@plune-ai/cli 0.11.0 → 0.11.1

package/dist/cli.cjs

@@ -597,7 +597,7 @@ function makeAnthropicProvider(config, env, pricing) {
           () => client.messages.create({
             model: req.model,
             max_tokens: req.max_tokens,
-            temperature: req.temperature,
+            ...req.temperature !== void 0 ? { temperature: req.temperature } : {},
             messages: [{ role: "user", content: req.prompt_resolved }]
           }),
           { max_retries: maxRetries }
@@ -657,7 +657,7 @@ function makeOpenAiCompatibleProvider(opts) {
         const res = await withRetry(
           () => client.chat.completions.create({
             model: req.model,
-            temperature: req.temperature,
+            ...req.temperature !== void 0 ? { temperature: req.temperature } : {},
             max_tokens: req.max_tokens,
             messages: [{ role: "user", content: req.prompt_resolved }],
             // OpenRouter extension: ask it to include the call's actual cost in `usage` (ADR-PRC02).
@@ -1547,7 +1547,7 @@ function buildJudge(provider, cfg, onUsage) {
       const res = await provider.complete({
         provider: cfg.type,
         model: cfg.model,
-        temperature: 0,
+        ...cfg.temperature !== void 0 ? { temperature: cfg.temperature } : {},
         max_tokens: cfg.max_tokens ?? DEFAULT_MAX_TOKENS,
         prompt_resolved: prompt5
       });
@@ -1609,7 +1609,7 @@ function estimateUsage(prompt5, maxTokens) {
 async function runRow(p) {
   const t0 = p.now();
   const prompt5 = resolvePrompt(p.template, p.row.vars);
-  const temperature = p.providerConfig.temperature ?? 0;
+  const temperature = p.providerConfig.temperature;
   const maxTokens = p.providerConfig.max_tokens ?? DEFAULT_MAX_TOKENS2;
   if (p.dryRun) {
     const est = estimateUsage(prompt5, maxTokens);
@@ -1619,7 +1619,7 @@ async function runRow(p) {
   const key = cacheKey({
     provider: p.providerConfig.type,
     model: p.providerConfig.model,
-    temperature,
+    temperature: temperature ?? 0,
     max_tokens: maxTokens,
     prompt_resolved: prompt5
   });
@@ -1640,7 +1640,7 @@ async function runRow(p) {
       const res = await p.provider.complete({
         provider: p.providerConfig.type,
         model: p.providerConfig.model,
-        temperature,
+        ...temperature !== void 0 ? { temperature } : {},
         max_tokens: maxTokens,
         prompt_resolved: prompt5
       });

package/dist/index.cjs

@@ -480,7 +480,7 @@ function makeAnthropicProvider(config, env, pricing) {
           () => client.messages.create({
             model: req.model,
             max_tokens: req.max_tokens,
-            temperature: req.temperature,
+            ...req.temperature !== void 0 ? { temperature: req.temperature } : {},
             messages: [{ role: "user", content: req.prompt_resolved }]
           }),
           { max_retries: maxRetries }
@@ -529,7 +529,7 @@ function makeOpenAiCompatibleProvider(opts) {
         const res = await withRetry(
           () => client.chat.completions.create({
             model: req.model,
-            temperature: req.temperature,
+            ...req.temperature !== void 0 ? { temperature: req.temperature } : {},
             max_tokens: req.max_tokens,
             messages: [{ role: "user", content: req.prompt_resolved }],
             // OpenRouter extension: ask it to include the call's actual cost in `usage` (ADR-PRC02).
@@ -1216,7 +1216,7 @@ function buildJudge(provider, cfg, onUsage) {
       const res = await provider.complete({
         provider: cfg.type,
         model: cfg.model,
-        temperature: 0,
+        ...cfg.temperature !== void 0 ? { temperature: cfg.temperature } : {},
         max_tokens: cfg.max_tokens ?? DEFAULT_MAX_TOKENS,
         prompt_resolved: prompt5
       });
@@ -1273,7 +1273,7 @@ function estimateUsage(prompt5, maxTokens) {
 async function runRow(p) {
   const t0 = p.now();
   const prompt5 = resolvePrompt(p.template, p.row.vars);
-  const temperature = p.providerConfig.temperature ?? 0;
+  const temperature = p.providerConfig.temperature;
   const maxTokens = p.providerConfig.max_tokens ?? DEFAULT_MAX_TOKENS2;
   if (p.dryRun) {
     const est = estimateUsage(prompt5, maxTokens);
@@ -1283,7 +1283,7 @@ async function runRow(p) {
   const key = cacheKey({
     provider: p.providerConfig.type,
     model: p.providerConfig.model,
-    temperature,
+    temperature: temperature ?? 0,
     max_tokens: maxTokens,
     prompt_resolved: prompt5
   });
@@ -1304,7 +1304,7 @@ async function runRow(p) {
       const res = await p.provider.complete({
         provider: p.providerConfig.type,
         model: p.providerConfig.model,
-        temperature,
+        ...temperature !== void 0 ? { temperature } : {},
         max_tokens: maxTokens,
         prompt_resolved: prompt5
       });

package/dist/index.d.cts

@@ -146,7 +146,12 @@ interface Config {
 interface CompletionRequest {
     provider: string;
     model: string;
-    temperature: number;
+    /**
+     * Only when the config asked for one. Models released after Claude Opus 4.6 do not take a
+     * temperature at all — any value but 1.0 is a 400 — so a default here would be a default that
+     * breaks every eval on a current model. Absent means "the model's own".
+     */
+    temperature?: number;
     max_tokens: number;
     prompt_resolved: string;
 }

package/dist/index.d.ts

@@ -146,7 +146,12 @@ interface Config {
 interface CompletionRequest {
     provider: string;
     model: string;
-    temperature: number;
+    /**
+     * Only when the config asked for one. Models released after Claude Opus 4.6 do not take a
+     * temperature at all — any value but 1.0 is a 400 — so a default here would be a default that
+     * breaks every eval on a current model. Absent means "the model's own".
+     */
+    temperature?: number;
     max_tokens: number;
     prompt_resolved: string;
 }

package/dist/index.js

@@ -427,7 +427,7 @@ function makeAnthropicProvider(config, env, pricing) {
           () => client.messages.create({
             model: req.model,
             max_tokens: req.max_tokens,
-            temperature: req.temperature,
+            ...req.temperature !== void 0 ? { temperature: req.temperature } : {},
             messages: [{ role: "user", content: req.prompt_resolved }]
           }),
           { max_retries: maxRetries }
@@ -476,7 +476,7 @@ function makeOpenAiCompatibleProvider(opts) {
         const res = await withRetry(
           () => client.chat.completions.create({
             model: req.model,
-            temperature: req.temperature,
+            ...req.temperature !== void 0 ? { temperature: req.temperature } : {},
             max_tokens: req.max_tokens,
             messages: [{ role: "user", content: req.prompt_resolved }],
             // OpenRouter extension: ask it to include the call's actual cost in `usage` (ADR-PRC02).
@@ -1163,7 +1163,7 @@ function buildJudge(provider, cfg, onUsage) {
       const res = await provider.complete({
         provider: cfg.type,
         model: cfg.model,
-        temperature: 0,
+        ...cfg.temperature !== void 0 ? { temperature: cfg.temperature } : {},
         max_tokens: cfg.max_tokens ?? DEFAULT_MAX_TOKENS,
         prompt_resolved: prompt5
       });
@@ -1220,7 +1220,7 @@ function estimateUsage(prompt5, maxTokens) {
 async function runRow(p) {
   const t0 = p.now();
   const prompt5 = resolvePrompt(p.template, p.row.vars);
-  const temperature = p.providerConfig.temperature ?? 0;
+  const temperature = p.providerConfig.temperature;
   const maxTokens = p.providerConfig.max_tokens ?? DEFAULT_MAX_TOKENS2;
   if (p.dryRun) {
     const est = estimateUsage(prompt5, maxTokens);
@@ -1230,7 +1230,7 @@ async function runRow(p) {
   const key = cacheKey({
     provider: p.providerConfig.type,
     model: p.providerConfig.model,
-    temperature,
+    temperature: temperature ?? 0,
     max_tokens: maxTokens,
     prompt_resolved: prompt5
   });
@@ -1251,7 +1251,7 @@ async function runRow(p) {
       const res = await p.provider.complete({
         provider: p.providerConfig.type,
         model: p.providerConfig.model,
-        temperature,
+        ...temperature !== void 0 ? { temperature } : {},
         max_tokens: maxTokens,
         prompt_resolved: prompt5
       });

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@plune-ai/cli",
-  "version": "0.11.0",
+  "version": "0.11.1",
   "description": "AI-powered assertion testing for LLM apps — write assertions, run them against any provider, get a pass/fail report (local, CI, or diff).",
   "keywords": [
     "ai",