byterover-cli 3.10.2 → 3.11.0

package/README.md

@@ -34,7 +34,7 @@ Or download our self-hosted PDF version of the paper [here](https://byterover.de
 - 🖥️ Interactive TUI with REPL interface (React/Ink)
 - 🧠 Context tree and knowledge storage management
 - 🔀 Git-like version control for the context tree (branch, commit, merge, push/pull)
-- 🤖 18 LLM providers (Anthropic, OpenAI, Google, Groq, Mistral, xAI, and more)
+- 🤖 20 LLM providers (Anthropic, OpenAI, Google, Groq, Mistral, xAI, DeepSeek, and more)
 - 🛠️ 24 built-in agent tools (code exec, file ops, knowledge search, memory management)
 - 🔄 Cloud sync with push/pull
 - 👀 Review workflow for curate operations (approve/reject pending changes)
@@ -220,7 +220,7 @@ Run `brv --help` for the full command reference.
 <details>
 <summary><h2>Supported LLM Providers</h2></summary>
 
-ByteRover CLI supports 18 LLM providers out of the box. Connect and switch providers from the dashboard, or use `brv providers connect` / `brv providers switch`.
+ByteRover CLI supports 20 LLM providers out of the box. Connect and switch providers from the dashboard, or use `brv providers connect` / `brv providers switch`.
 
 | Provider | Description |
 |----------|-------------|
@@ -233,6 +233,7 @@ ByteRover CLI supports 18 LLM providers out of the box. Connect and switch provi
 | Cerebras | Fast inference |
 | Cohere | Command models |
 | DeepInfra | Open-source model hosting |
+| DeepSeek | DeepSeek V3 and R1 reasoning models |
 | OpenRouter | Multi-provider gateway |
 | Perplexity | Search-augmented models |
 | TogetherAI | Open-source model hosting |
@@ -240,6 +241,7 @@ ByteRover CLI supports 18 LLM providers out of the box. Connect and switch provi
 | Minimax | Minimax models |
 | Moonshot | Kimi models |
 | GLM | GLM models |
+| GLM Coding Plan | GLM models on Z.AI Coding Plan subscription |
 | OpenAI-Compatible | Any OpenAI-compatible API |
 | ByteRover | ByteRover's hosted models |
 

package/dist/agent/core/domain/llm/registry.d.ts

@@ -119,3 +119,15 @@ export declare function resolveRegistryProvider(model: string, explicitProvider?
  * @returns true if provider accepts arbitrary models
  */
 export declare function acceptsAnyModel(provider: LLMProvider): boolean;
+/**
+ * Check whether a model accepts the sampling parameters `temperature`, `top_p`,
+ * and `top_k`.
+ *
+ * Some models (e.g. Claude Opus 4.7) reject any non-default value with a 400
+ * error — callers must omit these parameters entirely when this returns false.
+ *
+ * Handles both bare model ids (`claude-opus-4-7`) and OpenRouter-style prefixed
+ * ids (`anthropic/claude-opus-4-7`). Unknown models default to true so we don't
+ * regress arbitrary OpenRouter-routed models.
+ */
+export declare function modelAcceptsSamplingParameters(modelId: string): boolean;

package/dist/agent/core/domain/llm/registry.js

@@ -25,6 +25,21 @@ export const LLM_REGISTRY = {
         defaultModel: '',
         models: [
             // Claude 4.x series
+            {
+                capabilities: {
+                    acceptsSamplingParameters: false,
+                    supportsAudio: false,
+                    supportsImages: true,
+                    supportsPdf: true,
+                    supportsStreaming: true,
+                },
+                charsPerToken: 3.5,
+                displayName: 'Claude Opus 4.7',
+                maxInputTokens: 200_000,
+                maxOutputTokens: 128_000,
+                name: 'claude-opus-4-7',
+                supportedFileTypes: ['image', 'pdf'],
+            },
             {
                 capabilities: { supportsAudio: false, supportsImages: true, supportsPdf: true, supportsStreaming: true },
                 charsPerToken: 3.5,
@@ -684,3 +699,37 @@ export function acceptsAnyModel(provider) {
     // OpenAI provider type accepts arbitrary models (OpenRouter, direct OpenAI, xAI, etc.)
     return provider === 'openai';
 }
+/**
+ * Strip an OpenRouter-style `provider/` prefix from a model id.
+ * Returns the input unchanged if no slash is present.
+ */
+function stripRouterPrefix(modelId) {
+    const slash = modelId.lastIndexOf('/');
+    return slash === -1 ? modelId : modelId.slice(slash + 1);
+}
+/**
+ * Check whether a model accepts the sampling parameters `temperature`, `top_p`,
+ * and `top_k`.
+ *
+ * Some models (e.g. Claude Opus 4.7) reject any non-default value with a 400
+ * error — callers must omit these parameters entirely when this returns false.
+ *
+ * Handles both bare model ids (`claude-opus-4-7`) and OpenRouter-style prefixed
+ * ids (`anthropic/claude-opus-4-7`). Unknown models default to true so we don't
+ * regress arbitrary OpenRouter-routed models.
+ */
+export function modelAcceptsSamplingParameters(modelId) {
+    const bare = stripRouterPrefix(modelId);
+    for (const provider of PROVIDER_TYPES) {
+        const info = getModelInfo(provider, bare);
+        if (info) {
+            return info.capabilities.acceptsSamplingParameters !== false;
+        }
+    }
+    // Family-level fallback: catches date-suffixed snapshots not yet in the registry
+    // (e.g. `claude-opus-4-7-20260101`). Extend this list as new families deprecate
+    // sampling params.
+    if (bare.startsWith('claude-opus-4-7'))
+        return false;
+    return true;
+}

package/dist/agent/core/domain/llm/types.d.ts

@@ -34,6 +34,12 @@ export type SupportedFileType = (typeof SUPPORTED_FILE_TYPES)[number];
  * Defines what features a specific model supports.
  */
 export interface ModelCapabilities {
+    /**
+     * Whether the model accepts the sampling parameters `temperature`, `top_p`, and `top_k`.
+     * When false, callers must omit these from the request — Claude Opus 4.7 returns 400
+     * on any non-default value. Defaults to true when omitted.
+     */
+    acceptsSamplingParameters?: boolean;
     /** Whether the model supports audio input */
     supportsAudio: boolean;
     /** Whether the model supports image input */

package/dist/agent/core/interfaces/i-content-generator.d.ts

@@ -65,6 +65,14 @@ export interface GenerateContentResponse {
     finishReason: 'error' | 'max_tokens' | 'stop' | 'tool_calls';
     /** Raw response from provider (for debugging) */
     rawResponse?: unknown;
+    /**
+     * Reasoning / thinking text emitted by the model (e.g. DeepSeek-R1's
+     * `reasoning_content`, OpenAI o1's reasoning summary). Required to be
+     * passed back to the API on the next turn for some providers — DeepSeek-R1
+     * rejects the next call with "The reasoning_content in the thinking mode
+     * must be passed back to the API" if absent.
+     */
+    reasoning?: string;
     /** Tool calls requested by the model */
     toolCalls?: ToolCall[];
     /** Token usage statistics */

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

@@ -388,9 +388,13 @@ export class AgentLLMService {
     async callLLMAndParseResponse(request) {
         try {
             const response = await this.generator.generateContent(request);
-            // Convert response to InternalMessage format
+            // Convert response to InternalMessage format. The reasoning field must
+            // round-trip on the next turn for some providers (e.g. DeepSeek-R1
+            // rejects with "reasoning_content must be passed back to the API"
+            // otherwise).
             const message = {
                 content: response.content,
+                ...(response.reasoning && { reasoning: response.reasoning }),
                 role: 'assistant',
                 toolCalls: response.toolCalls,
             };
@@ -423,11 +427,15 @@ export class AgentLLMService {
     async callLLMAndParseResponseStreaming(request, taskId) {
         try {
             let accumulatedContent = '';
+            let accumulatedReasoning = '';
             let accumulatedToolCalls = [];
             // Stream chunks and accumulate content
             for await (const chunk of this.generator.generateContentStream(request)) {
-                // Emit thinking/reasoning chunks as events for TUI display
+                // Emit thinking/reasoning chunks as events for TUI display + accumulate
+                // for the InternalMessage so it round-trips on the next turn (DeepSeek-R1
+                // requires reasoning_content to be passed back).
                 if (chunk.type === StreamChunkType.THINKING && chunk.reasoning) {
+                    accumulatedReasoning += chunk.reasoning;
                     this.sessionEventBus.emit('llmservice:chunk', {
                         content: chunk.reasoning,
                         isComplete: chunk.isComplete,
@@ -454,6 +462,7 @@ export class AgentLLMService {
             // Convert accumulated response to InternalMessage format
             const message = {
                 content: accumulatedContent || null,
+                ...(accumulatedReasoning && { reasoning: accumulatedReasoning }),
                 role: 'assistant',
                 toolCalls: accumulatedToolCalls.length > 0 ? accumulatedToolCalls : undefined,
             };
@@ -960,8 +969,10 @@ export class AgentLLMService {
             provider: this.providerId,
             taskId: taskId || undefined,
         });
-        // Add assistant message to context
-        await this.contextManager.addAssistantMessage(content);
+        // Add assistant message to context. Pass reasoning so it round-trips to
+        // providers that demand it (DeepSeek-R1 rejects with "reasoning_content
+        // must be passed back to the API" otherwise).
+        await this.contextManager.addAssistantMessage(content, undefined, lastMessage.reasoning);
         return content;
     }
     /**
@@ -1084,9 +1095,10 @@ export class AgentLLMService {
         }
         // Emit thought events if present
         this.handleThoughts(lastMessage, taskId);
-        // Has tool calls - add assistant message with tool calls
+        // Has tool calls - add assistant message with tool calls. Pass reasoning
+        // so it round-trips to providers that demand it.
         const assistantContent = this.extractTextContent(lastMessage);
-        await this.contextManager.addAssistantMessage(assistantContent, lastMessage.toolCalls);
+        await this.contextManager.addAssistantMessage(assistantContent, lastMessage.toolCalls, lastMessage.reasoning);
         // Step 1: Create pending tool parts for all tool calls
         for (const toolCall of lastMessage.toolCalls) {
             const toolArgs = JSON.parse(toolCall.function.arguments);

package/dist/agent/infra/llm/context/context-manager.d.ts

@@ -140,8 +140,11 @@ export declare class ContextManager<T> {
      *
      * @param content - Message content (text or null if only tool calls)
      * @param toolCalls - Optional tool calls made by the assistant
+     * @param reasoning - Optional reasoning/thinking trace from the model.
+     *   Required to round-trip for providers like DeepSeek-R1 that reject
+     *   the next turn unless reasoning_content is replayed.
      */
-    addAssistantMessage(content: null | string, toolCalls?: InternalMessage['toolCalls']): Promise<void>;
+    addAssistantMessage(content: null | string, toolCalls?: InternalMessage['toolCalls'], reasoning?: string): Promise<void>;
     /**
      * Add a system message to the conversation.
      *

package/dist/agent/infra/llm/context/context-manager.js

@@ -83,10 +83,14 @@ export class ContextManager {
      *
      * @param content - Message content (text or null if only tool calls)
      * @param toolCalls - Optional tool calls made by the assistant
+     * @param reasoning - Optional reasoning/thinking trace from the model.
+     *   Required to round-trip for providers like DeepSeek-R1 that reject
+     *   the next turn unless reasoning_content is replayed.
      */
-    async addAssistantMessage(content, toolCalls) {
+    async addAssistantMessage(content, toolCalls, reasoning) {
         const message = {
             content,
+            ...(reasoning && { reasoning }),
             role: 'assistant',
             toolCalls,
         };

package/dist/agent/infra/llm/generators/ai-sdk-content-generator.d.ts

@@ -21,6 +21,12 @@ export declare function prependCachedSystemMessage(systemPrompt: string | undefi
 export interface AiSdkContentGeneratorConfig {
     /** Characters per token ratio for token estimation */
     charsPerToken?: number;
+    /**
+     * Drop the sampling request parameters (`temperature`, `top_p`, `top_k`)
+     * before calling the model. Set when targeting models that reject these
+     * (e.g. Claude Opus 4.7 returns 400 on any non-default value).
+     */
+    excludeSamplingParameters?: boolean;
     /** AI SDK LanguageModel instance */
     model: LanguageModel;
 }
@@ -34,11 +40,18 @@ export interface AiSdkContentGeneratorConfig {
  */
 export declare class AiSdkContentGenerator implements IContentGenerator {
     private readonly charsPerToken;
+    private readonly excludeSamplingParameters;
     private readonly model;
     constructor(config: AiSdkContentGeneratorConfig);
     estimateTokensSync(content: string): number;
     generateContent(request: GenerateContentRequest): Promise<GenerateContentResponse>;
     generateContentStream(request: GenerateContentRequest): AsyncGenerator<GenerateContentChunk>;
+    /**
+     * Build the sampling-parameter slice for an AI SDK request. Returns an
+     * empty object when the model-level exclusion is set, so callers can spread
+     * the result into the request payload without conditionally emitting fields.
+     */
+    private buildSamplingParams;
 }
 /**
  * Extract a human-readable message from an AI SDK stream error.

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

@@ -37,10 +37,12 @@ export function prependCachedSystemMessage(systemPrompt, messages) {
  */
 export class AiSdkContentGenerator {
     charsPerToken;
+    excludeSamplingParameters;
     model;
     constructor(config) {
         this.model = config.model;
         this.charsPerToken = config.charsPerToken ?? DEFAULT_CHARS_PER_TOKEN;
+        this.excludeSamplingParameters = config.excludeSamplingParameters ?? false;
     }
     estimateTokensSync(content) {
         return Math.ceil(content.length / this.charsPerToken);
@@ -53,10 +55,8 @@ export class AiSdkContentGenerator {
             maxRetries: 0, // RetryableContentGenerator handles retries
             messages,
             model: this.model,
-            temperature: request.config.temperature,
+            ...this.buildSamplingParams(request.config),
             ...(tools && { tools }),
-            ...(request.config.topK !== undefined && { topK: request.config.topK }),
-            ...(request.config.topP !== undefined && { topP: request.config.topP }),
         });
         // Map AI SDK tool calls to our ToolCall format
         // Preserve thoughtSignature from providerMetadata (required by Gemini 3+ models)
@@ -77,6 +77,7 @@ export class AiSdkContentGenerator {
             content: result.text,
             finishReason: mapFinishReason(result.finishReason, toolCalls.length > 0),
             rawResponse: result.response,
+            ...(result.reasoningText && { reasoning: result.reasoningText }),
             toolCalls: toolCalls.length > 0 ? toolCalls : undefined,
             usage: {
                 completionTokens: result.usage.outputTokens ?? 0,
@@ -93,10 +94,8 @@ export class AiSdkContentGenerator {
             maxRetries: 0,
             messages,
             model: this.model,
-            temperature: request.config.temperature,
+            ...this.buildSamplingParams(request.config),
             ...(tools && { tools }),
-            ...(request.config.topK !== undefined && { topK: request.config.topK }),
-            ...(request.config.topP !== undefined && { topP: request.config.topP }),
         });
         // Accumulate tool calls during streaming
         const pendingToolCalls = [];
@@ -155,6 +154,20 @@ export class AiSdkContentGenerator {
             }
         }
     }
+    /**
+     * Build the sampling-parameter slice for an AI SDK request. Returns an
+     * empty object when the model-level exclusion is set, so callers can spread
+     * the result into the request payload without conditionally emitting fields.
+     */
+    buildSamplingParams(config) {
+        if (this.excludeSamplingParameters)
+            return {};
+        return {
+            ...(config.temperature !== undefined && { temperature: config.temperature }),
+            ...(config.topK !== undefined && { topK: config.topK }),
+            ...(config.topP !== undefined && { topP: config.topP }),
+        };
+    }
 }
 /**
  * Extract a human-readable message from an AI SDK stream error.

package/dist/agent/infra/llm/generators/ai-sdk-message-converter.js

@@ -129,19 +129,31 @@ function convertUserMessage(msg) {
 }
 /**
  * Convert an internal assistant message to AI SDK format.
- * Handles text content and tool calls.
+ * Handles reasoning, text content, and tool calls.
+ *
+ * The reasoning part is required when the message is replayed to providers
+ * that demand the previous turn's thinking trace round-trip back — DeepSeek-R1
+ * rejects requests with "The reasoning_content in the thinking mode must be
+ * passed back to the API" if the assistant message in history lacks the
+ * reasoning that was emitted on the prior turn.
  */
 function convertAssistantMessage(msg) {
     const textContent = extractTextContent(msg);
     const hasToolCalls = msg.toolCalls && msg.toolCalls.length > 0;
-    if (!textContent && !hasToolCalls) {
+    const hasReasoning = Boolean(msg.reasoning);
+    if (!textContent && !hasToolCalls && !hasReasoning) {
         return undefined;
     }
-    // Simple text-only case
-    if (textContent && !hasToolCalls) {
+    // Simple text-only case (no reasoning, no tools)
+    if (textContent && !hasToolCalls && !hasReasoning) {
         return { content: textContent, role: 'assistant' };
     }
     const parts = [];
+    // Reasoning must come first — providers that consume it expect it at the
+    // start of the assistant turn, before any text/tool-call output.
+    if (msg.reasoning) {
+        parts.push({ text: msg.reasoning, type: 'reasoning' });
+    }
     if (textContent) {
         parts.push({ text: textContent, type: 'text' });
     }

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

@@ -31,6 +31,7 @@ export interface ByteRoverContentGeneratorConfig {
  * - Response parsing to unified format
  */
 export declare class ByteRoverContentGenerator implements IContentGenerator {
+    private readonly acceptsSamplingParameters;
     private readonly config;
     private readonly formatter;
     private readonly httpService;

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

@@ -5,6 +5,7 @@
  * Supports both Claude and Gemini models through the unified HTTP interface.
  */
 import { FunctionCallingConfigMode } from '@google/genai';
+import { modelAcceptsSamplingParameters } from '../../../core/domain/llm/registry.js';
 import { ClaudeMessageFormatter } from '../formatters/claude-formatter.js';
 import { ensureActiveLoopHasThoughtSignatures, GeminiMessageFormatter } from '../formatters/gemini-formatter.js';
 import { ThinkingConfigManager } from '../thought-parser.js';
@@ -21,6 +22,7 @@ import { GeminiTokenizer } from '../tokenizers/gemini-tokenizer.js';
  * - Response parsing to unified format
  */
 export class ByteRoverContentGenerator {
+    acceptsSamplingParameters;
     config;
     formatter;
     httpService;
@@ -40,6 +42,7 @@ export class ByteRoverContentGenerator {
             temperature: config.temperature ?? 0.7,
             thinkingConfig: config.thinkingConfig,
         };
+        this.acceptsSamplingParameters = modelAcceptsSamplingParameters(this.config.model);
         // Detect provider type from model name
         this.providerType = this.detectProviderType(this.config.model);
         // Initialize formatter and tokenizer based on provider type
@@ -160,7 +163,7 @@ export class ByteRoverContentGenerator {
             messages,
             model: this.config.model,
             system: systemPrompt,
-            temperature: this.config.temperature,
+            ...(this.acceptsSamplingParameters && { temperature: this.config.temperature }),
             ...(claudeTools.length > 0 && { tools: claudeTools }),
         };
         /* eslint-enable camelcase */

package/dist/agent/infra/llm/model-capabilities.d.ts

@@ -9,7 +9,8 @@
  * - Grok: `reasoning_content` or `reasoning_details` fields
  * - Gemini via OpenRouter: `reasoning_details` array or `thoughts` field
  * - GLM (Zhipu AI): `reasoning_content` field in API response
- * - Claude/DeepSeek/MiniMax: `<think>...</think>` XML tags in content
+ * - DeepSeek (R1/Reasoner): `reasoning_content` field in API response (OpenAI-compatible)
+ * - Claude/MiniMax: `<think>...</think>` XML tags in content
  */
 /**
  * Reasoning format types

package/dist/agent/infra/llm/model-capabilities.js

@@ -9,7 +9,8 @@
  * - Grok: `reasoning_content` or `reasoning_details` fields
  * - Gemini via OpenRouter: `reasoning_details` array or `thoughts` field
  * - GLM (Zhipu AI): `reasoning_content` field in API response
- * - Claude/DeepSeek/MiniMax: `<think>...</think>` XML tags in content
+ * - DeepSeek (R1/Reasoner): `reasoning_content` field in API response (OpenAI-compatible)
+ * - Claude/MiniMax: `<think>...</think>` XML tags in content
  */
 /**
  * Get model capabilities for a given model ID.
@@ -95,13 +96,14 @@ export function getModelCapabilities(modelId) {
             reasoningFormat: 'think-tags',
         };
     }
-    // DeepSeek models use think tags
+    // DeepSeek models — reasoning models stream `reasoning_content` natively
+    // (OpenAI-compatible field), not <think> tags.
     if (id.includes('deepseek')) {
-        // DeepSeek-R1 and reasoning models
         if (id.includes('r1') || id.includes('reasoner')) {
             return {
                 reasoning: true,
-                reasoningFormat: 'think-tags',
+                reasoningField: 'reasoning_content',
+                reasoningFormat: 'native-field',
             };
         }
         return {

package/dist/agent/infra/llm/providers/anthropic.js

@@ -4,6 +4,7 @@
  * Direct access to Claude models via @ai-sdk/anthropic.
  */
 import { createAnthropic } from '@ai-sdk/anthropic';
+import { modelAcceptsSamplingParameters } from '../../../core/domain/llm/registry.js';
 import { AiSdkContentGenerator } from '../generators/ai-sdk-content-generator.js';
 export const anthropicProvider = {
     apiKeyUrl: 'https://console.anthropic.com/settings/keys',
@@ -14,6 +15,7 @@ export const anthropicProvider = {
         const provider = createAnthropic({ apiKey: config.apiKey });
         return new AiSdkContentGenerator({
             charsPerToken: 3.5,
+            excludeSamplingParameters: !modelAcceptsSamplingParameters(config.model),
             model: provider(config.model),
         });
     },

package/dist/agent/infra/llm/providers/deepseek.d.ts

@@ -0,0 +1,10 @@
+/**
+ * DeepSeek Provider Module
+ *
+ * Access to DeepSeek V3 (deepseek-chat) and R1 (deepseek-reasoner) via their
+ * OpenAI-compatible API. The reasoner model streams thinking through the
+ * native `reasoning_content` field rather than `<think>` tags — see
+ * model-capabilities.ts for the parser routing.
+ */
+import type { ProviderModule } from './types.js';
+export declare const deepseekProvider: ProviderModule;

package/dist/agent/infra/llm/providers/deepseek.js

@@ -0,0 +1,33 @@
+/**
+ * DeepSeek Provider Module
+ *
+ * Access to DeepSeek V3 (deepseek-chat) and R1 (deepseek-reasoner) via their
+ * OpenAI-compatible API. The reasoner model streams thinking through the
+ * native `reasoning_content` field rather than `<think>` tags — see
+ * model-capabilities.ts for the parser routing.
+ */
+import { createOpenAICompatible } from '@ai-sdk/openai-compatible';
+import { AiSdkContentGenerator } from '../generators/ai-sdk-content-generator.js';
+export const deepseekProvider = {
+    apiKeyUrl: 'https://platform.deepseek.com/api_keys',
+    authType: 'api-key',
+    baseUrl: 'https://api.deepseek.com/v1',
+    category: 'other',
+    createGenerator(config) {
+        const provider = createOpenAICompatible({
+            apiKey: config.apiKey,
+            baseURL: 'https://api.deepseek.com/v1',
+            name: 'deepseek',
+        });
+        return new AiSdkContentGenerator({
+            model: provider.chatModel(config.model),
+        });
+    },
+    defaultModel: 'deepseek-chat',
+    description: 'DeepSeek V3 and R1 reasoning models',
+    envVars: ['DEEPSEEK_API_KEY'],
+    id: 'deepseek',
+    name: 'DeepSeek',
+    priority: 19,
+    providerType: 'openai',
+};

package/dist/agent/infra/llm/providers/glm-coding-plan.d.ts

@@ -0,0 +1,9 @@
+/**
+ * GLM Coding Plan (Z.AI) Provider Module
+ *
+ * Same Z.AI account as the standard `glm` provider but routes through the
+ * coding-plan endpoint so subscription quota is consumed instead of
+ * pay-per-token billing.
+ */
+import type { ProviderModule } from './types.js';
+export declare const glmCodingPlanProvider: ProviderModule;

package/dist/agent/infra/llm/providers/glm-coding-plan.js

@@ -0,0 +1,32 @@
+/**
+ * GLM Coding Plan (Z.AI) Provider Module
+ *
+ * Same Z.AI account as the standard `glm` provider but routes through the
+ * coding-plan endpoint so subscription quota is consumed instead of
+ * pay-per-token billing.
+ */
+import { createOpenAICompatible } from '@ai-sdk/openai-compatible';
+import { AiSdkContentGenerator } from '../generators/ai-sdk-content-generator.js';
+export const glmCodingPlanProvider = {
+    apiKeyUrl: 'https://z.ai/manage-apikey/apikey-list',
+    authType: 'api-key',
+    baseUrl: 'https://api.z.ai/api/coding/paas/v4',
+    category: 'other',
+    createGenerator(config) {
+        const provider = createOpenAICompatible({
+            apiKey: config.apiKey,
+            baseURL: 'https://api.z.ai/api/coding/paas/v4',
+            name: 'glm-coding-plan',
+        });
+        return new AiSdkContentGenerator({
+            model: provider.chatModel(config.model),
+        });
+    },
+    defaultModel: 'glm-4.7',
+    description: 'GLM models on the Z.AI Coding Plan subscription',
+    envVars: ['ZHIPU_API_KEY'],
+    id: 'glm-coding-plan',
+    name: 'GLM Coding Plan (Z.AI)',
+    priority: 17.5,
+    providerType: 'openai',
+};

package/dist/agent/infra/llm/providers/index.js

@@ -10,6 +10,8 @@ import { byteroverProvider } from './byterover.js';
 import { cerebrasProvider } from './cerebras.js';
 import { cohereProvider } from './cohere.js';
 import { deepinfraProvider } from './deepinfra.js';
+import { deepseekProvider } from './deepseek.js';
+import { glmCodingPlanProvider } from './glm-coding-plan.js';
 import { glmProvider } from './glm.js';
 import { googleProvider } from './google.js';
 import { groqProvider } from './groq.js';
@@ -33,7 +35,9 @@ const PROVIDER_MODULES = {
     cerebras: cerebrasProvider,
     cohere: cohereProvider,
     deepinfra: deepinfraProvider,
+    deepseek: deepseekProvider,
     glm: glmProvider,
+    'glm-coding-plan': glmCodingPlanProvider,
     google: googleProvider,
     groq: groqProvider,
     minimax: minimaxProvider,

package/dist/agent/infra/llm/providers/openrouter.js

@@ -4,6 +4,7 @@
  * Access 200+ models via the OpenRouter aggregator using @openrouter/ai-sdk-provider.
  */
 import { createOpenRouter } from '@openrouter/ai-sdk-provider';
+import { modelAcceptsSamplingParameters } from '../../../core/domain/llm/registry.js';
 import { AiSdkContentGenerator } from '../generators/ai-sdk-content-generator.js';
 export const openrouterProvider = {
     apiKeyUrl: 'https://openrouter.ai/keys',
@@ -13,6 +14,7 @@ export const openrouterProvider = {
     createGenerator(config) {
         const provider = createOpenRouter({ apiKey: config.apiKey });
         return new AiSdkContentGenerator({
+            excludeSamplingParameters: !modelAcceptsSamplingParameters(config.model),
             model: provider.chat(config.model),
         });
     },

package/dist/oclif/commands/query.js

@@ -110,7 +110,7 @@ Bad:
             client,
             command: 'query',
             format,
-            onCompleted: ({ result, taskId: tid }) => {
+            onCompleted: ({ durationMs, matchedDocs, result, taskId: tid, tier, topScore }) => {
                 const previousResult = finalResult;
                 // Always prefer the completed payload — it carries the attribution footer
                 // that may not be present in the earlier llmservice:response event.
@@ -133,11 +133,17 @@ Bad:
                 if (format === 'json') {
                     writeJsonResponse({
                         command: 'query',
+                        // Recall metadata is only present on query tasks; older daemons omit it. Spread
+                        // conditionally so JSON consumers do not see undefined keys.
                         data: {
+                            ...(durationMs === undefined ? {} : { durationMs }),
                             event: 'completed',
+                            ...(matchedDocs === undefined ? {} : { matchedDocs }),
                             result: finalResult,
                             status: 'completed',
                             taskId: tid,
+                            ...(tier === undefined ? {} : { tier }),
+                            ...(topScore === undefined ? {} : { topScore }),
                         },
                         success: true,
                     });