codeep 2.12.0 → 2.13.0

package/dist/config/providers.d.ts

@@ -26,6 +26,10 @@ export interface ProviderConfig {
     maxOutputTokens?: number;
     useMaxCompletionTokens?: boolean;
     requiresDefaultTemperature?: boolean;
+    /** Provider's OpenAI-compatible endpoint rejects `tools` together with
+     *  `stream: true` (Alibaba/Qwen DashScope). When true, agent turns that send
+     *  tools are issued non-streamed (we buffer the full response). */
+    noStreamWithTools?: boolean;
     envKey?: string;
     subscribeUrl?: string;
     noApiKey?: boolean;
@@ -68,6 +72,12 @@ export declare function usesMaxCompletionTokens(providerId: string): boolean;
  * (e.g. OpenAI GPT-5+ only accepts the default of 1).
  */
 export declare function requiresDefaultTemperature(providerId: string): boolean;
+/**
+ * Returns true if the provider's OpenAI-compatible endpoint rejects `tools`
+ * together with `stream: true` (Alibaba/Qwen) — callers must issue tool-bearing
+ * agent turns non-streamed.
+ */
+export declare function providerNoStreamWithTools(providerId: string): boolean;
 export declare function modelRejectsSamplingParams(model: string): boolean;
 /**
  * Returns the effective max output tokens for a provider, capped by the provider's limit.

package/dist/config/providers.js

@@ -203,6 +203,196 @@ export const PROVIDERS = {
         groupLabel: 'DeepSeek',
         hint: 'Pay-per-use via DeepSeek API key (platform.deepseek.com).',
     },
+    // ── Kimi (Moonshot AI) ────────────────────────────────────────────
+    // Subscription (Kimi Code) mirrors the Z.AI GLM-Coding-Plan shape: a
+    // dedicated coding base URL + a separate key, model id ALWAYS
+    // `kimi-for-coding` (a backend alias). OpenAI-compatible is the
+    // battle-tested path so we don't expose the Anthropic surface here.
+    'kimi': {
+        name: 'Kimi (Moonshot) — Coding Plan',
+        description: 'Kimi Code subscription',
+        protocols: {
+            openai: { baseUrl: 'https://api.kimi.com/coding/v1', authHeader: 'Bearer', supportsNativeTools: true },
+        },
+        models: [
+            { id: 'kimi-for-coding', name: 'Kimi Code', description: 'Subscription alias — auto-maps to the latest Kimi coding model (K2.7 Code)' },
+        ],
+        defaultModel: 'kimi-for-coding',
+        defaultProtocol: 'openai',
+        maxOutputTokens: 32_768,
+        envKey: 'KIMI_CODE_API_KEY',
+        subscribeUrl: 'https://www.kimi.com/code',
+        groupLabel: 'Kimi — Subscription (Kimi Code)',
+        hint: 'Uses your Kimi Code subscription — no per-token charges. Key from kimi.com/code/console.',
+    },
+    'kimi-api': {
+        name: 'Kimi (Moonshot) API (pay-per-use)',
+        description: 'Moonshot AI Kimi models via API key',
+        protocols: {
+            openai: { baseUrl: 'https://api.moonshot.ai/v1', authHeader: 'Bearer', supportsNativeTools: true },
+        },
+        models: [
+            { id: 'kimi-k2.7-code', name: 'Kimi K2.7 Code', description: 'Flagship agentic coding model (256K context)' },
+            { id: 'kimi-k2.7-code-highspeed', name: 'Kimi K2.7 Code (High-Speed)', description: 'Throughput-tuned K2.7 Code for latency-sensitive loops' },
+            { id: 'kimi-k2.6', name: 'Kimi K2.6', description: 'Previous-gen multimodal reasoning model' },
+            { id: 'kimi-k2.5', name: 'Kimi K2.5', description: 'Older general-purpose model (cheaper)' },
+        ],
+        defaultModel: 'kimi-k2.7-code',
+        defaultProtocol: 'openai',
+        maxOutputTokens: 32_768,
+        envKey: 'MOONSHOT_API_KEY',
+        subscribeUrl: 'https://platform.kimi.ai/console/api-keys',
+        groupLabel: 'Kimi — API (pay-per-use)',
+        hint: 'Pay-per-use via Moonshot API key (platform.kimi.ai).',
+    },
+    'kimi-cn': {
+        name: 'Kimi China (Moonshot)',
+        description: 'Moonshot AI Kimi models (China)',
+        protocols: {
+            openai: { baseUrl: 'https://api.moonshot.cn/v1', authHeader: 'Bearer', supportsNativeTools: true },
+        },
+        models: [
+            { id: 'kimi-k2.7-code', name: 'Kimi K2.7 Code', description: 'Flagship agentic coding model (256K context)' },
+            { id: 'kimi-k2.7-code-highspeed', name: 'Kimi K2.7 Code (High-Speed)', description: 'Throughput-tuned K2.7 Code' },
+            { id: 'kimi-k2.6', name: 'Kimi K2.6', description: 'Previous-gen multimodal reasoning model' },
+            { id: 'kimi-k2.5', name: 'Kimi K2.5', description: 'Older general-purpose model' },
+        ],
+        defaultModel: 'kimi-k2.7-code',
+        defaultProtocol: 'openai',
+        maxOutputTokens: 32_768,
+        envKey: 'MOONSHOT_CN_API_KEY',
+        subscribeUrl: 'https://platform.moonshot.cn/console/api-keys',
+        groupLabel: 'Kimi China — API (pay-per-use)',
+        hint: 'Pay-per-use via Moonshot China API key (platform.moonshot.cn).',
+    },
+    // ── Grok (xAI) ────────────────────────────────────────────────────
+    // Pay-per-use today (console.x.ai key). The SuperGrok / X Premium+
+    // subscription is OAuth-based — added separately. Reasoning models
+    // require max_completion_tokens (like GPT-5), so useMaxCompletionTokens.
+    'grok': {
+        name: 'Grok (xAI)',
+        description: 'xAI Grok models',
+        protocols: {
+            openai: { baseUrl: 'https://api.x.ai/v1', authHeader: 'Bearer', supportsNativeTools: true },
+        },
+        models: [
+            { id: 'grok-build-0.1', name: 'Grok Build 0.1', description: 'Agentic coding model — fast, 256K context' },
+            { id: 'grok-4.3', name: 'Grok 4.3', description: 'Flagship — highest quality, 1M context' },
+            { id: 'grok-code-fast-1', name: 'Grok Code Fast 1', description: 'Low-cost speed-first coder (alias of Build 0.1)' },
+            { id: 'grok-4-fast-reasoning', name: 'Grok 4 Fast (reasoning)', description: 'Cheap reasoning model, very large context' },
+        ],
+        defaultModel: 'grok-build-0.1',
+        defaultProtocol: 'openai',
+        useMaxCompletionTokens: true, // reasoning models reject max_tokens
+        envKey: 'XAI_API_KEY',
+        subscribeUrl: 'https://console.x.ai',
+        groupLabel: 'xAI Grok',
+        hint: 'Pay-per-use via xAI API key (console.x.ai).',
+    },
+    // ── Qwen (Alibaba Model Studio / DashScope) ───────────────────────
+    // Coding Plan subscription = dedicated base URL + sk-sp- key (mirrors
+    // Z.AI). Qwen's OpenAI-compatible surface CANNOT combine tools with
+    // streaming, so all Qwen entries set noStreamWithTools.
+    'qwen': {
+        name: 'Qwen (Alibaba) — Coding Plan',
+        description: 'Qwen Coding Plan subscription',
+        protocols: {
+            openai: { baseUrl: 'https://coding-intl.dashscope.aliyuncs.com/v1', authHeader: 'Bearer', supportsNativeTools: true },
+        },
+        models: [
+            { id: 'qwen3-coder-plus', name: 'Qwen3-Coder Plus', description: 'Flagship coding model — best quality' },
+            { id: 'qwen3-coder-next', name: 'Qwen3-Coder Next', description: 'Balanced quality/speed/cost' },
+            { id: 'qwen3-max', name: 'Qwen3-Max', description: 'Flagship general model (code + reasoning)' },
+        ],
+        defaultModel: 'qwen3-coder-plus',
+        defaultProtocol: 'openai',
+        maxOutputTokens: 65_536,
+        noStreamWithTools: true,
+        envKey: 'BAILIAN_CODING_PLAN_API_KEY',
+        subscribeUrl: 'https://www.alibabacloud.com/help/en/model-studio/qwen-code-coding-plan',
+        groupLabel: 'Qwen — Subscription (Coding Plan)',
+        hint: 'Uses your Qwen Coding Plan — no per-token charges. sk-sp-… key from Model Studio. Interactive coding use only.',
+    },
+    'qwen-api': {
+        name: 'Qwen (Alibaba) API (pay-per-use)',
+        description: 'Alibaba Model Studio Qwen models via API key',
+        protocols: {
+            openai: { baseUrl: 'https://dashscope-intl.aliyuncs.com/compatible-mode/v1', authHeader: 'Bearer', supportsNativeTools: true },
+        },
+        models: [
+            { id: 'qwen3-coder-plus', name: 'Qwen3-Coder Plus', description: 'Flagship coding model (256K, up to 1M)' },
+            { id: 'qwen3-coder-next', name: 'Qwen3-Coder Next', description: 'Balanced quality/speed/cost' },
+            { id: 'qwen3-coder-flash', name: 'Qwen3-Coder Flash', description: 'Fast/cheap coder' },
+            { id: 'qwen3-max', name: 'Qwen3-Max', description: 'Flagship general model' },
+        ],
+        defaultModel: 'qwen3-coder-plus',
+        defaultProtocol: 'openai',
+        maxOutputTokens: 65_536,
+        noStreamWithTools: true,
+        envKey: 'DASHSCOPE_API_KEY',
+        subscribeUrl: 'https://modelstudio.console.alibabacloud.com/',
+        groupLabel: 'Qwen — API (pay-per-use)',
+        hint: 'Pay-per-use via Alibaba Model Studio key (DASHSCOPE_API_KEY).',
+    },
+    'qwen-cn': {
+        name: 'Qwen China — Coding Plan',
+        description: 'Qwen Coding Plan subscription (China)',
+        protocols: {
+            openai: { baseUrl: 'https://coding.dashscope.aliyuncs.com/v1', authHeader: 'Bearer', supportsNativeTools: true },
+        },
+        models: [
+            { id: 'qwen3-coder-plus', name: 'Qwen3-Coder Plus', description: 'Flagship coding model — best quality' },
+            { id: 'qwen3-coder-next', name: 'Qwen3-Coder Next', description: 'Balanced quality/speed/cost' },
+            { id: 'qwen3-max', name: 'Qwen3-Max', description: 'Flagship general model' },
+        ],
+        defaultModel: 'qwen3-coder-plus',
+        defaultProtocol: 'openai',
+        maxOutputTokens: 65_536,
+        noStreamWithTools: true,
+        envKey: 'BAILIAN_CODING_PLAN_CN_API_KEY',
+        subscribeUrl: 'https://bailian.console.aliyun.com/',
+        groupLabel: 'Qwen China — Subscription (Coding Plan)',
+        hint: 'Uses your Qwen Coding Plan (China). sk-sp-… key from Bailian.',
+    },
+    'qwen-cn-api': {
+        name: 'Qwen China API (pay-per-use)',
+        description: 'Alibaba Model Studio Qwen models via API key (China)',
+        protocols: {
+            openai: { baseUrl: 'https://dashscope.aliyuncs.com/compatible-mode/v1', authHeader: 'Bearer', supportsNativeTools: true },
+        },
+        models: [
+            { id: 'qwen3-coder-plus', name: 'Qwen3-Coder Plus', description: 'Flagship coding model' },
+            { id: 'qwen3-coder-next', name: 'Qwen3-Coder Next', description: 'Balanced quality/speed/cost' },
+            { id: 'qwen3-coder-flash', name: 'Qwen3-Coder Flash', description: 'Fast/cheap coder' },
+            { id: 'qwen3-max', name: 'Qwen3-Max', description: 'Flagship general model' },
+        ],
+        defaultModel: 'qwen3-coder-plus',
+        defaultProtocol: 'openai',
+        maxOutputTokens: 65_536,
+        noStreamWithTools: true,
+        envKey: 'DASHSCOPE_CN_API_KEY',
+        subscribeUrl: 'https://bailian.console.aliyun.com/',
+        groupLabel: 'Qwen China — API (pay-per-use)',
+        hint: 'Pay-per-use via Alibaba Model Studio China key.',
+    },
+    'modelscope': {
+        name: 'ModelScope (free Qwen)',
+        description: 'Free Qwen3-Coder inference via ModelScope',
+        protocols: {
+            openai: { baseUrl: 'https://api-inference.modelscope.cn/v1', authHeader: 'Bearer', supportsNativeTools: true },
+        },
+        models: [
+            { id: 'Qwen/Qwen3-Coder-480B-A35B-Instruct', name: 'Qwen3-Coder 480B', description: 'Open MoE coder — free tier (~2000 req/day)' },
+        ],
+        defaultModel: 'Qwen/Qwen3-Coder-480B-A35B-Instruct',
+        defaultProtocol: 'openai',
+        maxOutputTokens: 65_536,
+        noStreamWithTools: true,
+        envKey: 'MODELSCOPE_API_KEY',
+        subscribeUrl: 'https://modelscope.cn/my/myaccesstoken',
+        groupLabel: 'ModelScope — Free (Qwen)',
+        hint: 'Free tier (~2000 req/day) via ModelScope token (modelscope.cn). Needs a bound Aliyun account.',
+    },
     'openai': {
         name: 'OpenAI',
         description: 'GPT and o-series models',
@@ -361,14 +551,24 @@ const DISPLAY_ORDER = [
     'openrouter', // 100+ models, one key — surfaced high on purpose for 2.0.0.
     'z.ai',
     'z.ai-api',
+    'kimi',
+    'kimi-api',
+    'qwen',
+    'qwen-api',
+    'grok',
     'deepseek',
     'google',
     'minimax',
     'minimax-api',
+    'modelscope',
     'ollama',
     'custom',
+    // Regional + parameter-variant entries trail.
     'z.ai-cn',
     'z.ai-cn-api',
+    'kimi-cn',
+    'qwen-cn',
+    'qwen-cn-api',
     'minimax-cn',
 ];
 export function getProviderList() {
@@ -441,14 +641,26 @@ export function usesMaxCompletionTokens(providerId) {
 export function requiresDefaultTemperature(providerId) {
     return PROVIDERS[providerId]?.requiresDefaultTemperature ?? false;
 }
+/**
+ * Returns true if the provider's OpenAI-compatible endpoint rejects `tools`
+ * together with `stream: true` (Alibaba/Qwen) — callers must issue tool-bearing
+ * agent turns non-streamed.
+ */
+export function providerNoStreamWithTools(providerId) {
+    return PROVIDERS[providerId]?.noStreamWithTools ?? false;
+}
 /**
  * Models that reject sampling parameters (temperature/top_p/top_k) with a 400.
  * Anthropic removed them on Fable 5 and Opus 4.7+; older Claude models still
  * accept them, so this must be a MODEL-level check, not a provider-level one
  * (requiresDefaultTemperature can't express it). Omitting the field is always
- * safe — the API treats omission as default.
+ * safe — the API treats omission as default. Kimi K2.x code/thinking models
+ * fix temperature internally and 400 on any custom value, so they're here too.
  */
-const SAMPLING_PARAMS_REJECTED = ['claude-fable-5', 'claude-opus-4-8', 'claude-opus-4-7'];
+const SAMPLING_PARAMS_REJECTED = [
+    'claude-fable-5', 'claude-opus-4-8', 'claude-opus-4-7',
+    'kimi-k2.7-code', 'kimi-for-coding',
+];
 export function modelRejectsSamplingParams(model) {
     return SAMPLING_PARAMS_REJECTED.some(id => model === id || model.startsWith(`${id}-`));
 }
@@ -508,6 +720,12 @@ export function modelSupportsReasoningEffort(providerId, model) {
             // GLM-5.2 added graded High/Max effort. glm-5-turbo is a plain thinking
             // toggle (no graded levels) so it stays out.
             return idMatches(id, 'glm-5-2');
+        case 'grok':
+            // Grok reasoning models accept reasoning_effort (none/low/medium/high).
+            // Explicit *-non-reasoning variants don't think → excluded.
+            return id.startsWith('grok') && !id.includes('non-reasoning');
+        // Kimi (thinking on/off, not graded) and Qwen coders (non-thinking) have
+        // no graded knob → fall through to default false.
         case 'openrouter':
             // OpenRouter normalizes a unified `reasoning` field and silently ignores
             // it for non-reasoning models, so the control is always safe to expose.
@@ -547,6 +765,9 @@ export function reasoningParamsFor(providerId, model, tier) {
         case 'z.ai-cn-api':
             // Graded thinking depth: high (default) or max. Lower tiers collapse to high.
             return { reasoning_effort: tier === 'max' ? 'max' : 'high' };
+        case 'grok':
+            // none/low/medium/high — no "max"; map our Max → high (the ceiling).
+            return { reasoning_effort: tier === 'max' ? 'high' : tier };
         case 'openrouter':
             // Unified reasoning object; no "max" effort → cap at high.
             return { reasoning: { effort: tier === 'max' ? 'high' : tier } };
@@ -580,6 +801,8 @@ export function availableReasoningTiers(providerId, model) {
         case 'z.ai-cn':
         case 'z.ai-cn-api':
             return ['auto', 'high', 'max'];
+        case 'grok':
+            return ['auto', 'low', 'medium', 'high'];
         case 'openrouter':
             return ['auto', 'low', 'medium', 'high'];
         default:

package/dist/utils/agentChat.js

@@ -18,7 +18,7 @@ import { config, getApiKey, resolveBaseUrl } from '../config/index.js';
 import { loadProjectIntelligence, generateContextFromIntelligence } from './projectIntelligence.js';
 import { formatCommandIndex } from './commandIndex.js';
 import { syncProgress, generateProjectId } from './codeepCloud.js';
-import { getProviderAuthHeader, supportsNativeTools, getEffectiveMaxTokens, usesMaxCompletionTokens, requiresDefaultTemperature, modelRejectsSamplingParams, isNoApiKeyProvider, reasoningParamsFor } from '../config/providers.js';
+import { getProviderAuthHeader, supportsNativeTools, getEffectiveMaxTokens, usesMaxCompletionTokens, requiresDefaultTemperature, modelRejectsSamplingParams, isNoApiKeyProvider, reasoningParamsFor, providerNoStreamWithTools } from '../config/providers.js';
 import { recordTokenUsage, extractOpenAIUsage, extractAnthropicUsage } from './tokenTracker.js';
 import { parseOpenAIToolCalls, parseAnthropicToolCalls, parseToolCalls } from './toolParsing.js';
 import { formatToolDefinitions, getOpenAITools, getAnthropicTools } from './tools.js';
@@ -367,7 +367,10 @@ additionalTools) {
     try {
         let endpoint;
         let body;
-        const useStreaming = Boolean(onChunk);
+        // Qwen/DashScope reject `tools` + `stream:true` together; this path always
+        // sends tools, so force a non-streamed request there (the non-streaming
+        // branch below still emits the content via onChunk). Other providers stream.
+        const useStreaming = Boolean(onChunk) && !providerNoStreamWithTools(providerId);
         // Provider-level guard (OpenAI GPT-5+) OR model-level guard — Anthropic's
         // Fable 5 / Opus 4.7+ reject temperature with a 400; omission is safe.
         const tempParam = (requiresDefaultTemperature(providerId) || modelRejectsSamplingParams(model)) ? {} : { temperature: config.get('temperature') };

package/dist/utils/tokenTracker.js

@@ -26,6 +26,23 @@ const MODEL_CONTEXT_WINDOWS = {
     'gemini-3-flash-preview': 1_000_000,
     // MiniMax
     'MiniMax-M3': 524_288,
+    // Kimi (Moonshot) — 256K across the K2.x line
+    'kimi-k2.7-code': 262_144,
+    'kimi-k2.7-code-highspeed': 262_144,
+    'kimi-k2.6': 262_144,
+    'kimi-k2.5': 262_144,
+    'kimi-for-coding': 262_144,
+    // Grok (xAI)
+    'grok-build-0.1': 256_000,
+    'grok-4.3': 1_000_000,
+    'grok-code-fast-1': 256_000,
+    'grok-4-fast-reasoning': 2_000_000,
+    // Qwen (Alibaba) — 256K native (1M with extrapolation)
+    'qwen3-coder-plus': 262_144,
+    'qwen3-coder-next': 262_144,
+    'qwen3-coder-flash': 262_144,
+    'qwen3-max': 262_144,
+    'Qwen/Qwen3-Coder-480B-A35B-Instruct': 262_144,
 };
 const DEFAULT_CONTEXT_WINDOW = 128_000;
 /**
@@ -62,6 +79,26 @@ const MODEL_PRICING = {
     'gemini-3-flash-preview': { inputPer1M: 0.50, outputPer1M: 3.00 },
     // MiniMax
     'MiniMax-M3': { inputPer1M: 0.60, outputPer1M: 2.40 },
+    // Kimi (Moonshot) — pay-per-use cache-miss rates; `kimi-for-coding` is the
+    // subscription alias (flat-fee in reality, priced notionally like K2.7 Code).
+    'kimi-k2.7-code': { inputPer1M: 0.60, outputPer1M: 2.50 },
+    'kimi-k2.7-code-highspeed': { inputPer1M: 0.60, outputPer1M: 2.50 },
+    'kimi-k2.6': { inputPer1M: 0.55, outputPer1M: 2.20 },
+    'kimi-k2.5': { inputPer1M: 0.40, outputPer1M: 1.90 },
+    'kimi-for-coding': { inputPer1M: 0.60, outputPer1M: 2.50 },
+    // Grok (xAI)
+    'grok-build-0.1': { inputPer1M: 1.00, outputPer1M: 2.00 },
+    'grok-4.3': { inputPer1M: 1.25, outputPer1M: 2.50 },
+    'grok-code-fast-1': { inputPer1M: 0.20, outputPer1M: 1.50 },
+    'grok-4-fast-reasoning': { inputPer1M: 0.20, outputPer1M: 0.50 },
+    // Qwen (Alibaba) — qwen3-coder-* 0–256K tier; the Coding Plan is flat-fee so
+    // this only affects the pay-per-use estimate.
+    'qwen3-coder-plus': { inputPer1M: 0.28, outputPer1M: 1.65 },
+    'qwen3-coder-next': { inputPer1M: 0.28, outputPer1M: 1.65 },
+    'qwen3-coder-flash': { inputPer1M: 0.10, outputPer1M: 0.50 },
+    'qwen3-max': { inputPer1M: 1.20, outputPer1M: 6.00 },
+    // ModelScope free tier — no per-token charge.
+    'Qwen/Qwen3-Coder-480B-A35B-Instruct': { inputPer1M: 0, outputPer1M: 0 },
 };
 export function getPricingTable() {
     return Object.entries(MODEL_PRICING).map(([model, p]) => ({ model, ...p }));

package/dist/version.d.ts

@@ -1 +1 @@
-export declare const VERSION = "2.12.0";
+export declare const VERSION = "2.13.0";

package/dist/version.js

@@ -1,4 +1,4 @@
 // AUTO-GENERATED by scripts/gen-version.js — do not edit by hand.
 // Baked from package.json at build time so the bun-compiled binary reports
 // the right version (it has no package.json on disk to read at runtime).
-export const VERSION = '2.12.0';
+export const VERSION = '2.13.0';

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "codeep",
-  "version": "2.12.0",
+  "version": "2.13.0",
   "description": "AI-powered coding assistant built for the terminal. Multiple LLM providers, project-aware context, and a seamless development workflow.",
   "type": "module",
   "main": "dist/index.js",