feat(server): improve embedding & rerank speed (#12666)

fix AI-109

packages/backend/server/src/plugins/copilot/providers/openai.ts

@@ -103,7 +103,7 @@ export class OpenAIProvider extends CopilotProvider<OpenAIConfig> {
       capabilities: [
         {
           input: [ModelInputType.Text, ModelInputType.Image],
-          output: [ModelOutputType.Text],
+          output: [ModelOutputType.Text, ModelOutputType.Structured],
           defaultForOutputType: true,
         },
       ],
@@ -113,7 +113,7 @@ export class OpenAIProvider extends CopilotProvider<OpenAIConfig> {
       capabilities: [
         {
           input: [ModelInputType.Text, ModelInputType.Image],
-          output: [ModelOutputType.Text],
+          output: [ModelOutputType.Text, ModelOutputType.Structured],
         },
       ],
     },
@@ -122,7 +122,16 @@ export class OpenAIProvider extends CopilotProvider<OpenAIConfig> {
       capabilities: [
         {
           input: [ModelInputType.Text, ModelInputType.Image],
-          output: [ModelOutputType.Text],
+          output: [ModelOutputType.Text, ModelOutputType.Structured],
+        },
+      ],
+    },
+    {
+      id: 'gpt-4.1-nano',
+      capabilities: [
+        {
+          input: [ModelInputType.Text, ModelInputType.Image],
+          output: [ModelOutputType.Text, ModelOutputType.Structured],
         },
       ],
     },
@@ -283,8 +292,8 @@ export class OpenAIProvider extends CopilotProvider<OpenAIConfig> {
         model: modelInstance,
         system,
         messages: msgs,
-        temperature: options.temperature || 0,
-        maxTokens: options.maxTokens || 4096,
+        temperature: options.temperature ?? 0,
+        maxTokens: options.maxTokens ?? 4096,
         providerOptions: {
           openai: this.getOpenAIOptions(options, model.id),
         },
@@ -322,10 +331,10 @@ export class OpenAIProvider extends CopilotProvider<OpenAIConfig> {
         model: modelInstance,
         system,
         messages: msgs,
-        frequencyPenalty: options.frequencyPenalty || 0,
-        presencePenalty: options.presencePenalty || 0,
-        temperature: options.temperature || 0,
-        maxTokens: options.maxTokens || 4096,
+        frequencyPenalty: options.frequencyPenalty ?? 0,
+        presencePenalty: options.presencePenalty ?? 0,
+        temperature: options.temperature ?? 0,
+        maxTokens: options.maxTokens ?? 4096,
         providerOptions: {
           openai: this.getOpenAIOptions(options, model.id),
         },
@@ -388,8 +397,9 @@ export class OpenAIProvider extends CopilotProvider<OpenAIConfig> {
         model: modelInstance,
         system,
         messages: msgs,
-        temperature: ('temperature' in options && options.temperature) || 0,
-        maxTokens: ('maxTokens' in options && options.maxTokens) || 4096,
+        temperature: options.temperature ?? 0,
+        maxTokens: options.maxTokens ?? 4096,
+        maxRetries: options.maxRetries ?? 3,
         schema,
         providerOptions: {
           openai: options.user ? { user: options.user } : {},

packages/backend/server/src/plugins/copilot/providers/perplexity.ts

@@ -124,8 +124,8 @@ export class PerplexityProvider extends CopilotProvider<PerplexityConfig> {
         model: modelInstance,
         system,
         messages: msgs,
-        temperature: options.temperature || 0,
-        maxTokens: options.maxTokens || 4096,
+        temperature: options.temperature ?? 0,
+        maxTokens: options.maxTokens ?? 4096,
         abortSignal: options.signal,
       });
 
@@ -164,8 +164,8 @@ export class PerplexityProvider extends CopilotProvider<PerplexityConfig> {
         model: modelInstance,
         system,
         messages: msgs,
-        temperature: options.temperature || 0,
-        maxTokens: options.maxTokens || 4096,
+        temperature: options.temperature ?? 0,
+        maxTokens: options.maxTokens ?? 4096,
         abortSignal: options.signal,
       });
 

packages/backend/server/src/plugins/copilot/providers/provider.ts

@@ -172,7 +172,7 @@ export abstract class CopilotProvider<C = any> {
   structure(
     _cond: ModelConditions,
     _messages: PromptMessage[],
-    _options: CopilotStructuredOptions
+    _options?: CopilotStructuredOptions
   ): Promise<string> {
     throw new CopilotProviderNotSupported({
       provider: this.type,
@@ -193,7 +193,7 @@ export abstract class CopilotProvider<C = any> {
 
   embedding(
     _model: ModelConditions,
-    _text: string,
+    _text: string | string[],
     _options?: CopilotEmbeddingOptions
   ): Promise<number[][]> {
     throw new CopilotProviderNotSupported({

packages/backend/server/src/plugins/copilot/providers/types.ts

@@ -61,6 +61,8 @@ export const PromptConfigStrictSchema = z.object({
   // params requirements
   requireContent: z.boolean().nullable().optional(),
   requireAttachment: z.boolean().nullable().optional(),
+  // structure output
+  maxRetries: z.number().nullable().optional(),
   // openai
   frequencyPenalty: z.number().nullable().optional(),
   presencePenalty: z.number().nullable().optional(),