feat: implement OpenAI compatible API provider support (#500)

* feat: implement OpenAI compatible API provider support

- Add WHODB_OPENAI_COMPATIBLE_LABEL environment variable for custom labeling
- Create OpenAI compatible provider when API key and endpoint are configured
- Add OpenAI-Compatible LLM type with request/response handling
- Support custom models from WHODB_CUSTOM_MODELS environment variable
- Reuse existing OpenAI-compatible API structure without requiring new API code

Fixes #496

Co-authored-by: modelorona <modelorona@users.noreply.github.com>

* fix: handle SSE format in OpenAI-compatible streaming responses

- Strip "data: " prefix from SSE lines before JSON parsing
- Handle "[DONE]" control message to terminate stream properly
- Fixes JSON unmarshaling errors in streaming responses

Co-authored-by: modelorona <modelorona@users.noreply.github.com>

* fix: prevent truncated streaming responses in OpenAI-compatible client

- Send accumulated response before breaking on "[DONE]" message
- Ensure complete response is returned when stream ends normally
- Fixes issue where last part of response could be lost

Co-authored-by: modelorona <modelorona@users.noreply.github.com>

* add openai compatible url func

* move openai compatible to chatgpt logic

---------

Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com>
Co-authored-by: modelorona <modelorona@users.noreply.github.com>
Co-authored-by: modelorona <angel.hristozov@gmail.com>

Author: 3 people

.gitignore

@@ -1,3 +1,4 @@
 /.idea
 /.vscode
 /dev/elasticsearch
+DS_Store

core/src/env/env.go

@@ -42,6 +42,7 @@ var OpenAIEndpoint = os.Getenv("WHODB_OPENAI_ENDPOINT")
 
 var OpenAICompatibleEndpoint = os.Getenv("WHODB_OPENAI_COMPATIBLE_ENDPOINT")
 var OpenAICompatibleAPIKey = os.Getenv("WHODB_OPENAI_COMPATIBLE_API_KEY")
+var OpenAICompatibleLabel = os.Getenv("WHODB_OPENAI_COMPATIBLE_LABEL")
 
 var CustomModels = common.FilterList(strings.Split(os.Getenv("WHODB_CUSTOM_MODELS"), ","), func(item string) bool {
 	return strings.TrimSpace(item) != ""
@@ -58,6 +59,7 @@ type ChatProvider struct {
 	ProviderId string
 }
 
+// TODO: need to make this more dynamic so users can configure more than one key for each provider
 func GetConfiguredChatProviders() []ChatProvider {
 	providers := []ChatProvider{}
 
@@ -79,6 +81,19 @@ func GetConfiguredChatProviders() []ChatProvider {
 		})
 	}
 
+	if len(OpenAICompatibleAPIKey) > 0 && len(OpenAICompatibleEndpoint) > 0 {
+		label := OpenAICompatibleLabel
+		if label == "" {
+			label = "OpenAI-Compatible API"
+		}
+		providers = append(providers, ChatProvider{
+			Type:       "OpenAI-Compatible",
+			APIKey:     OpenAICompatibleAPIKey,
+			Endpoint:   GetOpenAICompatibleEndpoint(),
+			ProviderId: "openai-compatible-1",
+		})
+	}
+
 	providers = append(providers, ChatProvider{
 		Type:       "Ollama",
 		APIKey:     "",
@@ -115,16 +130,19 @@ func GetAnthropicEndpoint() string {
 }
 
 func GetOpenAIEndpoint() string {
-	// If the OpenAI compatible endpoint is set, use it. Otherwise, use the OpenAI endpoint.
-	if OpenAICompatibleEndpoint != "" {
-		return OpenAICompatibleEndpoint
-	}
 	if OpenAIEndpoint != "" {
 		return OpenAIEndpoint
 	}
 	return "https://api.openai.com/v1"
 }
 
+func GetOpenAICompatibleEndpoint() string {
+	if OpenAICompatibleEndpoint != "" && OpenAICompatibleAPIKey != "" {
+		return OpenAICompatibleEndpoint
+	}
+	return "https://api.openai.com/v1"
+}
+
 func GetClideyQuickContainerImage() string {
 	image := os.Getenv("CLIDEY_QUICK_CONTAINER_IMAGE")
 	if len(image) == 0 {

core/src/llm/chatgpt_client.go

@@ -27,7 +27,7 @@ import (
 	"github.com/clidey/whodb/core/src/env"
 )
 
-func prepareChatGPTRequest(c *LLMClient, prompt string, model LLMModel, receiverChan *chan string) (string, []byte, map[string]string, error) {
+func prepareChatGPTRequest(c *LLMClient, prompt string, model LLMModel, receiverChan *chan string, isOpenAICompatible bool) (string, []byte, map[string]string, error) {
 	requestBody, err := json.Marshal(map[string]interface{}{
 		"model":    string(model),
 		"messages": []map[string]string{{"role": "user", "content": prompt}},
@@ -37,6 +37,9 @@ func prepareChatGPTRequest(c *LLMClient, prompt string, model LLMModel, receiver
 		return "", nil, nil, err
 	}
 	url := fmt.Sprintf("%v/chat/completions", env.GetOpenAIEndpoint())
+	if isOpenAICompatible {
+		url = fmt.Sprintf("%v/chat/completions", env.GetOpenAICompatibleEndpoint())
+	}
 	headers := map[string]string{
 		"Authorization": fmt.Sprintf("Bearer %s", c.APIKey),
 		"Content-Type":  "application/json",

core/src/llm/llm_client.go

@@ -26,9 +26,10 @@ import (
 type LLMType string
 
 const (
-	Ollama_LLMType    LLMType = "Ollama"
-	ChatGPT_LLMType   LLMType = "ChatGPT"
-	Anthropic_LLMType LLMType = "Anthropic"
+	Ollama_LLMType           LLMType = "Ollama"
+	ChatGPT_LLMType          LLMType = "ChatGPT"
+	Anthropic_LLMType        LLMType = "Anthropic"
+	OpenAICompatible_LLMType LLMType = "OpenAI-Compatible"
 )
 
 type LLMModel string
@@ -49,9 +50,11 @@ func (c *LLMClient) Complete(prompt string, model LLMModel, receiverChan *chan s
 	case Ollama_LLMType:
 		url, requestBody, headers, err = prepareOllamaRequest(prompt, model)
 	case ChatGPT_LLMType:
-		url, requestBody, headers, err = prepareChatGPTRequest(c, prompt, model, receiverChan)
+		url, requestBody, headers, err = prepareChatGPTRequest(c, prompt, model, receiverChan, false)
 	case Anthropic_LLMType:
 		url, requestBody, headers, err = prepareAnthropicRequest(c, prompt, model)
+	case OpenAICompatible_LLMType:
+		url, requestBody, headers, err = prepareChatGPTRequest(c, prompt, model, receiverChan, true)
 	default:
 		return nil, errors.New("unsupported LLM type")
 	}
@@ -88,6 +91,8 @@ func (c *LLMClient) GetSupportedModels() ([]string, error) {
 		url, headers = prepareChatGPTModelsRequest(c.APIKey)
 	case Anthropic_LLMType:
 		return getAnthropicModels(c.APIKey)
+	case OpenAICompatible_LLMType:
+		return getOpenAICompatibleModels()
 	default:
 		return nil, errors.New("unsupported LLM type")
 	}
@@ -118,6 +123,8 @@ func (c *LLMClient) parseResponse(body io.ReadCloser, receiverChan *chan string)
 		return parseChatGPTResponse(body, receiverChan, &responseBuilder)
 	case Anthropic_LLMType:
 		return parseAnthropicResponse(body, receiverChan, &responseBuilder)
+	case OpenAICompatible_LLMType:
+		return parseChatGPTResponse(body, receiverChan, &responseBuilder)
 	default:
 		return nil, errors.New("unsupported LLM type")
 	}

core/src/llm/openai_compatible_client.go

@@ -0,0 +1,28 @@
+/*
+ * Copyright 2025 Clidey, Inc.
+ *
+ * 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 llm
+
+import (
+	"github.com/clidey/whodb/core/src/env"
+)
+
+func getOpenAICompatibleModels() ([]string, error) {
+	if len(env.CustomModels) > 0 {
+		return env.CustomModels, nil
+	}
+	return []string{}, nil
+}