reasonix 0.5.20 → 0.5.21

package/dist/index.d.ts

@@ -97,6 +97,21 @@ interface ChatRequestOptions {
     responseFormat?: {
         type: "json_object" | "text";
     };
+    /**
+     * Explicitly toggle V4 thinking mode. Serialized as
+     * `extra_body.thinking.type = enabled|disabled`. Omit to let the
+     * server default apply (thinking enabled). Mainly used so the loop
+     * can pin the mode per model: `deepseek-chat` → disabled (legacy
+     * non-thinking compat), everything else → enabled.
+     */
+    thinking?: "enabled" | "disabled";
+    /**
+     * Per-request reasoning-effort cap. Serialized as the top-level
+     * `reasoning_effort` field. DeepSeek accepts `high` (standard) or
+     * `max` (Agent-class, auto-applied to Claude-Code-style flows per
+     * the V4 docs). Reasonix pins `max` because every turn is agent-like.
+     */
+    reasoningEffort?: "high" | "max";
 }
 
 declare class Usage {
@@ -848,6 +863,11 @@ interface CacheFirstLoopOptions {
      * since the default selector scores samples by plan-state uncertainty.
      */
     branch?: number | BranchOptions;
+    /**
+     * Reasoning-effort cap. See {@link ReconfigurableOptions} — default
+     * `max` for Reasonix (agent-class use per DeepSeek V4 docs).
+     */
+    reasoningEffort?: "high" | "max";
     /**
      * Session name. When set, the loop pre-loads the session's prior messages
      * into its log on construction, and appends every new log entry to
@@ -884,6 +904,14 @@ interface ReconfigurableOptions {
     harvest?: boolean | HarvestOptions;
     branch?: number | BranchOptions;
     stream?: boolean;
+    /**
+     * Reasoning-effort cap sent per turn (V4 thinking mode only;
+     * deepseek-chat ignores it). Reasonix pins `max` by default because
+     * DeepSeek's V4 docs flag Claude-Code-style agent loops as the
+     * canonical `max` use case. `/effort high` lets a user step down
+     * mid-session for cheaper, faster turns on simple tasks.
+     */
+    reasoningEffort?: "high" | "max";
 }
 declare class CacheFirstLoop {
     readonly client: DeepSeekClient;
@@ -900,6 +928,8 @@ declare class CacheFirstLoop {
     harvestOptions: HarvestOptions;
     branchEnabled: boolean;
     branchOptions: BranchOptions;
+    /** See ReconfigurableOptions — mutable so `/effort` can flip mid-session. */
+    reasoningEffort: "high" | "max";
     sessionName: string | null;
     /**
      * Hook list, mutable so `/hooks reload` can swap it without
@@ -1620,7 +1650,7 @@ interface SubagentToolOptions {
     defaultSystem?: string;
     /** Project root for `applyProjectMemory` lookup. Omit in chat mode. */
     projectRoot?: string;
-    /** Default model. `deepseek-chat` (V3) by default. */
+    /** Default model. `deepseek-v4-pro` by default. */
     defaultModel?: string;
     /** Iteration ceiling. Lower than the parent (16 by default). */
     maxToolIters?: number;

package/dist/index.js

@@ -131,6 +131,12 @@ var DeepSeekClient = class {
     if (opts.temperature !== void 0) payload.temperature = opts.temperature;
     if (opts.maxTokens !== void 0) payload.max_tokens = opts.maxTokens;
     if (opts.responseFormat) payload.response_format = opts.responseFormat;
+    if (opts.thinking) {
+      payload.extra_body = { thinking: { type: opts.thinking } };
+    }
+    if (opts.reasoningEffort) {
+      payload.reasoning_effort = opts.reasoningEffort;
+    }
     return payload;
   }
   /**
@@ -345,6 +351,13 @@ async function harvest(reasoningContent, client, options = {}, signal) {
       responseFormat: { type: "json_object" },
       temperature: 0,
       maxTokens: 600,
+      // Pin mode + effort so a future default-model swap (e.g. someone
+      // sets `options.model = "deepseek-v4-pro"`) can't accidentally
+      // turn this micro-extraction into a multi-thousand-reasoning-
+      // token call. DeepSeek ignores these on non-thinking models, so
+      // the request stays valid regardless of the chosen model.
+      thinking: "disabled",
+      reasoningEffort: "high",
       signal
     });
     return parsePlanState(resp.content, maxItems, maxItemLen);
@@ -1704,6 +1717,8 @@ var CacheFirstLoop = class {
   harvestOptions;
   branchEnabled;
   branchOptions;
+  /** See ReconfigurableOptions — mutable so `/effort` can flip mid-session. */
+  reasoningEffort;
   sessionName;
   /**
    * Hook list, mutable so `/hooks reload` can swap it without
@@ -1729,7 +1744,8 @@ var CacheFirstLoop = class {
     this.client = opts.client;
     this.prefix = opts.prefix;
     this.tools = opts.tools ?? new ToolRegistry();
-    this.model = opts.model ?? "deepseek-chat";
+    this.model = opts.model ?? "deepseek-v4-pro";
+    this.reasoningEffort = opts.reasoningEffort ?? "max";
     this.maxToolIters = opts.maxToolIters ?? 64;
     this.hooks = opts.hooks ?? [];
     this.hookCwd = opts.hookCwd ?? process.cwd();
@@ -1845,6 +1861,7 @@ var CacheFirstLoop = class {
   configure(opts) {
     if (opts.model !== void 0) this.model = opts.model;
     if (opts.stream !== void 0) this._streamPreference = opts.stream;
+    if (opts.reasoningEffort !== void 0) this.reasoningEffort = opts.reasoningEffort;
     if (opts.branch !== void 0) {
       if (typeof opts.branch === "number") {
         this.branchOptions = { budget: opts.branch };
@@ -2023,7 +2040,9 @@ var CacheFirstLoop = class {
               model: this.model,
               messages,
               tools: toolSpecs.length ? toolSpecs : void 0,
-              signal
+              signal,
+              thinking: thinkingModeForModel(this.model),
+              reasoningEffort: this.reasoningEffort
             },
             {
               ...this.branchOptions,
@@ -2075,7 +2094,9 @@ var CacheFirstLoop = class {
             model: this.model,
             messages,
             tools: toolSpecs.length ? toolSpecs : void 0,
-            signal
+            signal,
+            thinking: thinkingModeForModel(this.model),
+            reasoningEffort: this.reasoningEffort
           })) {
             if (chunk.contentDelta) {
               assistantContent += chunk.contentDelta;
@@ -2129,7 +2150,9 @@ var CacheFirstLoop = class {
             model: this.model,
             messages,
             tools: toolSpecs.length ? toolSpecs : void 0,
-            signal
+            signal,
+            thinking: thinkingModeForModel(this.model),
+            reasoningEffort: this.reasoningEffort
           });
           assistantContent = resp.content;
           reasoningContent = resp.reasoningContent ?? "";
@@ -2322,7 +2345,9 @@ ${reason}`;
         model: this.model,
         messages,
         // no tools → model is forced to answer in text
-        signal: this._turnAbort.signal
+        signal: this._turnAbort.signal,
+        thinking: thinkingModeForModel(this.model),
+        reasoningEffort: this.reasoningEffort
       });
       const rawContent = resp.content?.trim() ?? "";
       const cleaned = stripHallucinatedToolMarkup(rawContent);
@@ -2390,6 +2415,12 @@ function isThinkingModeModel(model) {
   if (model === "deepseek-v4-flash" || model === "deepseek-v4-pro") return true;
   return false;
 }
+function thinkingModeForModel(model) {
+  if (model === "deepseek-chat") return "disabled";
+  if (model.includes("reasoner")) return "enabled";
+  if (model === "deepseek-v4-flash" || model === "deepseek-v4-pro") return "enabled";
+  return void 0;
+}
 function stripHallucinatedToolMarkup(s) {
   let out = s;
   out = out.replace(/<｜DSML｜function_calls>[\s\S]*?<\/?｜DSML｜function_calls>/g, "");
@@ -4003,7 +4034,7 @@ Formatting rules (the parent renders your reply in a TUI with a real markdown re
 - For flow charts and diagrams: use a markdown bullet list with \`\u2192\` or \`\u2193\` between steps. Don't try to draw boxes-and-arrows in ASCII; it never survives word-wrap.`;
 var DEFAULT_MAX_RESULT_CHARS2 = 8e3;
 var DEFAULT_MAX_ITERS = 16;
-var DEFAULT_SUBAGENT_MODEL = "deepseek-chat";
+var DEFAULT_SUBAGENT_MODEL = "deepseek-v4-pro";
 var SUBAGENT_TOOL_NAME = "spawn_subagent";
 var NEVER_INHERITED_TOOLS = /* @__PURE__ */ new Set([SUBAGENT_TOOL_NAME, "submit_plan"]);
 async function spawnSubagent(opts) {
@@ -4156,7 +4187,7 @@ function registerSubagentTool(parentRegistry, opts) {
         model: {
           type: "string",
           enum: ["deepseek-v4-flash", "deepseek-v4-pro", "deepseek-chat", "deepseek-reasoner"],
-          description: "Which DeepSeek model the subagent runs on. 'deepseek-v4-flash' (default; thinking mode) is fast and cheap and is what the legacy 'deepseek-chat' / 'deepseek-reasoner' aliases route to today. Use 'deepseek-v4-pro' only when the subtask needs the strongest model \u2014 roughly 12\xD7 the input cost and 12\xD7 the output cost vs flash."
+          description: "Which DeepSeek model the subagent runs on. Default is 'deepseek-v4-pro' \u2014 the strongest model, best for complex subtasks. Override to 'deepseek-v4-flash' (or the legacy 'deepseek-chat' / 'deepseek-reasoner' aliases, which route to flash non-thinking / thinking modes) when the subtask is simple enough that flash's quality suffices \u2014 flash is roughly 12\xD7 cheaper."
         }
       },
       required: ["task"]