Skip to content

RAGFlow is an agent-native database: a Retrieval-Augmented Generation (RAG) engine that grew into context infrastructure.

feat: add Runcrate as a model provider (#21446)

Nirupam Bhowmick396525c7e0ba
diff --git a/conf/llm_factories.json b/conf/llm_factories.json
index c83a075713..fc656450c4 100644
--- a/conf/llm_factories.json
+++ b/conf/llm_factories.json
@@ -3524,6 +3524,14 @@
"llm": [],
"url": "https://api.empiriolabs.ai/v1"
},
+ {
+ "name": "Runcrate",
+ "logo": "",
+ "tags": "LLM,TEXT EMBEDDING,IMAGE2TEXT",
+ "status": "1",
+ "llm": [],
+ "url": "https://api.runcrate.ai/v1"
+ },
{
"name": "StepFun",
"logo": "",
diff --git a/conf/model.md b/conf/model.md
index 7715986566..a5191abcd0 100644
--- a/conf/model.md
+++ b/conf/model.md
@@ -193,6 +193,7 @@ The following providers are **aggregators** — they host models from multiple u
| **PPIO** | Edge AI platform |
| **Qiniu** | Chinese cloud platform |
| **Replicate** | Open-source model hosting |
+| **Runcrate** | GPU cloud with an OpenAI-compatible inference API for open models |
| **SiliconFlow** | Chinese aggregator |
| **TogetherAI** | Open-source model hosting |
| **TokenHub** | Multi-provider aggregator |
diff --git a/conf/models/runcrate.json b/conf/models/runcrate.json
new file mode 100644
index 0000000000..62c2cce7ec
--- /dev/null
+++ b/conf/models/runcrate.json
@@ -0,0 +1,114 @@
+{
+ "name": "Runcrate",
+ "url": {
+ "default": "https://api.runcrate.ai/v1"
+ },
+ "url_suffix": {
+ "chat": "chat/completions",
+ "models": "models",
+ "embedding": "embeddings"
+ },
+ "class": "runcrate",
+ "models": [
+ {
+ "name": "deepseek-ai/DeepSeek-V4.1-Flash",
+ "context_length": 1048576,
+ "model_types": [
+ "chat",
+ "vision"
+ ],
+ "tools": {
+ "support": true
+ }
+ },
+ {
+ "name": "deepseek-ai/DeepSeek-V4.1-Flash-1017",
+ "context_length": 1048576,
+ "model_types": [
+ "chat",
+ "vision"
+ ],
+ "tools": {
+ "support": true
+ }
+ },
+ {
+ "name": "zai-org/GLM-5.3",
+ "context_length": 1048576,
+ "model_types": [
+ "chat"
+ ],
+ "tools": {
+ "support": true
+ }
+ },
+ {
+ "name": "zai-org/GLM-5.3-Flash",
+ "context_length": 1048576,
+ "model_types": [
+ "chat",
+ "vision"
+ ],
+ "tools": {
+ "support": true
+ }
+ },
+ {
+ "name": "Qwen/Qwen3.8-Flash",
+ "context_length": 1000000,
+ "model_types": [
+ "chat"
+ ],
+ "tools": {
+ "support": true
+ }
+ },
+ {
+ "name": "openai/gpt-oss-120b",
+ "context_length": 131072,
+ "model_types": [
+ "chat"
+ ],
+ "tools": {
+ "support": true
+ }
+ },
+ {
+ "name": "moonshotai/Kimi-K2.6",
+ "context_length": 262144,
+ "max_output": 16384,
+ "model_types": [
+ "chat",
+ "vision"
+ ],
+ "tools": {
+ "support": true
+ }
+ },
+ {
+ "name": "Qwen/Qwen3-Coder-480B-A35B-Instruct-Turbo",
+ "context_length": 262144,
+ "max_output": 65536,
+ "model_types": [
+ "chat"
+ ],
+ "tools": {
+ "support": true
+ }
+ },
+ {
+ "name": "Qwen/Qwen3-Embedding-8B",
+ "max_tokens": 32768,
+ "model_types": [
+ "embedding"
+ ]
+ },
+ {
+ "name": "BAAI/bge-m3",
+ "max_tokens": 8192,
+ "model_types": [
+ "embedding"
+ ]
+ }
+ ]
+}
diff --git a/internal/entity/models/factory.go b/internal/entity/models/factory.go
index 87758ad80b..7dcda49b28 100644
--- a/internal/entity/models/factory.go
+++ b/internal/entity/models/factory.go
@@ -199,6 +199,8 @@ func (f *ModelFactory) CreateModelDriver(providerName string, baseURL map[string
return NewOpperModel(baseURL, urlSuffix), nil
case "empiriolabs ai":
return NewEmpirioLabsModel(baseURL, urlSuffix), nil
+ case "runcrate":
+ return NewRuncrateModel(baseURL, urlSuffix), nil
case "new api":
return NewNewAPIModel(baseURL, urlSuffix), nil
default:
diff --git a/internal/entity/models/runcrate.go b/internal/entity/models/runcrate.go
new file mode 100644
index 0000000000..3f059235f0
--- /dev/null
+++ b/internal/entity/models/runcrate.go
@@ -0,0 +1,42 @@
+//
+// Copyright 2026 The InfiniFlow Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+
+package models
+
+// RuncrateModel drives Runcrate (https://api.runcrate.ai/v1), a GPU cloud
+// whose inference API serves open models through OpenAI-compatible chat,
+// embedding and model-list endpoints, so the embedded
+// OpenAIAPICompatibleModel covers every call.
+type RuncrateModel struct {
+ *OpenAIAPICompatibleModel
+}
+
+// NewRuncrateModel creates a Runcrate driver.
+func NewRuncrateModel(baseURL map[string]string, urlSuffix URLSuffix) *RuncrateModel {
+ return &RuncrateModel{
+ OpenAIAPICompatibleModel: NewOpenAIAPICompatibleModel(baseURL, urlSuffix),
+ }
+}
+
+// Name returns the driver name.
+func (m *RuncrateModel) Name() string {
+ return "runcrate"
+}
+
+// NewInstance returns a driver bound to the given base URL.
+func (m *RuncrateModel) NewInstance(baseURL map[string]string) ModelDriver {
+ return NewRuncrateModel(baseURL, m.baseModel.URLSuffix)
+}
diff --git a/internal/entity/models/runcrate_test.go b/internal/entity/models/runcrate_test.go
new file mode 100644
index 0000000000..6421eaff34
--- /dev/null
+++ b/internal/entity/models/runcrate_test.go
@@ -0,0 +1,181 @@
+//
+// Copyright 2026 The InfiniFlow Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+
+package models
+
+import (
+ "encoding/json"
+ "net/http"
+ "net/http/httptest"
+ "ragflow/internal/common"
+ "testing"
+)
+
+// newRuncrateForTest creates a RuncrateModel instance for testing.
+func newRuncrateForTest(baseURL string) *RuncrateModel {
+ return NewRuncrateModel(
+ map[string]string{"default": baseURL},
+ URLSuffix{Chat: "chat/completions", Models: "models", Embedding: "embeddings"},
+ )
+}
+
+// TestRuncrateName tests the driver name.
+func TestRuncrateName(t *testing.T) {
+ if got := newRuncrateForTest("http://unused").Name(); got != "runcrate" {
+ t.Errorf("Name()=%q", got)
+ }
+}
+
+// TestRuncrateFactory tests creating the driver from factory.
+func TestRuncrateFactory(t *testing.T) {
+ driver, err := NewModelFactory().CreateModelDriver("runcrate", map[string]string{"default": "http://unused"}, URLSuffix{})
+ if err != nil {
+ t.Fatalf("CreateModelDriver: %v", err)
+ }
+ if _, ok := driver.(*RuncrateModel); !ok {
+ t.Fatalf("driver type=%T, want *RuncrateModel", driver)
+ }
+ if _, ok := driver.NewInstance(map[string]string{"default": "http://other"}).(*RuncrateModel); !ok {
+ t.Fatal("NewInstance did not return *RuncrateModel")
+ }
+}
+
+// The shipped config must resolve to the Runcrate driver. CreateModelDriver
+// dispatches on strings.ToLower(name), so the "Runcrate" provider name has
+// to match the "runcrate" case; anything else silently falls through to
+// DummyModel.
+func TestRuncrateProviderConfig(t *testing.T) {
+ dir, restore := setupProviderTestDir(t, "runcrate.json")
+ defer restore()
+
+ if err := InitProviderManager(dir); err != nil {
+ t.Fatalf("InitProviderManager: %v", err)
+ }
+
+ provider := GetProviderManager().FindProvider("Runcrate")
+ if provider == nil {
+ t.Fatal("Runcrate provider not found")
+ }
+ if _, ok := provider.ModelDriver.(*RuncrateModel); !ok {
+ t.Fatalf("ModelDriver=%T, want *models.RuncrateModel", provider.ModelDriver)
+ }
+}
+
+// TestRuncrateChatSendsAPIKey tests chat completion authentication and that
+// the organization-prefixed model id reaches the API unchanged.
+func TestRuncrateChatSendsAPIKey(t *testing.T) {
+ withSSRFBypass(t)
+ srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+ if r.URL.Path != "/chat/completions" {
+ t.Errorf("path=%s", r.URL.Path)
+ }
+ if got := r.Header.Get("Authorization"); got != "Bearer sk-gw-test" {
+ t.Errorf("Authorization=%q", got)
+ }
+ var body struct {
+ Model string `json:"model"`
+ }
+ if err := json.NewDecoder(r.Body).Decode(&body); err != nil {
+ t.Errorf("decode request: %v", err)
+ }
+ if body.Model != "deepseek-ai/DeepSeek-V4.1-Flash" {
+ t.Errorf("model=%q", body.Model)
+ }
+ _ = json.NewEncoder(w).Encode(map[string]any{
+ "id": "chat-runcrate",
+ "choices": []map[string]any{{"message": map[string]any{"content": "pong"}}},
+ "usage": map[string]any{"prompt_tokens": 3, "completion_tokens": 5, "total_tokens": 8},
+ })
+ }))
+ defer srv.Close()
+
+ usage := &common.ModelUsage{}
+ resp, err := newRuncrateForTest(srv.URL).ChatWithMessages(
+ t.Context(),
+ "deepseek-ai/DeepSeek-V4.1-Flash",
+ []Message{{Role: "user", Content: "ping"}},
+ &APIConfig{ApiKey: &testAPIKey},
+ nil,
+ usage,
+ )
+ if err != nil {
+ t.Fatalf("ChatWithMessages: %v", err)
+ }
+ if *resp.Answer != "pong" {
+ t.Errorf("Answer=%q", *resp.Answer)
+ }
+ assertModelUsage(t, usage, 3, 5, 8)
+}
+
+// TestRuncrateEmbed tests embedding requests.
+func TestRuncrateEmbed(t *testing.T) {
+ withSSRFBypass(t)
+ srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+ if r.URL.Path != "/embeddings" {
+ t.Errorf("path=%s", r.URL.Path)
+ }
+ _ = json.NewEncoder(w).Encode(map[string]any{
+ "data": []map[string]any{{"embedding": []float64{0.1, 0.2}, "index": 0}},
+ "usage": map[string]any{"prompt_tokens": 7, "total_tokens": 7},
+ })
+ }))
+ defer srv.Close()
+
+ modelName := "Qwen/Qwen3-Embedding-8B"
+ embeddings, err := newRuncrateForTest(srv.URL).Embed(
+ t.Context(),
+ &modelName,
+ EmbedRequest{Texts: []string{"document"}},
+ &APIConfig{ApiKey: &testAPIKey},
+ nil,
+ &common.ModelUsage{},
+ )
+ if err != nil {
+ t.Fatalf("Embed: %v", err)
+ }
+ if len(embeddings) != 1 || len(embeddings[0].Embedding) != 2 {
+ t.Fatalf("embeddings=%#v", embeddings)
+ }
+}
+
+// TestRuncrateListModelsAndCheckConnection tests model listing and connectivity check.
+func TestRuncrateListModelsAndCheckConnection(t *testing.T) {
+ withSSRFBypass(t)
+ srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+ if r.Method != http.MethodGet || r.URL.Path != "/models" {
+ t.Errorf("%s %s", r.Method, r.URL.Path)
+ }
+ _ = json.NewEncoder(w).Encode(map[string]any{
+ "data": []map[string]any{
+ {"id": "deepseek-ai/DeepSeek-V4.1-Flash"},
+ {"id": "Qwen/Qwen3-Embedding-8B"},
+ },
+ })
+ }))
+ defer srv.Close()
+
+ model := newRuncrateForTest(srv.URL)
+ list, err := model.ListModels(t.Context(), &APIConfig{ApiKey: &testAPIKey})
+ if err != nil {
+ t.Fatalf("ListModels: %v", err)
+ }
+ if joinModelNames(list, ",") != "deepseek-ai/DeepSeek-V4.1-Flash,Qwen/Qwen3-Embedding-8B" {
+ t.Errorf("models=%v", list)
+ }
+ if err := model.CheckConnection(t.Context(), &APIConfig{ApiKey: &testAPIKey}); err != nil {
+ t.Fatalf("CheckConnection: %v", err)
+ }
+}
diff --git a/web/src/assets/svg/llm/runcrate.svg b/web/src/assets/svg/llm/runcrate.svg
new file mode 100644
index 0000000000..dd2dff8681
--- /dev/null
+++ b/web/src/assets/svg/llm/runcrate.svg
@@ -0,0 +1 @@
+<svg xmlns="http://www.w3.org/2000/svg" width="48" height="48" viewBox="0 0 24 24"><rect x="1" y="1" width="22" height="22" rx="6" fill="#1A2B78"/><path d="M5.793 9.207A1 1 0 0 1 7.207 7.793L11.414 12 7.207 16.207A1 1 0 0 1 5.793 14.793L8.586 12ZM13 14.5h4.5a1 1 0 0 1 0 2H13a1 1 0 0 1 0-2Z" fill="#5F84FF"/></svg>
diff --git a/web/src/components/svg-icon.tsx b/web/src/components/svg-icon.tsx
index 7898e1ab17..0775a9601c 100644
--- a/web/src/components/svg-icon.tsx
+++ b/web/src/components/svg-icon.tsx
@@ -121,6 +121,7 @@ const svgIcons = [
LLMFactory.TrustedRouter,
LLMFactory.EmpirioLabs,
LLMFactory.AIStupidLevel,
+ LLMFactory.Runcrate,
];
export const LlmIcon = ({
diff --git a/web/src/constants/llm.ts b/web/src/constants/llm.ts
index 77e28d369c..42402aeed4 100644
--- a/web/src/constants/llm.ts
+++ b/web/src/constants/llm.ts
@@ -116,6 +116,7 @@ export enum LLMFactory {
TrustedRouter = 'TrustedRouter',
EmpirioLabs = 'EmpirioLabs AI',
AIStupidLevel = 'AI Stupid Level',
+ Runcrate = 'Runcrate',
}
// Please lowercase the file name
@@ -214,6 +215,7 @@ export const IconMap = {
[LLMFactory.TrustedRouter]: 'trustedrouter',
[LLMFactory.EmpirioLabs]: 'empiriolabs',
[LLMFactory.AIStupidLevel]: 'aistupidlevel',
+ [LLMFactory.Runcrate]: 'runcrate',
};
/**
@@ -264,6 +266,8 @@ export const APIMapUrl = {
[LLMFactory.EmpirioLabs]:
'https://platform.empiriolabs.ai/dashboard/api-keys',
[LLMFactory.AIStupidLevel]: 'https://aistupidlevel.info/router/keys',
+ [LLMFactory.Runcrate]:
+ 'https://www.runcrate.ai/dashboard/api-keys?utm_source=ragflow&utm_medium=directory&utm_campaign=gpu_pricing',
[LLMFactory.MWS]:
'https://mws.ru/docs/cloud-platform/gpt/general/inference-text.html',
[LLMFactory.Anthropic]: 'https://console.anthropic.com/settings/keys',