/null
+++ b/internal/providers/chutes/chutes_test.go
@@ -0,0 +1,167 @@
+package chutes
+
+import (
+	"context"
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
+	"testing"
+
+	"github.com/enterpilot/gomodel/internal/core"
+	"github.com/enterpilot/gomodel/internal/llmclient"
+)
+
+func TestChatCompletion_UsesBearerAuthAndChatEndpoint(t *testing.T) {
+	var gotPath, gotAuth string
+
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		gotPath = r.URL.Path
+		gotAuth = r.Header.Get("Authorization")
+		w.Header().Set("Content-Type", "application/json")
+		_, _ = w.Write([]byte(`{
+			"id":"chatcmpl-chutes",
+			"created":1677652288,
+			"model":"Qwen/Qwen3-32B-TEE",
+			"choices":[{"index":0,"message":{"role":"assistant","content":"hello"},"finish_reason":"stop"}],
+			"usage":{"prompt_tokens":3,"completion_tokens":1,"total_tokens":4}
+		}`))
+	}))
+	defer server.Close()
+
+	provider := NewWithHTTPClient("cpk_test", server.URL, server.Client(), llmclient.Hooks{})
+	resp, err := provider.ChatCompletion(context.Background(), &core.ChatRequest{
+		Model:    "Qwen/Qwen3-32B-TEE",
+		Messages: []core.Message{{Role: "user", Content: "hi"}},
+	})
+	if err != nil {
+		t.Fatalf("ChatCompletion() error = %v", err)
+	}
+	if gotPath != "/chat/completions" {
+		t.Fatalf("path = %q, want /chat/completions", gotPath)
+	}
+	if gotAuth != "Bearer cpk_test" {
+		t.Fatalf("authorization = %q, want Bearer cpk_test", gotAuth)
+	}
+	if resp.Model != "Qwen/Qwen3-32B-TEE" || resp.Usage.TotalTokens != 4 {
+		t.Fatalf("response = %+v, want model and usage preserved", resp)
+	}
+}
+
+func TestResponses_TranslatesToChatCompletions(t *testing.T) {
+	var gotPath string
+	var gotBody map[string]any
+
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		gotPath = r.URL.Path
+		if err := json.NewDecoder(r.Body).Decode(&gotBody); err != nil {
+			http.Error(w, "decode error", http.StatusBadRequest)
+			return
+		}
+		w.Header().Set("Content-Type", "application/json")
+		_, _ = w.Write([]byte(`{
+			"id":"chatcmpl-chutes",
+			"created":1677652288,
+			"model":"Qwen/Qwen3-32B-TEE",
+			"choices":[{"index":0,"message":{"role":"assistant","content":"translated"},"finish_reason":"stop"}],
+			"usage":{"prompt_tokens":3,"completion_tokens":2,"total_tokens":5}
+		}`))
+	}))
+	defer server.Close()
+
+	provider := NewWithHTTPClient("cpk_test", server.URL, server.Client(), llmclient.Hooks{})
+	resp, err := provider.Responses(context.Background(), &core.ResponsesRequest{
+		Model: "Qwen/Qwen3-32B-TEE",
+		Input: "hi",
+	})
+	if err != nil {
+		t.Fatalf("Responses() error = %v", err)
+	}
+	if gotPath != "/chat/completions" {
+		t.Fatalf("path = %q, want /chat/completions", gotPath)
+	}
+	if gotBody["model"] != "Qwen/Qwen3-32B-TEE" {
+		t.Fatalf("request model = %#v, want Qwen/Qwen3-32B-TEE", gotBody["model"])
+	}
+	if resp.Object != "response" || resp.Status != "completed" {
+		t.Fatalf("response metadata = object %q status %q, want response/completed", resp.Object, resp.Status)
+	}
+}
+
+func TestListModels_PreservesChutesMetadata(t *testing.T) {
+	var gotPath, gotAuth string
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		gotPath = r.URL.Path
+		gotAuth = r.Header.Get("Authorization")
+		w.Header().Set("Content-Type", "application/json")
+		_, _ = w.Write([]byte(`{
+			"object":"list",
+			"data":[{
+				"id":"Qwen/Qwen3.5-397B-A17B-TEE",
+				"owned_by":"sglang",
+				"created":1677652288,
+				"context_length":262144,
+				"max_output_length":65536,
+				"input_modalities":["text","image"],
+				"supported_features":["json_mode","tools","structured_outputs","reasoning"],
+				"confidential_compute":true,
+				"pricing":{"prompt":0.45,"completion":3.0,"input_cache_read":0.045}
+			}]
+		}`))
+	}))
+	defer server.Close()
+
+	provider := NewWithHTTPClient("cpk_test", server.URL, server.Client(), llmclient.Hooks{})
+	resp, err := provider.ListModels(context.Background())
+	if err != nil {
+		t.Fatalf("ListModels() error = %v", err)
+	}
+	if gotPath != "/models" || gotAuth != "Bearer cpk_test" {
+		t.Fatalf("request path/auth = %q/%q, want /models/Bearer cpk_test", gotPath, gotAuth)
+	}
+	if len(resp.Data) != 1 {
+		t.Fatalf("len(resp.Data) = %d, want 1", len(resp.Data))
+	}
+	model := resp.Data[0]
+	if model.Object != "model" {
+		t.Fatalf("model.Object = %q, want model", model.Object)
+	}
+	if model.Metadata == nil || model.Metadata.ContextWindow == nil || *model.Metadata.ContextWindow != 262144 {
+		t.Fatalf("model context metadata = %+v, want 262144", model.Metadata)
+	}
+	if model.Metadata.MaxOutputTokens == nil || *model.Metadata.MaxOutputTokens != 65536 {
+		t.Fatalf("max output tokens = %+v, want 65536", model.Metadata.MaxOutputTokens)
+	}
+	if !model.Metadata.Capabilities["tools"] || !model.Metadata.Capabilities["vision"] || !model.Metadata.Capabilities["confidential_compute"] {
+		t.Fatalf("capabilities = %v, want tools, vision, and confidential_compute", model.Metadata.Capabilities)
+	}
+	pricing := model.Metadata.Pricing
+	if pricing == nil || pricing.Currency != "USD" || pricing.InputPerMtok == nil || *pricing.InputPerMtok != 0.45 ||
+		pricing.OutputPerMtok == nil || *pricing.OutputPerMtok != 3.0 ||
+		pricing.CachedInputPerMtok == nil || *pricing.CachedInputPerMtok != 0.045 {
+		t.Fatalf("pricing = %+v, want Chutes per-MTok USD pricing", pricing)
+	}
+}
+
+func TestEmbeddings_ReturnsUnsupportedError(t *testing.T) {
+	provider := NewWithHTTPClient("cpk_test", "", nil, llmclient.Hooks{})
+	if _, err := provider.Embeddings(context.Background(), &core.EmbeddingRequest{}); err == nil {
+		t.Fatal("Embeddings() error = nil, want unsupported error")
+	}
+}
+
+func TestProvider_DoesNotExposeUnsupportedOptionalInterfaces(t *testing.T) {
+	provider := NewWithHTTPClient("cpk_test", "", nil, llmclient.Hooks{})
+
+	if _, ok := any(provider).(core.NativeBatchProvider); ok {
+		t.Fatal("chutes provider should not implement native batch provider")
+	}
+	if _, ok := any(provider).(core.NativeFileProvider); ok {
+		t.Fatal("chutes provider should not implement native file provider")
+	}
+	if _, ok := any(provider).(core.NativeResponseLifecycleProvider); ok {
+		t.Fatal("chutes provider should not implement native response lifecycle provider")
+	}
+	if _, ok := any(provider).(core.AudioProvider); ok {
+		t.Fatal("chutes provider should not implement audio provider")
+	}
+}
diff --git a/internal/providers/chutes/models.go b/internal/providers/chutes/models.go
new file mode 100644
index 000000000..2904e746e
--- /dev/null
+++ b/internal/providers/chutes/models.go
@@ -0,0 +1,134 @@
+package chutes
+
+import (
+	"context"
+	"net/http"
+	"strings"
+
+	"github.com/enterpilot/gomodel/internal/core"
+	"github.com/enterpilot/gomodel/internal/llmclient"
+)
+
+type modelsResponse struct {
+	Object string      `json:"object"`
+	Data   []modelInfo `json:"data"`
+}
+
+type modelInfo struct {
+	ID                  string        `json:"id"`
+	Object              string        `json:"object"`
+	OwnedBy             string        `json:"owned_by"`
+	Created             int64         `json:"created"`
+	ContextLength       int           `json:"context_length"`
+	MaxOutputLength     int           `json:"max_output_length"`
+	InputModalities     []string      `json:"input_modalities"`
+	SupportedFeatures   []string      `json:"supported_features"`
+	ConfidentialCompute bool          `json:"confidential_compute"`
+	Pricing             *modelPricing `json:"pricing"`
+}
+
+type modelPricing struct {
+	Prompt         *float64 `json:"prompt"`
+	Completion     *float64 `json:"completion"`
+	InputCacheRead *float64 `json:"input_cache_read"`
+}
+
+// ListModels returns Chutes' live model catalog while retaining the context,
+// feature, and per-million-token pricing data included in its response.
+func (p *Provider) ListModels(ctx context.Context) (*core.ModelsResponse, error) {
+	var upstream modelsResponse
+	if err := p.compat.Do(ctx, llmclient.Request{
+		Method:   http.MethodGet,
+		Endpoint: "/models",
+	}, &upstream); err != nil {
+		return nil, err
+	}
+
+	result := &core.ModelsResponse{Object: strings.TrimSpace(upstream.Object)}
+	if result.Object == "" {
+		result.Object = "list"
+	}
+	result.Data = make([]core.Model, 0, len(upstream.Data))
+	for _, model := range upstream.Data {
+		if strings.TrimSpace(model.ID) == "" {
+			continue
+		}
+		result.Data = append(result.Data, model.toCore())
+	}
+	return result, nil
+}
+
+func (m modelInfo) toCore() core.Model {
+	object := strings.TrimSpace(m.Object)
+	if object == "" {
+		object = "model"
+	}
+
+	modes := []string{"chat", "responses"}
+	metadata := &core.ModelMetadata{
+		Modes:        modes,
+		Categories:   core.CategoriesForModes(modes),
+		Capabilities: modelCapabilities(m),
+		Pricing:      m.Pricing.toCore(),
+	}
+	if m.ContextLength > 0 {
+		metadata.ContextWindow = new(m.ContextLength)
+	}
+	if m.MaxOutputLength > 0 {
+		metadata.MaxOutputTokens = new(m.MaxOutputLength)
+	}
+
+	return core.Model{
+		ID:       strings.TrimSpace(m.ID),
+		Object:   object,
+		OwnedBy:  strings.TrimSpace(m.OwnedBy),
+		Created:  m.Created,
+		Metadata: metadata,
+	}
+}
+
+func modelCapabilities(m modelInfo) map[string]bool {
+	capabilities := make(map[string]bool, len(m.SupportedFeatures)+3)
+	for _, feature := range m.SupportedFeatures {
+		if normalized := strings.ToLower(strings.TrimSpace(feature)); normalized != "" {
+			capabilities[normalized] = true
+		}
+	}
+	for _, modality := range m.InputModalities {
+		switch strings.ToLower(strings.TrimSpace(modality)) {
+		case "image":
+			capabilities["vision"] = true
+		case "video":
+			capabilities["video"] = true
+		case "audio":
+			capabilities["audio"] = true
+		}
+	}
+	if m.ConfidentialCompute {
+		capabilities["confidential_compute"] = true
+	}
+	if len(capabilities) == 0 {
+		return nil
+	}
+	return capabilities
+}
+
+func (p *modelPricing) toCore() *core.ModelPricing {
+	if p == nil || (p.Prompt == nil && p.Completion == nil && p.InputCacheRead == nil) {
+		return nil
+	}
+	return &core.ModelPricing{
+		Currency:           "USD",
+		InputPerMtok:       cloneFloat(p.Prompt),
+		OutputPerMtok:      cloneFloat(p.Completion),
+		CachedInputPerMtok: cloneFloat(p.InputCacheRead),
+	}
+}
+
+func cloneFloat(value *float64) *float64 {
+	if value == nil {
+		return nil
+	}
+	cloned := *value
+	return &cloned
+}
diff --git a/internal/providers/config_test.go b/internal/providers/config_test.go
index 9f4cea48d..63db6714d 100644
--- a/internal/providers/config_test.go
+++ b/internal/providers/config_test.go
@@ -35,6 +35,9 @@ var testDiscoveryConfigs = map[string]DiscoveryConfig{
 	"deepseek": {
 		DefaultBaseURL: "https://api.deepseek.com",
 	},
+	"chutes": {
+		DefaultBaseURL: "https://llm.chutes.ai/v1",
+	},
 	"xai": {
 		DefaultBaseURL: "https://api.x.ai/v1",
 	},
@@ -744,6 +747,26 @@ func TestApplyProviderEnvVars_DiscoversDeepSeekFromAPIKey(t *testing.T) {
 	}
 }
 
+func TestApplyProviderEnvVars_DiscoversChutesFromAPIKey(t *testing.T) {
+	t.Setenv("CHUTES_API_KEY", "cpk_test")
+
+	got := applyProviderEnvVars(map[string]config.RawProviderConfig{}, testDiscoveryConfigs)
+
+	p, exists := got["chutes"]
+	if !exists {
+		t.Fatal("expected chutes to be discovered from env var")
+	}
+	if p.APIKey != "cpk_test" {
+		t.Errorf("APIKey = %q, want cpk_test", p.APIKey)
+	}
+	if p.Type != "chutes" {
+		t.Errorf("Type = %q, want chutes", p.Type)
+	}
+	if p.BaseURL != testDiscoveryConfigs["chutes"].DefaultBaseURL {
+		t.Errorf("BaseURL = %q, want %q", p.BaseURL, testDiscoveryConfigs["chutes"].DefaultBaseURL)
+	}
+}
+
 func TestApplyProviderEnvVars_DiscoversZAIFromAPIKey(t *testing.T) {
 	t.Setenv("ZAI_API_KEY", "zai-key")
 
diff --git a/run/providers.go b/run/providers.go
index b0331baf3..d76388477 100644
--- a/run/providers.go
+++ b/run/providers.go
@@ -9,6 +9,7 @@ import (
 	"github.com/enterpilot/gomodel/internal/providers/bailian"
 	"github.com/enterpilot/gomodel/internal/providers/bedrock"
 	"github.com/enterpilot/gomodel/internal/providers/bedrockmantle"
+	"github.com/enterpilot/gomodel/internal/providers/chutes"
 	"github.com/enterpilot/gomodel/internal/providers/cohere"
 	"github.com/enterpilot/gomodel/internal/providers/deepseek"
 	"github.com/enterpilot/gomodel/internal/providers/fireworks"
@@ -49,6 +50,7 @@ func defaultProviderFactory(cfg *config.Config) *providers.ProviderFactory {
 	factory.Add(anthropic.Registration)
 	factory.Add(bedrock.Registration)
 	factory.Add(bedrockmantle.Registration)
+	factory.Add(chutes.Registration)
 	factory.Add(cohere.Registration)
 	factory.Add(deepseek.Registration)
 	factory.Add(fireworks.Registration)
diff --git a/run/providers_test.go b/run/providers_test.go
index 4aa00429b..285038e83 100644
--- a/run/providers_test.go
+++ b/run/providers_test.go
@@ -155,7 +155,7 @@ var credentialPayloadFields = []string{
 
 func TestDefaultProviderFactoryRegistersAllProviderTypes(t *testing.T) {
 	expected := []string{
-		"anthropic", "azure", "bailian", "bedrock", "bedrock-mantle", "cohere", "deepseek", "fireworks",
+		"anthropic", "azure", "bailian", "bedrock", "bedrock-mantle", "chutes", "cohere", "deepseek", "fireworks",
 		"gemini", "groq", "kilo", "kimicode", "llmd", "meta", "minimax", "ollama", "openai", "opencode_go",
 		"openrouter", "oracle", "sglang", "vertex", "vllm", "xai", "xiaomi", "zai",
 	}

From 829102ac308f2fdb9ff070c7a75c4e3a462bb493 Mon Sep 17 00:00:00 2001
From: "Jakub A. W" <jakubwasek@gmail.com>
Date: Sat, 8 Aug 2026 20:33:24 +0200
Subject: [PATCH 2/7] fix(chutes): address provider review feedback

---
 config/config.example.yaml               |   3 +-
 docs/advanced/configuration.mdx          |   3 +-
 internal/providers/chutes/chutes.go      |   2 +
 internal/providers/chutes/chutes_test.go | 118 ++++++++++++++++++++++-
 internal/providers/chutes/models.go      |   9 +-
 5 files changed, 128 insertions(+), 7 deletions(-)

diff --git a/config/config.example.yaml b/config/config.example.yaml
index d43412956..8142a5d38 100644
--- a/config/config.example.yaml
+++ b/config/config.example.yaml
@@ -383,7 +383,8 @@ providers:
   chutes:
     type: chutes
     api_key: "${CHUTES_API_KEY}"
-    # base_url defaults to "https://llm.chutes.ai/v1"
+    # base_url defaults to "https://llm.chutes.ai/v1".
+    # Set base_url when using a different compatible endpoint.
 
   meta:
     type: meta
diff --git a/docs/advanced/configuration.mdx b/docs/advanced/configuration.mdx
index b5c35aa47..e148ef840 100644
--- a/docs/advanced/configuration.mdx
+++ b/docs/advanced/configuration.mdx
@@ -276,6 +276,7 @@ Set these to automatically register providers. No YAML configuration required.
 | `ZAI_API_KEY`        | Z.ai                                               |
 | `XAI_API_KEY`        | xAI (Grok)                                         |
 | `GROQ_API_KEY`       | Groq                                               |
+| `CHUTES_API_KEY`     | Chutes AI (`CHUTES_BASE_URL` optional)             |
 | `AZURE_API_KEY`      | Azure OpenAI (`AZURE_BASE_URL` also required)     |
 | `ORACLE_API_KEY`     | Oracle GenAI (`ORACLE_BASE_URL` also required)    |
 | `OLLAMA_BASE_URL`    | Ollama (no API key needed)                        |
@@ -283,7 +284,7 @@ Set these to automatically register providers. No YAML configuration required.
 | `VLLM_BASE_URL`      | vLLM (no API key needed unless upstream requires) |
 | `LLMD_BASE_URL`      | llm-d Router/EPP (no API key needed unless its Gateway requires one) |
 
-Most providers can use a custom base URL via `<PROVIDER>_BASE_URL` (for example `OPENAI_BASE_URL`). DeepSeek defaults to `https://api.deepseek.com`; set `DEEPSEEK_BASE_URL` only for a compatible proxy or alternate DeepSeek endpoint. OpenRouter defaults to `https://openrouter.ai/api/v1` and can be overridden with `OPENROUTER_BASE_URL`. Kilo AI defaults to `https://api.kilo.ai/api/gateway` and can be overridden with `KILO_BASE_URL`. Z.ai defaults to `https://api.z.ai/api/paas/v4`; set `ZAI_BASE_URL=https://api.z.ai/api/coding/paas/v4` for the GLM Coding Plan endpoint. SGLang defaults to `http://localhost:30000/v1` when `SGLANG_API_KEY` is set, but keyless deployments should set `SGLANG_BASE_URL` explicitly to register the provider. vLLM follows the same pattern at `http://localhost:8000/v1`. llm-d has no universal endpoint, so `LLMD_BASE_URL` is always required; `LLMD_API_KEY` is optional. Azure uses `AZURE_BASE_URL` for its deployment base URL and accepts an optional `AZURE_API_VERSION` override; otherwise it defaults to `2024-10-21`. Oracle requires `ORACLE_BASE_URL` because its OpenAI-compatible endpoint is region-specific.
+Most providers can use a custom base URL via `<PROVIDER>_BASE_URL` (for example `OPENAI_BASE_URL`). Chutes AI defaults to `https://llm.chutes.ai/v1` and can be overridden with `CHUTES_BASE_URL`. DeepSeek defaults to `https://api.deepseek.com`; set `DEEPSEEK_BASE_URL` only for a compatible proxy or alternate DeepSeek endpoint. OpenRouter defaults to `https://openrouter.ai/api/v1` and can be overridden with `OPENROUTER_BASE_URL`. Kilo AI defaults to `https://api.kilo.ai/api/gateway` and can be overridden with `KILO_BASE_URL`. Z.ai defaults to `https://api.z.ai/api/paas/v4`; set `ZAI_BASE_URL=https://api.z.ai/api/coding/paas/v4` for the GLM Coding Plan endpoint. SGLang defaults to `http://localhost:30000/v1` when `SGLANG_API_KEY` is set, but keyless deployments should set `SGLANG_BASE_URL` explicitly to register the provider. vLLM follows the same pattern at `http://localhost:8000/v1`. llm-d has no universal endpoint, so `LLMD_BASE_URL` is always required; `LLMD_API_KEY` is optional. Azure uses `AZURE_BASE_URL` for its deployment base URL and accepts an optional `AZURE_API_VERSION` override; otherwise it defaults to `2024-10-21`. Oracle requires `ORACLE_BASE_URL` because its OpenAI-compatible endpoint is region-specific.
 
 Every provider type also accepts a comma-separated configured model list via
 `<PROVIDER>_MODELS`, for example `OPENROUTER_MODELS`, `ORACLE_MODELS`,
diff --git a/internal/providers/chutes/chutes.go b/internal/providers/chutes/chutes.go
index dc172c80b..60701d0c6 100644
--- a/internal/providers/chutes/chutes.go
+++ b/internal/providers/chutes/chutes.go
@@ -48,6 +48,7 @@ func NewWithHTTPClient(apiKey string, baseURL string, httpClient *http.Client, h
 	))}
 }
 
+// compatibleConfig returns the shared OpenAI-compatible transport settings for Chutes.
 func compatibleConfig(baseURL string) openai.CompatibleProviderConfig {
 	return openai.CompatibleProviderConfig{
 		ProviderName: "chutes",
@@ -56,6 +57,7 @@ func compatibleConfig(baseURL string) openai.CompatibleProviderConfig {
 	}
 }
 
+// setHeaders applies Chutes' bearer-token authentication.
 func setHeaders(req *http.Request, apiKey string) {
 	providers.SetAuthHeaders(req, apiKey, providers.AuthHeaderConfig{AuthScheme: "Bearer "})
 }
diff --git a/internal/providers/chutes/chutes_test.go b/internal/providers/chutes/chutes_test.go
index 63376f4ac..3654a4fb0 100644
--- a/internal/providers/chutes/chutes_test.go
+++ b/internal/providers/chutes/chutes_test.go
@@ -3,8 +3,10 @@ package chutes
 import (
 	"context"
 	"encoding/json"
+	"io"
 	"net/http"
 	"net/http/httptest"
+	"strings"
 	"testing"
 
 	"github.com/enterpilot/gomodel/internal/core"
@@ -47,6 +49,117 @@ func TestChatCompletion_UsesBearerAuthAndChatEndpoint(t *testing.T) {
 	}
 }
 
+func TestStreamChatCompletion_UsesBearerAuthAndChatEndpoint(t *testing.T) {
+	var gotPath, gotAuth string
+	var gotBody map[string]any
+
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		gotPath = r.URL.Path
+		gotAuth = r.Header.Get("Authorization")
+		if err := json.NewDecoder(r.Body).Decode(&gotBody); err != nil {
+			http.Error(w, "decode error", http.StatusBadRequest)
+			return
+		}
+		w.Header().Set("Content-Type", "text/event-stream")
+		_, _ = io.WriteString(w, "data: {\"id\":\"chatcmpl-chutes\",\"object\":\"chat.completion.chunk\",\"choices\":[]}\n\ndata: [DONE]\n\n")
+	}))
+	defer server.Close()
+
+	provider := NewWithHTTPClient("cpk_test", server.URL, server.Client(), llmclient.Hooks{})
+	stream, err := provider.StreamChatCompletion(context.Background(), &core.ChatRequest{
+		Model:    "Qwen/Qwen3-32B-TEE",
+		Messages: []core.Message{{Role: "user", Content: "hi"}},
+	})
+	if err != nil {
+		t.Fatalf("StreamChatCompletion() error = %v", err)
+	}
+	defer stream.Close()
+
+	body, err := io.ReadAll(stream)
+	if err != nil {
+		t.Fatalf("ReadAll() error = %v", err)
+	}
+	if gotPath != "/chat/completions" || gotAuth != "Bearer cpk_test" {
+		t.Fatalf("request path/auth = %q/%q, want /chat/completions/Bearer cpk_test", gotPath, gotAuth)
+	}
+	if gotBody["model"] != "Qwen/Qwen3-32B-TEE" || gotBody["stream"] != true {
+		t.Fatalf("stream request body = %#v", gotBody)
+	}
+	if !strings.Contains(string(body), "data: [DONE]") {
+		t.Fatalf("stream body = %q, want SSE terminator", body)
+	}
+}
+
+func TestChatCompletion_ReturnsUpstreamError(t *testing.T) {
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
+		w.Header().Set("Content-Type", "application/json")
+		w.WriteHeader(http.StatusTooManyRequests)
+		_, _ = w.Write([]byte(`{"error":{"message":"rate limited","type":"rate_limit_error"}}`))
+	}))
+	defer server.Close()
+
+	provider := NewWithHTTPClient("cpk_test", server.URL, server.Client(), llmclient.Hooks{})
+	_, err := provider.ChatCompletion(context.Background(), &core.ChatRequest{
+		Model:    "Qwen/Qwen3-32B-TEE",
+		Messages: []core.Message{{Role: "user", Content: "hi"}},
+	})
+	if err == nil {
+		t.Fatal("ChatCompletion() error = nil, want upstream error")
+	}
+	gatewayErr, ok := err.(*core.GatewayError)
+	if !ok {
+		t.Fatalf("error type = %T, want *core.GatewayError", err)
+	}
+	if gatewayErr.StatusCode != http.StatusTooManyRequests || gatewayErr.Type != core.ErrorTypeRateLimit {
+		t.Fatalf("gateway error = %+v, want 429 rate_limit_error", gatewayErr)
+	}
+}
+
+func TestPassthrough_ForwardsOpaqueRequest(t *testing.T) {
+	var gotURI, gotAuth, gotBeta, gotBody string
+
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		gotURI = r.URL.RequestURI()
+		gotAuth = r.Header.Get("Authorization")
+		gotBeta = r.Header.Get("X-Chutes-Beta")
+		body, _ := io.ReadAll(r.Body)
+		gotBody = string(body)
+		w.Header().Set("Content-Type", "application/json")
+		w.WriteHeader(http.StatusAccepted)
+		_, _ = w.Write([]byte(`{"accepted":true}`))
+	}))
+	defer server.Close()
+
+	provider := NewWithHTTPClient("cpk_test", server.URL, server.Client(), llmclient.Hooks{})
+	resp, err := provider.Passthrough(context.Background(), &core.PassthroughRequest{
+		Method:   http.MethodPost,
+		Endpoint: "chat/completions?trace=true",
+		Body:     io.NopCloser(strings.NewReader(`{"model":"Qwen/Qwen3-32B-TEE"}`)),
+		Headers: http.Header{
+			"Content-Type":  {"application/json"},
+			"X-Chutes-Beta": {"test"},
+		},
+	})
+	if err != nil {
+		t.Fatalf("Passthrough() error = %v", err)
+	}
+	defer resp.Body.Close()
+
+	responseBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		t.Fatalf("ReadAll() error = %v", err)
+	}
+	if gotURI != "/chat/completions?trace=true" || gotAuth != "Bearer cpk_test" {
+		t.Fatalf("request URI/auth = %q/%q", gotURI, gotAuth)
+	}
+	if gotBeta != "test" || gotBody != `{"model":"Qwen/Qwen3-32B-TEE"}` {
+		t.Fatalf("request beta/body = %q/%q", gotBeta, gotBody)
+	}
+	if resp.StatusCode != http.StatusAccepted || string(responseBody) != `{"accepted":true}` {
+		t.Fatalf("response status/body = %d/%q", resp.StatusCode, responseBody)
+	}
+}
+
 func TestResponses_TranslatesToChatCompletions(t *testing.T) {
 	var gotPath string
 	var gotBody map[string]any
@@ -94,7 +207,7 @@ func TestListModels_PreservesChutesMetadata(t *testing.T) {
 		gotAuth = r.Header.Get("Authorization")
 		w.Header().Set("Content-Type", "application/json")
 		_, _ = w.Write([]byte(`{
-			"object":"list",
+			"object":"chutes-model-catalog",
 			"data":[{
 				"id":"Qwen/Qwen3.5-397B-A17B-TEE",
 				"owned_by":"sglang",
@@ -121,6 +234,9 @@ func TestListModels_PreservesChutesMetadata(t *testing.T) {
 	if len(resp.Data) != 1 {
 		t.Fatalf("len(resp.Data) = %d, want 1", len(resp.Data))
 	}
+	if resp.Object != "list" {
+		t.Fatalf("resp.Object = %q, want list", resp.Object)
+	}
 	model := resp.Data[0]
 	if model.Object != "model" {
 		t.Fatalf("model.Object = %q, want model", model.Object)
diff --git a/internal/providers/chutes/models.go b/internal/providers/chutes/models.go
index 2904e746e..aec4684e0 100644
--- a/internal/providers/chutes/models.go
+++ b/internal/providers/chutes/models.go
@@ -44,10 +44,7 @@ func (p *Provider) ListModels(ctx context.Context) (*core.ModelsResponse, error)
 		return nil, err
 	}
 
-	result := &core.ModelsResponse{Object: strings.TrimSpace(upstream.Object)}
-	if result.Object == "" {
-		result.Object = "list"
-	}
+	result := &core.ModelsResponse{Object: "list"}
 	result.Data = make([]core.Model, 0, len(upstream.Data))
 	for _, model := range upstream.Data {
 		if strings.TrimSpace(model.ID) == "" {
@@ -58,6 +55,7 @@ func (p *Provider) ListModels(ctx context.Context) (*core.ModelsResponse, error)
 	return result, nil
 }
 
+// toCore normalizes a Chutes catalog entry into GoModel's provider-neutral model shape.
 func (m modelInfo) toCore() core.Model {
 	object := strings.TrimSpace(m.Object)
 	if object == "" {
@@ -87,6 +85,7 @@ func (m modelInfo) toCore() core.Model {
 	}
 }
 
+// modelCapabilities maps Chutes features and input modalities to GoModel capabilities.
 func modelCapabilities(m modelInfo) map[string]bool {
 	capabilities := make(map[string]bool, len(m.SupportedFeatures)+3)
 	for _, feature := range m.SupportedFeatures {
@@ -113,6 +112,7 @@ func modelCapabilities(m modelInfo) map[string]bool {
 	return capabilities
 }
 
+// toCore converts Chutes' per-million-token prices to GoModel pricing metadata.
 func (p *modelPricing) toCore() *core.ModelPricing {
 	if p == nil || (p.Prompt == nil && p.Completion == nil && p.InputCacheRead == nil) {
 		return nil
@@ -125,6 +125,7 @@ func (p *modelPricing) toCore() *core.ModelPricing {
 	}
 }
 
+// cloneFloat copies optional prices so the normalized model does not retain upstream pointers.
 func cloneFloat(value *float64) *float64 {
 	if value == nil {
 		return nil

From a7c34ff246f6e5543026c225fbc01452b2fcbf91 Mon Sep 17 00:00:00 2001
From: "Jakub A. W" <jakubwasek@gmail.com>
Date: Sat, 8 Aug 2026 21:00:46 +0200
Subject: [PATCH 3/7] test(chutes): cover provider integration branches

---
 internal/providers/chutes/chutes_test.go | 100 +++++++++-------
 internal/providers/chutes/models_test.go | 143 +++++++++++++++++++++++
 2 files changed, 201 insertions(+), 42 deletions(-)
 create mode 100644 internal/providers/chutes/models_test.go

diff --git a/internal/providers/chutes/chutes_test.go b/internal/providers/chutes/chutes_test.go
index 3654a4fb0..533fc8df9 100644
--- a/internal/providers/chutes/chutes_test.go
+++ b/internal/providers/chutes/chutes_test.go
@@ -11,8 +11,41 @@ import (
 
 	"github.com/enterpilot/gomodel/internal/core"
 	"github.com/enterpilot/gomodel/internal/llmclient"
+	"github.com/enterpilot/gomodel/internal/providers"
 )
 
+func TestNew_ConstructsRegisteredProvider(t *testing.T) {
+	provider, ok := New(providers.ProviderConfig{
+		APIKey:  "cpk_test",
+		BaseURL: "https://chutes.example/v1",
+	}, providers.ProviderOptions{}).(*Provider)
+	if !ok || provider.compat == nil {
+		t.Fatalf("New() = %T, want initialized *Provider", provider)
+	}
+	if Registration.Discovery.DefaultBaseURL != defaultBaseURL {
+		t.Fatalf("registration base URL = %q, want %q", Registration.Discovery.DefaultBaseURL, defaultBaseURL)
+	}
+}
+
+func TestSetBaseURL_ChangesRequestTarget(t *testing.T) {
+	var gotPath string
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		gotPath = r.URL.Path
+		w.Header().Set("Content-Type", "application/json")
+		_, _ = w.Write([]byte(`{"object":"list","data":[]}`))
+	}))
+	defer server.Close()
+
+	provider := NewWithHTTPClient("cpk_test", "https://unused.example/v1", server.Client(), llmclient.Hooks{})
+	provider.SetBaseURL(server.URL)
+	if _, err := provider.ListModels(context.Background()); err != nil {
+		t.Fatalf("ListModels() error = %v", err)
+	}
+	if gotPath != "/models" {
+		t.Fatalf("path = %q, want /models", gotPath)
+	}
+}
+
 func TestChatCompletion_UsesBearerAuthAndChatEndpoint(t *testing.T) {
 	var gotPath, gotAuth string
 
@@ -200,61 +233,44 @@ func TestResponses_TranslatesToChatCompletions(t *testing.T) {
 	}
 }
 
-func TestListModels_PreservesChutesMetadata(t *testing.T) {
+func TestStreamResponses_TranslatesToChatCompletions(t *testing.T) {
 	var gotPath, gotAuth string
+	var gotBody map[string]any
+
 	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		gotPath = r.URL.Path
 		gotAuth = r.Header.Get("Authorization")
-		w.Header().Set("Content-Type", "application/json")
-		_, _ = w.Write([]byte(`{
-			"object":"chutes-model-catalog",
-			"data":[{
-				"id":"Qwen/Qwen3.5-397B-A17B-TEE",
-				"owned_by":"sglang",
-				"created":1677652288,
-				"context_length":262144,
-				"max_output_length":65536,
-				"input_modalities":["text","image"],
-				"supported_features":["json_mode","tools","structured_outputs","reasoning"],
-				"confidential_compute":true,
-				"pricing":{"prompt":0.45,"completion":3.0,"input_cache_read":0.045}
-			}]
-		}`))
+		if err := json.NewDecoder(r.Body).Decode(&gotBody); err != nil {
+			http.Error(w, "decode error", http.StatusBadRequest)
+			return
+		}
+		w.Header().Set("Content-Type", "text/event-stream")
+		_, _ = io.WriteString(w, "data: {\"id\":\"chatcmpl-chutes\",\"object\":\"chat.completion.chunk\",\"created\":1677652288,\"model\":\"Qwen/Qwen3-32B-TEE\",\"choices\":[{\"index\":0,\"delta\":{\"content\":\"ok\"},\"finish_reason\":null}]}\n\ndata: [DONE]\n\n")
 	}))
 	defer server.Close()
 
 	provider := NewWithHTTPClient("cpk_test", server.URL, server.Client(), llmclient.Hooks{})
-	resp, err := provider.ListModels(context.Background())
+	stream, err := provider.StreamResponses(context.Background(), &core.ResponsesRequest{
+		Model: "Qwen/Qwen3-32B-TEE",
+		Input: "hi",
+	})
 	if err != nil {
-		t.Fatalf("ListModels() error = %v", err)
-	}
-	if gotPath != "/models" || gotAuth != "Bearer cpk_test" {
-		t.Fatalf("request path/auth = %q/%q, want /models/Bearer cpk_test", gotPath, gotAuth)
-	}
-	if len(resp.Data) != 1 {
-		t.Fatalf("len(resp.Data) = %d, want 1", len(resp.Data))
+		t.Fatalf("StreamResponses() error = %v", err)
 	}
-	if resp.Object != "list" {
-		t.Fatalf("resp.Object = %q, want list", resp.Object)
-	}
-	model := resp.Data[0]
-	if model.Object != "model" {
-		t.Fatalf("model.Object = %q, want model", model.Object)
-	}
-	if model.Metadata == nil || model.Metadata.ContextWindow == nil || *model.Metadata.ContextWindow != 262144 {
-		t.Fatalf("model context metadata = %+v, want 262144", model.Metadata)
+	defer stream.Close()
+
+	body, err := io.ReadAll(stream)
+	if err != nil {
+		t.Fatalf("ReadAll() error = %v", err)
 	}
-	if model.Metadata.MaxOutputTokens == nil || *model.Metadata.MaxOutputTokens != 65536 {
-		t.Fatalf("max output tokens = %+v, want 65536", model.Metadata.MaxOutputTokens)
+	if gotPath != "/chat/completions" || gotAuth != "Bearer cpk_test" {
+		t.Fatalf("request path/auth = %q/%q", gotPath, gotAuth)
 	}
-	if !model.Metadata.Capabilities["tools"] || !model.Metadata.Capabilities["vision"] || !model.Metadata.Capabilities["confidential_compute"] {
-		t.Fatalf("capabilities = %v, want tools, vision, and confidential_compute", model.Metadata.Capabilities)
+	if gotBody["model"] != "Qwen/Qwen3-32B-TEE" || gotBody["stream"] != true {
+		t.Fatalf("stream request body = %#v", gotBody)
 	}
-	pricing := model.Metadata.Pricing
-	if pricing == nil || pricing.Currency != "USD" || pricing.InputPerMtok == nil || *pricing.InputPerMtok != 0.45 ||
-		pricing.OutputPerMtok == nil || *pricing.OutputPerMtok != 3.0 ||
-		pricing.CachedInputPerMtok == nil || *pricing.CachedInputPerMtok != 0.045 {
-		t.Fatalf("pricing = %+v, want Chutes per-MTok USD pricing", pricing)
+	if raw := string(body); !strings.Contains(raw, "response.output_text.delta") || !strings.Contains(raw, "data: [DONE]") {
+		t.Fatalf("converted stream missing Responses events or done marker: %s", raw)
 	}
 }
 
diff --git a/internal/providers/chutes/models_test.go b/internal/providers/chutes/models_test.go
new file mode 100644
index 000000000..fb269eee5
--- /dev/null
+++ b/internal/providers/chutes/models_test.go
@@ -0,0 +1,143 @@
+package chutes
+
+import (
+	"context"
+	"net/http"
+	"net/http/httptest"
+	"testing"
+
+	"github.com/enterpilot/gomodel/internal/core"
+	"github.com/enterpilot/gomodel/internal/llmclient"
+)
+
+func TestListModels_PreservesChutesMetadata(t *testing.T) {
+	var gotPath, gotAuth string
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		gotPath = r.URL.Path
+		gotAuth = r.Header.Get("Authorization")
+		w.Header().Set("Content-Type", "application/json")
+		_, _ = w.Write([]byte(`{
+			"object":"chutes-model-catalog",
+			"data":[{
+				"id":"Qwen/Qwen3.5-397B-A17B-TEE",
+				"owned_by":"sglang",
+				"created":1677652288,
+				"context_length":262144,
+				"max_output_length":65536,
+				"input_modalities":["text","image"],
+				"supported_features":["json_mode","tools","structured_outputs","reasoning"],
+				"confidential_compute":true,
+				"pricing":{"prompt":0.45,"completion":3.0,"input_cache_read":0.045}
+			}]
+		}`))
+	}))
+	defer server.Close()
+
+	provider := NewWithHTTPClient("cpk_test", server.URL, server.Client(), llmclient.Hooks{})
+	resp, err := provider.ListModels(context.Background())
+	if err != nil {
+		t.Fatalf("ListModels() error = %v", err)
+	}
+	if gotPath != "/models" || gotAuth != "Bearer cpk_test" {
+		t.Fatalf("request path/auth = %q/%q, want /models/Bearer cpk_test", gotPath, gotAuth)
+	}
+	if len(resp.Data) != 1 {
+		t.Fatalf("len(resp.Data) = %d, want 1", len(resp.Data))
+	}
+	if resp.Object != "list" {
+		t.Fatalf("resp.Object = %q, want list", resp.Object)
+	}
+	model := resp.Data[0]
+	if model.Object != "model" {
+		t.Fatalf("model.Object = %q, want model", model.Object)
+	}
+	if model.Metadata == nil || model.Metadata.ContextWindow == nil || *model.Metadata.ContextWindow != 262144 {
+		t.Fatalf("model context metadata = %+v, want 262144", model.Metadata)
+	}
+	if model.Metadata.MaxOutputTokens == nil || *model.Metadata.MaxOutputTokens != 65536 {
+		t.Fatalf("max output tokens = %+v, want 65536", model.Metadata.MaxOutputTokens)
+	}
+	if !model.Metadata.Capabilities["tools"] || !model.Metadata.Capabilities["vision"] || !model.Metadata.Capabilities["confidential_compute"] {
+		t.Fatalf("capabilities = %v, want tools, vision, and confidential_compute", model.Metadata.Capabilities)
+	}
+	pricing := model.Metadata.Pricing
+	if pricing == nil || pricing.Currency != "USD" || pricing.InputPerMtok == nil || *pricing.InputPerMtok != 0.45 ||
+		pricing.OutputPerMtok == nil || *pricing.OutputPerMtok != 3.0 ||
+		pricing.CachedInputPerMtok == nil || *pricing.CachedInputPerMtok != 0.045 {
+		t.Fatalf("pricing = %+v, want Chutes per-MTok USD pricing", pricing)
+	}
+}
+
+func TestListModels_FiltersBlankIDsAndKeepsMinimalModels(t *testing.T) {
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
+		w.Header().Set("Content-Type", "application/json")
+		_, _ = w.Write([]byte(`{
+			"data":[
+				{"id":"   "},
+				{"id":" minimal-model ","object":"model","owned_by":" chutes ","pricing":{}}
+			]
+		}`))
+	}))
+	defer server.Close()
+
+	provider := NewWithHTTPClient("cpk_test", server.URL, server.Client(), llmclient.Hooks{})
+	resp, err := provider.ListModels(context.Background())
+	if err != nil {
+		t.Fatalf("ListModels() error = %v", err)
+	}
+	if len(resp.Data) != 1 {
+		t.Fatalf("models = %+v, want one non-blank model", resp.Data)
+	}
+	model := resp.Data[0]
+	if model.ID != "minimal-model" || model.Object != "model" || model.OwnedBy != "chutes" {
+		t.Fatalf("model identity = %+v, want trimmed minimal model", model)
+	}
+	if model.Metadata.ContextWindow != nil || model.Metadata.MaxOutputTokens != nil ||
+		model.Metadata.Capabilities != nil || model.Metadata.Pricing != nil {
+		t.Fatalf("optional metadata = %+v, want omitted zero values", model.Metadata)
+	}
+}
+
+func TestListModels_ReturnsUpstreamError(t *testing.T) {
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
+		w.WriteHeader(http.StatusServiceUnavailable)
+		_, _ = w.Write([]byte(`{"error":{"message":"catalog unavailable"}}`))
+	}))
+	defer server.Close()
+
+	provider := NewWithHTTPClient("cpk_test", server.URL, server.Client(), llmclient.Hooks{})
+	_, err := provider.ListModels(context.Background())
+	if err == nil {
+		t.Fatal("ListModels() error = nil, want upstream error")
+	}
+	gatewayErr, ok := err.(*core.GatewayError)
+	if !ok || gatewayErr.StatusCode != http.StatusServiceUnavailable {
+		t.Fatalf("error = %#v, want 503 *core.GatewayError", err)
+	}
+}
+
+func TestModelCapabilities_MapsOptionalModalities(t *testing.T) {
+	capabilities := modelCapabilities(modelInfo{
+		SupportedFeatures: []string{" JSON_Mode ", "   "},
+		InputModalities:   []string{"audio", "video", "unknown"},
+	})
+	if !capabilities["json_mode"] || !capabilities["audio"] || !capabilities["video"] {
+		t.Fatalf("capabilities = %v, want normalized json_mode, audio, and video", capabilities)
+	}
+}
+
+func TestModelPricing_HandlesNilAndPartialPrices(t *testing.T) {
+	var absent *modelPricing
+	if got := absent.toCore(); got != nil {
+		t.Fatalf("nil pricing = %+v, want nil", got)
+	}
+
+	prompt := 0.25
+	got := (&modelPricing{Prompt: &prompt}).toCore()
+	if got == nil || got.InputPerMtok == nil || *got.InputPerMtok != prompt {
+		t.Fatalf("partial pricing = %+v, want input price", got)
+	}
+	if got.OutputPerMtok != nil || got.CachedInputPerMtok != nil {
+		t.Fatalf("partial pricing = %+v, want absent optional prices", got)
+	}
+}

From d0dc94346849ecdcee815bb4e85cb6cd4099e428 Mon Sep 17 00:00:00 2001
From: "Jakub A. W" <jakubwasek@gmail.com>
Date: Sat, 8 Aug 2026 21:56:28 +0200
Subject: [PATCH 4/7] test(chutes): strengthen request contract assertions

---
 internal/providers/chutes/chutes_test.go | 36 +++++++++++++++---------
 internal/providers/chutes/models_test.go |  7 +++--
 2 files changed, 27 insertions(+), 16 deletions(-)

diff --git a/internal/providers/chutes/chutes_test.go b/internal/providers/chutes/chutes_test.go
index 533fc8df9..5e0416d85 100644
--- a/internal/providers/chutes/chutes_test.go
+++ b/internal/providers/chutes/chutes_test.go
@@ -28,8 +28,9 @@ func TestNew_ConstructsRegisteredProvider(t *testing.T) {
 }
 
 func TestSetBaseURL_ChangesRequestTarget(t *testing.T) {
-	var gotPath string
+	var gotMethod, gotPath string
 	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		gotMethod = r.Method
 		gotPath = r.URL.Path
 		w.Header().Set("Content-Type", "application/json")
 		_, _ = w.Write([]byte(`{"object":"list","data":[]}`))
@@ -41,8 +42,8 @@ func TestSetBaseURL_ChangesRequestTarget(t *testing.T) {
 	if _, err := provider.ListModels(context.Background()); err != nil {
 		t.Fatalf("ListModels() error = %v", err)
 	}
-	if gotPath != "/models" {
-		t.Fatalf("path = %q, want /models", gotPath)
+	if gotMethod != http.MethodGet || gotPath != "/models" {
+		t.Fatalf("method/path = %q/%q, want GET /models", gotMethod, gotPath)
 	}
 }
 
@@ -84,7 +85,10 @@ func TestChatCompletion_UsesBearerAuthAndChatEndpoint(t *testing.T) {
 
 func TestStreamChatCompletion_UsesBearerAuthAndChatEndpoint(t *testing.T) {
 	var gotPath, gotAuth string
-	var gotBody map[string]any
+	var gotBody struct {
+		Model  string `json:"model"`
+		Stream bool   `json:"stream"`
+	}
 
 	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		gotPath = r.URL.Path
@@ -115,7 +119,7 @@ func TestStreamChatCompletion_UsesBearerAuthAndChatEndpoint(t *testing.T) {
 	if gotPath != "/chat/completions" || gotAuth != "Bearer cpk_test" {
 		t.Fatalf("request path/auth = %q/%q, want /chat/completions/Bearer cpk_test", gotPath, gotAuth)
 	}
-	if gotBody["model"] != "Qwen/Qwen3-32B-TEE" || gotBody["stream"] != true {
+	if gotBody.Model != "Qwen/Qwen3-32B-TEE" || !gotBody.Stream {
 		t.Fatalf("stream request body = %#v", gotBody)
 	}
 	if !strings.Contains(string(body), "data: [DONE]") {
@@ -195,7 +199,9 @@ func TestPassthrough_ForwardsOpaqueRequest(t *testing.T) {
 
 func TestResponses_TranslatesToChatCompletions(t *testing.T) {
 	var gotPath string
-	var gotBody map[string]any
+	var gotBody struct {
+		Model string `json:"model"`
+	}
 
 	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		gotPath = r.URL.Path
@@ -225,8 +231,8 @@ func TestResponses_TranslatesToChatCompletions(t *testing.T) {
 	if gotPath != "/chat/completions" {
 		t.Fatalf("path = %q, want /chat/completions", gotPath)
 	}
-	if gotBody["model"] != "Qwen/Qwen3-32B-TEE" {
-		t.Fatalf("request model = %#v, want Qwen/Qwen3-32B-TEE", gotBody["model"])
+	if gotBody.Model != "Qwen/Qwen3-32B-TEE" {
+		t.Fatalf("request model = %q, want Qwen/Qwen3-32B-TEE", gotBody.Model)
 	}
 	if resp.Object != "response" || resp.Status != "completed" {
 		t.Fatalf("response metadata = object %q status %q, want response/completed", resp.Object, resp.Status)
@@ -234,10 +240,14 @@ func TestResponses_TranslatesToChatCompletions(t *testing.T) {
 }
 
 func TestStreamResponses_TranslatesToChatCompletions(t *testing.T) {
-	var gotPath, gotAuth string
-	var gotBody map[string]any
+	var gotMethod, gotPath, gotAuth string
+	var gotBody struct {
+		Model  string `json:"model"`
+		Stream bool   `json:"stream"`
+	}
 
 	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		gotMethod = r.Method
 		gotPath = r.URL.Path
 		gotAuth = r.Header.Get("Authorization")
 		if err := json.NewDecoder(r.Body).Decode(&gotBody); err != nil {
@@ -263,10 +273,10 @@ func TestStreamResponses_TranslatesToChatCompletions(t *testing.T) {
 	if err != nil {
 		t.Fatalf("ReadAll() error = %v", err)
 	}
-	if gotPath != "/chat/completions" || gotAuth != "Bearer cpk_test" {
-		t.Fatalf("request path/auth = %q/%q", gotPath, gotAuth)
+	if gotMethod != http.MethodPost || gotPath != "/chat/completions" || gotAuth != "Bearer cpk_test" {
+		t.Fatalf("request method/path/auth = %q/%q/%q", gotMethod, gotPath, gotAuth)
 	}
-	if gotBody["model"] != "Qwen/Qwen3-32B-TEE" || gotBody["stream"] != true {
+	if gotBody.Model != "Qwen/Qwen3-32B-TEE" || !gotBody.Stream {
 		t.Fatalf("stream request body = %#v", gotBody)
 	}
 	if raw := string(body); !strings.Contains(raw, "response.output_text.delta") || !strings.Contains(raw, "data: [DONE]") {
diff --git a/internal/providers/chutes/models_test.go b/internal/providers/chutes/models_test.go
index fb269eee5..1904df1fd 100644
--- a/internal/providers/chutes/models_test.go
+++ b/internal/providers/chutes/models_test.go
@@ -11,8 +11,9 @@ import (
 )
 
 func TestListModels_PreservesChutesMetadata(t *testing.T) {
-	var gotPath, gotAuth string
+	var gotMethod, gotPath, gotAuth string
 	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		gotMethod = r.Method
 		gotPath = r.URL.Path
 		gotAuth = r.Header.Get("Authorization")
 		w.Header().Set("Content-Type", "application/json")
@@ -38,8 +39,8 @@ func TestListModels_PreservesChutesMetadata(t *testing.T) {
 	if err != nil {
 		t.Fatalf("ListModels() error = %v", err)
 	}
-	if gotPath != "/models" || gotAuth != "Bearer cpk_test" {
-		t.Fatalf("request path/auth = %q/%q, want /models/Bearer cpk_test", gotPath, gotAuth)
+	if gotMethod != http.MethodGet || gotPath != "/models" || gotAuth != "Bearer cpk_test" {
+		t.Fatalf("request method/path/auth = %q/%q/%q, want GET /models/Bearer cpk_test", gotMethod, gotPath, gotAuth)
 	}
 	if len(resp.Data) != 1 {
 		t.Fatalf("len(resp.Data) = %d, want 1", len(resp.Data))

From a5a661da33b2e9fc46701c2601229bf2f8346e7e Mon Sep 17 00:00:00 2001
From: "Jakub A. W" <jakubwasek@gmail.com>
Date: Sat, 8 Aug 2026 22:42:41 +0200
Subject: [PATCH 5/7] feat(chutes): enable passthrough by default

---
 config/config.example.yaml             |  2 +-
 config/config.go                       |  1 +
 config/config_test.go                  |  4 ++--
 config/server.go                       |  2 +-
 docs/features/passthrough-api.mdx      |  4 ++--
 internal/server/handlers_test.go       | 31 +++++++++++++++++++++++++-
 internal/server/passthrough_support.go |  2 +-
 7 files changed, 38 insertions(+), 8 deletions(-)

diff --git a/config/config.example.yaml b/config/config.example.yaml
index 8142a5d38..8133ba427 100644
--- a/config/config.example.yaml
+++ b/config/config.example.yaml
@@ -13,7 +13,7 @@ server:
   enable_passthrough_routes: true # expose /p/{provider}/{endpoint} passthrough routes
   allow_passthrough_v1_alias: true # allow /p/{provider}/v1/... while keeping /p/{provider}/... canonical
   user_path_header: "X-GoModel-User-Path" # env: USER_PATH_HEADER; inbound header used for user_path scoping
-  enabled_passthrough_providers: ["openai", "anthropic", "cohere", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek", "bailian"] # providers enabled on /p/{provider}/...
+  enabled_passthrough_providers: ["openai", "anthropic", "chutes", "cohere", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek", "bailian"] # providers enabled on /p/{provider}/...
   realtime_enabled: true # env: REALTIME_ENABLED; expose /v1/realtime websocket and /p/{provider}/v1/realtime upgrades (OpenAI only)
   pid_file: "data/gomodel.pid" # env: PID_FILE; where the running gateway records its process id so `gomodel --reload` can find it. Set per instance when several gateways share a host; empty writes no pid file and disables --reload; changing it needs a restart, not a reload
 
diff --git a/config/config.go b/config/config.go
index b4d0b6e45..fc0d8f82b 100644
--- a/config/config.go
+++ b/config/config.go
@@ -97,6 +97,7 @@ func buildDefaultConfig() *Config {
 			EnabledPassthroughProviders: []string{
 				"openai",
 				"anthropic",
+				"chutes",
 				"openrouter",
 				"kilo",
 				"zai",
diff --git a/config/config_test.go b/config/config_test.go
index 1182bc288..044d31b51 100644
--- a/config/config_test.go
+++ b/config/config_test.go
@@ -126,7 +126,7 @@ func TestBuildDefaultConfig(t *testing.T) {
 	if !cfg.Server.AllowPassthroughV1Alias {
 		t.Error("expected Server.AllowPassthroughV1Alias=true")
 	}
-	if got, want := cfg.Server.EnabledPassthroughProviders, []string{"openai", "anthropic", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek"}; !reflect.DeepEqual(got, want) {
+	if got, want := cfg.Server.EnabledPassthroughProviders, []string{"openai", "anthropic", "chutes", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek"}; !reflect.DeepEqual(got, want) {
 		t.Errorf("expected Server.EnabledPassthroughProviders=%v, got %v", want, got)
 	}
 	if cfg.Models.ConfiguredProviderModelsMode != ConfiguredProviderModelsModeFallback {
@@ -1203,7 +1203,7 @@ func TestLoad_ConfigExample_UsesNestedModelCacheSettings(t *testing.T) {
 			t.Fatalf("expected Cache.Model.Redis to be nil in example config, got %+v", result.Config.Cache.Model.Redis)
 		}
 		gotProviders := result.Config.Server.EnabledPassthroughProviders
-		wantProviders := []string{"openai", "anthropic", "cohere", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek", "bailian"}
+		wantProviders := []string{"openai", "anthropic", "chutes", "cohere", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek", "bailian"}
 		if !reflect.DeepEqual(gotProviders, wantProviders) {
 			t.Fatalf("Server.EnabledPassthroughProviders = %v, want %v", gotProviders, wantProviders)
 		}
diff --git a/config/server.go b/config/server.go
index e0af13d47..bf9f657a0 100644
--- a/config/server.go
+++ b/config/server.go
@@ -39,7 +39,7 @@ type ServerConfig struct {
 	UserPathHeader string `yaml:"user_path_header" env:"USER_PATH_HEADER"`
 	// EnabledPassthroughProviders lists the provider types enabled on
 	// /p/{provider}/... passthrough routes. Default:
-	// ["openai", "anthropic", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek"].
+	// ["openai", "anthropic", "chutes", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek"].
 	EnabledPassthroughProviders []string `yaml:"enabled_passthrough_providers" env:"ENABLED_PASSTHROUGH_PROVIDERS"`
 	// RealtimeEnabled exposes the realtime (speech-to-speech) websocket endpoint
 	// at /v1/realtime and the /p/{provider}/v1/realtime passthrough upgrade.
diff --git a/docs/features/passthrough-api.mdx b/docs/features/passthrough-api.mdx
index 2f6e985d2..c1deb825a 100644
--- a/docs/features/passthrough-api.mdx
+++ b/docs/features/passthrough-api.mdx
@@ -131,7 +131,7 @@ from passthrough requests before forwarding them upstream.
 
 Passthrough is intentionally narrow while the API is in beta.
 
-- `openai`, `anthropic`, `openrouter`, `kilo`, `zai`, `sglang`, `vllm`, `llmd`, and `deepseek` are enabled by
+- `openai`, `anthropic`, `chutes`, `openrouter`, `kilo`, `zai`, `sglang`, `vllm`, `llmd`, and `deepseek` are enabled by
   default.
 - GoModel does not translate passthrough request bodies or response bodies.
 - Provider-native error bodies and status codes are proxied instead of converted
@@ -146,7 +146,7 @@ Passthrough routes are enabled by default:
 ```env
 ENABLE_PASSTHROUGH_ROUTES=true
 ALLOW_PASSTHROUGH_V1_ALIAS=true
-ENABLED_PASSTHROUGH_PROVIDERS=openai,anthropic,openrouter,kilo,zai,sglang,vllm,llmd,deepseek
+ENABLED_PASSTHROUGH_PROVIDERS=openai,anthropic,chutes,openrouter,kilo,zai,sglang,vllm,llmd,deepseek
 ```
 
 Set `ENABLED_PASSTHROUGH_PROVIDERS` to the provider types you want to expose.
diff --git a/internal/server/handlers_test.go b/internal/server/handlers_test.go
index 1b9335efb..26bc32961 100644
--- a/internal/server/handlers_test.go
+++ b/internal/server/handlers_test.go
@@ -7178,11 +7178,40 @@ func TestProviderPassthrough_RejectsUnsupportedProvider(t *testing.T) {
 	if !strings.Contains(rec.Body.String(), `provider passthrough for \"groq\" is not enabled`) {
 		t.Fatalf("unexpected error body: %s", rec.Body.String())
 	}
-	if !strings.Contains(rec.Body.String(), "anthropic, deepseek, kilo, llmd, openai, openrouter, sglang, vllm, zai") {
+	if !strings.Contains(rec.Body.String(), "anthropic, chutes, deepseek, kilo, llmd, openai, openrouter, sglang, vllm, zai") {
 		t.Fatalf("unexpected error body: %s", rec.Body.String())
 	}
 }
 
+func TestProviderPassthrough_DefaultAllowsChutes(t *testing.T) {
+	provider := &mockProvider{
+		passthroughResponse: &core.PassthroughResponse{
+			StatusCode: http.StatusOK,
+			Headers:    http.Header{"Content-Type": []string{"application/json"}},
+			Body:       io.NopCloser(strings.NewReader(`{"ok":true}`)),
+		},
+	}
+
+	e := echo.New()
+	handler := NewHandler(provider, nil, nil, nil)
+	e.POST("/p/:provider/*", handler.ProviderPassthrough)
+
+	req := httptest.NewRequest(http.MethodPost, "/p/chutes/chat/completions", strings.NewReader(`{"model":"Qwen/Qwen3-32B-TEE"}`))
+	req.Header.Set("Content-Type", "application/json")
+	rec := httptest.NewRecorder()
+	e.ServeHTTP(rec, req)
+
+	if rec.Code != http.StatusOK {
+		t.Fatalf("status = %d, want 200: %s", rec.Code, rec.Body.String())
+	}
+	if provider.lastPassthroughProvider != "chutes" {
+		t.Fatalf("providerType = %q, want chutes", provider.lastPassthroughProvider)
+	}
+	if provider.lastPassthroughReq == nil || provider.lastPassthroughReq.Endpoint != "chat/completions" {
+		t.Fatalf("passthrough request = %+v, want chat/completions endpoint", provider.lastPassthroughReq)
+	}
+}
+
 func TestProviderPassthrough_UsesConfiguredSupportedProviders(t *testing.T) {
 	provider := &mockProvider{
 		passthroughResponse: &core.PassthroughResponse{
diff --git a/internal/server/passthrough_support.go b/internal/server/passthrough_support.go
index 5bce2f747..1bdf0bb5a 100644
--- a/internal/server/passthrough_support.go
+++ b/internal/server/passthrough_support.go
@@ -16,7 +16,7 @@ import (
 	"github.com/enterpilot/gomodel/internal/usage"
 )
 
-var defaultEnabledPassthroughProviders = []string{"openai", "anthropic", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek"}
+var defaultEnabledPassthroughProviders = []string{"openai", "anthropic", "chutes", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek"}
 
 const llmdDroppedReasonHeader = "X-Llm-D-Request-Dropped-Reason"
 

From 9d421be3c795d133a07cf77146661f6606905494 Mon Sep 17 00:00:00 2001
From: "Jakub A. W" <jakubwasek@gmail.com>
Date: Sat, 8 Aug 2026 23:16:07 +0200
Subject: [PATCH 6/7] fix(chutes): require passthrough opt-in

---
 config/config.example.yaml             |  2 +-
 config/config.go                       |  1 -
 config/config_test.go                  |  4 ++--
 config/server.go                       |  2 +-
 docs/features/passthrough-api.mdx      |  8 ++++++--
 docs/providers/overview.mdx            |  4 +++-
 internal/server/handlers_test.go       | 18 +++++++++++++++---
 internal/server/passthrough_support.go |  2 +-
 8 files changed, 29 insertions(+), 12 deletions(-)

diff --git a/config/config.example.yaml b/config/config.example.yaml
index 8133ba427..8142a5d38 100644
--- a/config/config.example.yaml
+++ b/config/config.example.yaml
@@ -13,7 +13,7 @@ server:
   enable_passthrough_routes: true # expose /p/{provider}/{endpoint} passthrough routes
   allow_passthrough_v1_alias: true # allow /p/{provider}/v1/... while keeping /p/{provider}/... canonical
   user_path_header: "X-GoModel-User-Path" # env: USER_PATH_HEADER; inbound header used for user_path scoping
-  enabled_passthrough_providers: ["openai", "anthropic", "chutes", "cohere", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek", "bailian"] # providers enabled on /p/{provider}/...
+  enabled_passthrough_providers: ["openai", "anthropic", "cohere", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek", "bailian"] # providers enabled on /p/{provider}/...
   realtime_enabled: true # env: REALTIME_ENABLED; expose /v1/realtime websocket and /p/{provider}/v1/realtime upgrades (OpenAI only)
   pid_file: "data/gomodel.pid" # env: PID_FILE; where the running gateway records its process id so `gomodel --reload` can find it. Set per instance when several gateways share a host; empty writes no pid file and disables --reload; changing it needs a restart, not a reload
 
diff --git a/config/config.go b/config/config.go
index fc0d8f82b..b4d0b6e45 100644
--- a/config/config.go
+++ b/config/config.go
@@ -97,7 +97,6 @@ func buildDefaultConfig() *Config {
 			EnabledPassthroughProviders: []string{
 				"openai",
 				"anthropic",
-				"chutes",
 				"openrouter",
 				"kilo",
 				"zai",
diff --git a/config/config_test.go b/config/config_test.go
index 044d31b51..1182bc288 100644
--- a/config/config_test.go
+++ b/config/config_test.go
@@ -126,7 +126,7 @@ func TestBuildDefaultConfig(t *testing.T) {
 	if !cfg.Server.AllowPassthroughV1Alias {
 		t.Error("expected Server.AllowPassthroughV1Alias=true")
 	}
-	if got, want := cfg.Server.EnabledPassthroughProviders, []string{"openai", "anthropic", "chutes", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek"}; !reflect.DeepEqual(got, want) {
+	if got, want := cfg.Server.EnabledPassthroughProviders, []string{"openai", "anthropic", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek"}; !reflect.DeepEqual(got, want) {
 		t.Errorf("expected Server.EnabledPassthroughProviders=%v, got %v", want, got)
 	}
 	if cfg.Models.ConfiguredProviderModelsMode != ConfiguredProviderModelsModeFallback {
@@ -1203,7 +1203,7 @@ func TestLoad_ConfigExample_UsesNestedModelCacheSettings(t *testing.T) {
 			t.Fatalf("expected Cache.Model.Redis to be nil in example config, got %+v", result.Config.Cache.Model.Redis)
 		}
 		gotProviders := result.Config.Server.EnabledPassthroughProviders
-		wantProviders := []string{"openai", "anthropic", "chutes", "cohere", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek", "bailian"}
+		wantProviders := []string{"openai", "anthropic", "cohere", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek", "bailian"}
 		if !reflect.DeepEqual(gotProviders, wantProviders) {
 			t.Fatalf("Server.EnabledPassthroughProviders = %v, want %v", gotProviders, wantProviders)
 		}
diff --git a/config/server.go b/config/server.go
index bf9f657a0..e0af13d47 100644
--- a/config/server.go
+++ b/config/server.go
@@ -39,7 +39,7 @@ type ServerConfig struct {
 	UserPathHeader string `yaml:"user_path_header" env:"USER_PATH_HEADER"`
 	// EnabledPassthroughProviders lists the provider types enabled on
 	// /p/{provider}/... passthrough routes. Default:
-	// ["openai", "anthropic", "chutes", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek"].
+	// ["openai", "anthropic", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek"].
 	EnabledPassthroughProviders []string `yaml:"enabled_passthrough_providers" env:"ENABLED_PASSTHROUGH_PROVIDERS"`
 	// RealtimeEnabled exposes the realtime (speech-to-speech) websocket endpoint
 	// at /v1/realtime and the /p/{provider}/v1/realtime passthrough upgrade.
diff --git a/docs/features/passthrough-api.mdx b/docs/features/passthrough-api.mdx
index c1deb825a..6e59b89fc 100644
--- a/docs/features/passthrough-api.mdx
+++ b/docs/features/passthrough-api.mdx
@@ -131,8 +131,12 @@ from passthrough requests before forwarding them upstream.
 
 Passthrough is intentionally narrow while the API is in beta.
 
-- `openai`, `anthropic`, `chutes`, `openrouter`, `kilo`, `zai`, `sglang`, `vllm`, `llmd`, and `deepseek` are enabled by
+- `openai`, `anthropic`, `openrouter`, `kilo`, `zai`, `sglang`, `vllm`, `llmd`, and `deepseek` are enabled by
   default.
+- Chutes supports passthrough but requires explicit operator opt-in because
+  passthrough can forward provider-native routes that do not identify a model.
+  Add `chutes` to `ENABLED_PASSTHROUGH_PROVIDERS` only when you intend to expose
+  that surface.
 - GoModel does not translate passthrough request bodies or response bodies.
 - Provider-native error bodies and status codes are proxied instead of converted
   into OpenAI-compatible responses.
@@ -146,7 +150,7 @@ Passthrough routes are enabled by default:
 ```env
 ENABLE_PASSTHROUGH_ROUTES=true
 ALLOW_PASSTHROUGH_V1_ALIAS=true
-ENABLED_PASSTHROUGH_PROVIDERS=openai,anthropic,chutes,openrouter,kilo,zai,sglang,vllm,llmd,deepseek
+ENABLED_PASSTHROUGH_PROVIDERS=openai,anthropic,openrouter,kilo,zai,sglang,vllm,llmd,deepseek
 ```
 
 Set `ENABLED_PASSTHROUGH_PROVIDERS` to the provider types you want to expose.
diff --git a/docs/providers/overview.mdx b/docs/providers/overview.mdx
index 50e32aad6..bce882554 100644
--- a/docs/providers/overview.mdx
+++ b/docs/providers/overview.mdx
@@ -94,7 +94,9 @@ support, not every individual model capability exposed by an upstream provider.
 - **Chutes AI** — defaults to `https://llm.chutes.ai/v1` and discovers its
   current model IDs, context limits, capabilities, and pricing from the live
   catalog. GoModel translates `/v1/responses` requests to chat completions;
-  Chutes' shared LLM endpoint does not expose embeddings.
+  Chutes' shared LLM endpoint does not expose embeddings. Passthrough support
+  requires explicit operator opt-in by adding `chutes` to
+  `ENABLED_PASSTHROUGH_PROVIDERS`.
 - **Meta (Muse Spark)** — the Meta Model API is OpenAI-compatible; set
   `META_API_KEY` and route to `muse-spark-1.1`. Muse Spark models are not in
   the upstream model catalog yet, so declare `context_window` and `pricing`
diff --git a/internal/server/handlers_test.go b/internal/server/handlers_test.go
index 26bc32961..ba9d92673 100644
--- a/internal/server/handlers_test.go
+++ b/internal/server/handlers_test.go
@@ -7178,12 +7178,12 @@ func TestProviderPassthrough_RejectsUnsupportedProvider(t *testing.T) {
 	if !strings.Contains(rec.Body.String(), `provider passthrough for \"groq\" is not enabled`) {
 		t.Fatalf("unexpected error body: %s", rec.Body.String())
 	}
-	if !strings.Contains(rec.Body.String(), "anthropic, chutes, deepseek, kilo, llmd, openai, openrouter, sglang, vllm, zai") {
+	if !strings.Contains(rec.Body.String(), "anthropic, deepseek, kilo, llmd, openai, openrouter, sglang, vllm, zai") {
 		t.Fatalf("unexpected error body: %s", rec.Body.String())
 	}
 }
 
-func TestProviderPassthrough_DefaultAllowsChutes(t *testing.T) {
+func TestProviderPassthrough_ChutesRequiresExplicitOptIn(t *testing.T) {
 	provider := &mockProvider{
 		passthroughResponse: &core.PassthroughResponse{
 			StatusCode: http.StatusOK,
@@ -7196,13 +7196,25 @@ func TestProviderPassthrough_DefaultAllowsChutes(t *testing.T) {
 	handler := NewHandler(provider, nil, nil, nil)
 	e.POST("/p/:provider/*", handler.ProviderPassthrough)
 
+	blockedReq := httptest.NewRequest(http.MethodPost, "/p/chutes/provider-native/admin/keys", strings.NewReader(`{}`))
+	blockedRec := httptest.NewRecorder()
+	e.ServeHTTP(blockedRec, blockedReq)
+
+	if blockedRec.Code != http.StatusBadRequest {
+		t.Fatalf("default status = %d, want 400: %s", blockedRec.Code, blockedRec.Body.String())
+	}
+	if provider.lastPassthroughReq != nil {
+		t.Fatal("default Chutes passthrough reached provider, want rejection before forwarding")
+	}
+
+	handler.setEnabledPassthroughProviders([]string{"chutes"})
 	req := httptest.NewRequest(http.MethodPost, "/p/chutes/chat/completions", strings.NewReader(`{"model":"Qwen/Qwen3-32B-TEE"}`))
 	req.Header.Set("Content-Type", "application/json")
 	rec := httptest.NewRecorder()
 	e.ServeHTTP(rec, req)
 
 	if rec.Code != http.StatusOK {
-		t.Fatalf("status = %d, want 200: %s", rec.Code, rec.Body.String())
+		t.Fatalf("opt-in status = %d, want 200: %s", rec.Code, rec.Body.String())
 	}
 	if provider.lastPassthroughProvider != "chutes" {
 		t.Fatalf("providerType = %q, want chutes", provider.lastPassthroughProvider)
diff --git a/internal/server/passthrough_support.go b/internal/server/passthrough_support.go
index 1bdf0bb5a..5bce2f747 100644
--- a/internal/server/passthrough_support.go
+++ b/internal/server/passthrough_support.go
@@ -16,7 +16,7 @@ import (
 	"github.com/enterpilot/gomodel/internal/usage"
 )
 
-var defaultEnabledPassthroughProviders = []string{"openai", "anthropic", "chutes", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek"}
+var defaultEnabledPassthroughProviders = []string{"openai", "anthropic", "openrouter", "kilo", "zai", "sglang", "vllm", "llmd", "deepseek"}
 
 const llmdDroppedReasonHeader = "X-Llm-D-Request-Dropped-Reason"
 

From 85c5bcbf96415201faf2e7d5ea154c5424131d6c Mon Sep 17 00:00:00 2001
From: "Jakub A. W" <jakubwasek@gmail.com>
Date: Sun, 9 Aug 2026 21:18:36 +0200
Subject: [PATCH 7/7] test(dashboard): cover all provider types

---
 run/providers_test.go                        | 24 ++++++++++++++++++++
 web/dashboard/tests/providers-config.test.js | 14 ++++++++++++
 2 files changed, 38 insertions(+)

diff --git a/run/providers_test.go b/run/providers_test.go
index 285038e83..3d3c27eec 100644
--- a/run/providers_test.go
+++ b/run/providers_test.go
@@ -21,6 +21,7 @@ func TestDefaultProviderFactoryCredentialForms(t *testing.T) {
 
 	tests := []struct {
 		providerType string
+		defaultURL   string
 		fields       []string // exact, in display order
 		required     []string
 		absent       []string
@@ -29,10 +30,19 @@ func TestDefaultProviderFactoryCredentialForms(t *testing.T) {
 		{
 			// The plain shape every API-key provider derives.
 			providerType: "openai",
+			defaultURL:   "https://api.openai.com/v1",
 			fields:       []string{"api_keys", "base_url", "session_sticky_keys", "models"},
 			required:     []string{"api_keys"},
 			absent:       []string{"api_version", "vertex_project"},
 		},
+		{
+			// Newly registered API-key providers use the same schema feed as
+			// every other type, so the dashboard can offer them immediately.
+			providerType: "chutes",
+			defaultURL:   "https://llm.chutes.ai/v1",
+			fields:       []string{"api_keys", "base_url", "session_sticky_keys", "models"},
+			required:     []string{"api_keys"},
+		},
 		{
 			// A deployment URL is the provider, so it is required, and Azure
 			// is the one type that takes an API version.
@@ -107,6 +117,9 @@ func TestDefaultProviderFactoryCredentialForms(t *testing.T) {
 			if !ok {
 				t.Fatalf("no credential schema for provider type %q", tt.providerType)
 			}
+			if tt.defaultURL != "" && schema.DefaultBaseURL != tt.defaultURL {
+				t.Errorf("DefaultBaseURL = %q, want %q", schema.DefaultBaseURL, tt.defaultURL)
+			}
 
 			var names []string
 			for _, field := range schema.Fields {
@@ -171,5 +184,16 @@ func TestDefaultProviderFactoryRegistersAllProviderTypes(t *testing.T) {
 		if !slices.Equal(got, expected) {
 			t.Errorf("metrics=%v: registered types = %v, want %v", metricsEnabled, got, expected)
 		}
+
+		// CredentialSchemas is the source for
+		// GET /admin/provider-credentials/types, which drives the dashboard's
+		// Add Provider selector. Keep it in exact lockstep with construction.
+		dashboardTypes := make([]string, 0, len(expected))
+		for _, schema := range factory.CredentialSchemas() {
+			dashboardTypes = append(dashboardTypes, schema.Type)
+		}
+		if !slices.Equal(dashboardTypes, expected) {
+			t.Errorf("metrics=%v: dashboard provider types = %v, want %v", metricsEnabled, dashboardTypes, expected)
+		}
 	}
 }
diff --git a/web/dashboard/tests/providers-config.test.js b/web/dashboard/tests/providers-config.test.js
index 3de1499f2..1075ed563 100644
--- a/web/dashboard/tests/providers-config.test.js
+++ b/web/dashboard/tests/providers-config.test.js
@@ -527,6 +527,20 @@ test("providerCredentialTypeOptions always includes the current selection", () =
   assert.deepEqual(providerCredentialTypeOptions(null, " "), []);
 });
 
+test("providerCredentialTypeOptions lists every server-supplied provider type", () => {
+  const recentProviderSchemas = ["chutes", "cohere", "llmd", "sglang"].map((type) => ({
+    type,
+    fields: [],
+  }));
+
+  assert.deepEqual(providerCredentialTypeOptions(recentProviderSchemas, ""), [
+    "chutes",
+    "cohere",
+    "llmd",
+    "sglang",
+  ]);
+});
+
 test("splitCommaList trims and drops empties", () => {
   assert.deepEqual(
     splitCommaList(" gpt-4o, gpt-4o-mini ,,"),