byterover-cli 2.3.3 → 2.3.4

package/dist/agent/infra/http/internal-llm-http-service.d.ts

@@ -51,8 +51,6 @@ export declare class ByteRoverLlmHttpService {
      * Call ByteRover REST LLM service to generate content.
      *
      * Simple forward to remote REST API - delegates all formatting to backend.
-     * Supports both Gemini and Claude formats - the correct format is determined
-     * automatically based on the model name.
      *
      * Parameter structure differs by provider:
      * - Gemini: contents = Content[], config = GenerateContentConfig
@@ -60,27 +58,22 @@ export declare class ByteRoverLlmHttpService {
      *
      * @param contents - For Gemini: Content[]. For Claude: MessageCreateParamsNonStreaming (complete body)
      * @param config - For Gemini: GenerateContentConfig. For Claude: RequestOptions (optional HTTP options)
-     * @param model - Model to use (detects provider from model name)
      * @param executionMetadata - Optional execution metadata (mode, executionContext)
      * @returns Response in GenerateContentResponse format
      */
-    generateContent(contents: Content[] | MessageCreateParamsNonStreaming, config: GenerateContentConfig | RequestOptions, model: string, executionMetadata?: Record<string, unknown>): Promise<GenerateContentResponse>;
+    generateContent(contents: Content[] | MessageCreateParamsNonStreaming, config: GenerateContentConfig | RequestOptions, executionMetadata?: Record<string, unknown>): Promise<GenerateContentResponse>;
     /**
      * Call ByteRover REST LLM service to generate content with streaming.
      *
-     * Currently falls back to non-streaming endpoint since /api/llm/generate/stream
-     * doesn't exist on the backend yet. Extracts thinking/reasoning from the complete
-     * response and yields them as separate chunks.
-     *
-     * When backend streaming is available, this will use SSE for true streaming.
+     * Currently falls back to non-streaming endpoint. Extracts thinking/reasoning
+     * from the complete response and yields them as separate chunks.
      *
      * @param contents - For Gemini: Content[]. For Claude: MessageCreateParamsNonStreaming (complete body)
      * @param config - For Gemini: GenerateContentConfig. For Claude: RequestOptions (optional HTTP options)
-     * @param model - Model to use (detects provider from model name)
      * @param executionMetadata - Optional execution metadata (mode, executionContext)
      * @yields GenerateContentChunk objects as they are generated
      */
-    generateContentStream(contents: Content[] | MessageCreateParamsNonStreaming, config: GenerateContentConfig | RequestOptions, model: string, executionMetadata?: Record<string, unknown>): AsyncGenerator<GenerateContentChunk>;
+    generateContentStream(contents: Content[] | MessageCreateParamsNonStreaming, config: GenerateContentConfig | RequestOptions, executionMetadata?: Record<string, unknown>): AsyncGenerator<GenerateContentChunk>;
     /**
      * Call the ByteRover REST Generate endpoint.
      *
@@ -91,26 +84,6 @@ export declare class ByteRoverLlmHttpService {
      * @throws Error if the request fails
      */
     private callHttpGenerate;
-    /**
-     * Detect LLM provider from model identifier.
-     *
-     * Determines which provider (Claude or Gemini) to use based on the model name.
-     * Defaults to Gemini if the model doesn't match Claude patterns.
-     *
-     * @param model - Model identifier (e.g., 'claude-3-5-sonnet', 'gemini-2.5-flash')
-     * @returns Provider name: 'claude' or 'gemini'
-     */
-    private detectProviderFromModel;
-    /**
-     * Detect appropriate GCP region from model identifier.
-     *
-     * Routes Claude models to us-east5 and Gemini models to global.
-     * This ensures compatibility with the provider's available regions on Vertex AI.
-     *
-     * @param model - Model identifier (e.g., 'claude-3-5-sonnet', 'gemini-2.5-flash')
-     * @returns GCP region identifier ('us-east5' or 'global')
-     */
-    private detectRegionFromModel;
     /**
      * Extract content chunks from a complete response.
      *

package/dist/agent/infra/http/internal-llm-http-service.js

@@ -33,7 +33,7 @@ export class ByteRoverLlmHttpService {
         this.config = {
             apiBaseUrl: config.apiBaseUrl,
             projectId: config.projectId ?? 'byterover',
-            region: config.region ?? 'us-east1',
+            region: config.region ?? 'global',
             sessionKey: config.sessionKey,
             spaceId: config.spaceId,
             teamId: config.teamId,
@@ -44,8 +44,6 @@ export class ByteRoverLlmHttpService {
      * Call ByteRover REST LLM service to generate content.
      *
      * Simple forward to remote REST API - delegates all formatting to backend.
-     * Supports both Gemini and Claude formats - the correct format is determined
-     * automatically based on the model name.
      *
      * Parameter structure differs by provider:
      * - Gemini: contents = Content[], config = GenerateContentConfig
@@ -53,21 +51,18 @@ export class ByteRoverLlmHttpService {
      *
      * @param contents - For Gemini: Content[]. For Claude: MessageCreateParamsNonStreaming (complete body)
      * @param config - For Gemini: GenerateContentConfig. For Claude: RequestOptions (optional HTTP options)
-     * @param model - Model to use (detects provider from model name)
      * @param executionMetadata - Optional execution metadata (mode, executionContext)
      * @returns Response in GenerateContentResponse format
      */
-    async generateContent(contents, config, model, executionMetadata) {
+    async generateContent(contents, config, executionMetadata) {
         const request = {
             executionMetadata: JSON.stringify(executionMetadata ?? {}),
             params: {
                 config,
                 contents,
-                model,
             },
             project_id: typeof this.config.projectId === 'function' ? this.config.projectId() : this.config.projectId,
-            provider: this.detectProviderFromModel(model),
-            region: this.detectRegionFromModel(model),
+            region: this.config.region,
             spaceId: typeof this.config.spaceId === 'function' ? this.config.spaceId() : this.config.spaceId,
             teamId: typeof this.config.teamId === 'function' ? this.config.teamId() : this.config.teamId,
         };
@@ -76,22 +71,18 @@ export class ByteRoverLlmHttpService {
     /**
      * Call ByteRover REST LLM service to generate content with streaming.
      *
-     * Currently falls back to non-streaming endpoint since /api/llm/generate/stream
-     * doesn't exist on the backend yet. Extracts thinking/reasoning from the complete
-     * response and yields them as separate chunks.
-     *
-     * When backend streaming is available, this will use SSE for true streaming.
+     * Currently falls back to non-streaming endpoint. Extracts thinking/reasoning
+     * from the complete response and yields them as separate chunks.
      *
      * @param contents - For Gemini: Content[]. For Claude: MessageCreateParamsNonStreaming (complete body)
      * @param config - For Gemini: GenerateContentConfig. For Claude: RequestOptions (optional HTTP options)
-     * @param model - Model to use (detects provider from model name)
      * @param executionMetadata - Optional execution metadata (mode, executionContext)
      * @yields GenerateContentChunk objects as they are generated
      */
-    async *generateContentStream(contents, config, model, executionMetadata) {
+    async *generateContentStream(contents, config, executionMetadata) {
         // Fall back to non-streaming endpoint and simulate streaming
         // by extracting thinking from the complete response
-        const response = await this.generateContent(contents, config, model, executionMetadata);
+        const response = await this.generateContent(contents, config, executionMetadata);
         // Extract and yield thinking/reasoning chunks first
         yield* this.extractThinkingFromResponse(response);
         // Then yield the final content
@@ -115,30 +106,6 @@ export class ByteRoverLlmHttpService {
         });
         return httpResponse.data;
     }
-    /**
-     * Detect LLM provider from model identifier.
-     *
-     * Determines which provider (Claude or Gemini) to use based on the model name.
-     * Defaults to Gemini if the model doesn't match Claude patterns.
-     *
-     * @param model - Model identifier (e.g., 'claude-3-5-sonnet', 'gemini-2.5-flash')
-     * @returns Provider name: 'claude' or 'gemini'
-     */
-    detectProviderFromModel(model) {
-        return model.toLowerCase().startsWith('claude') ? 'claude' : 'gemini';
-    }
-    /**
-     * Detect appropriate GCP region from model identifier.
-     *
-     * Routes Claude models to us-east5 and Gemini models to global.
-     * This ensures compatibility with the provider's available regions on Vertex AI.
-     *
-     * @param model - Model identifier (e.g., 'claude-3-5-sonnet', 'gemini-2.5-flash')
-     * @returns GCP region identifier ('us-east5' or 'global')
-     */
-    detectRegionFromModel(model) {
-        return model.toLowerCase().startsWith('claude') ? 'us-east5' : 'global';
-    }
     /**
      * Extract content chunks from a complete response.
      *

package/dist/agent/infra/llm/generators/byterover-content-generator.js

@@ -87,7 +87,7 @@ export class ByteRoverContentGenerator {
         };
         // // Debug: Log taskId for tracking
         // appendFileSync('debug-taskid.log', `[${new Date().toISOString()}] taskId from request: "${request.taskId}"\n`)
-        const rawResponse = await this.httpService.generateContent(contents, config, this.config.model, executionMetadata);
+        const rawResponse = await this.httpService.generateContent(contents, config, executionMetadata);
         // Parse response to internal format
         const messages = this.formatter.parseResponse(rawResponse);
         const lastMessage = messages.at(-1);
@@ -143,7 +143,7 @@ export class ByteRoverContentGenerator {
         const contents = this.providerType === 'claude' ? genConfig : formattedMessages;
         const config = this.providerType === 'claude' ? {} : genConfig;
         // Stream from HTTP service
-        yield* this.httpService.generateContentStream(contents, config, this.config.model, executionMetadata);
+        yield* this.httpService.generateContentStream(contents, config, executionMetadata);
     }
     /**
      * Build Claude-specific generation configuration.

package/oclif.manifest.json

@@ -1636,5 +1636,5 @@
       ]
     }
   },
-  "version": "2.3.3"
+  "version": "2.3.4"
 }

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "byterover-cli",
   "description": "ByteRover's CLI",
-  "version": "2.3.3",
+  "version": "2.3.4",
   "author": "ByteRover",
   "bin": {
     "brv": "./bin/run.js"