@tuned-tensor/cli 0.18.0 → 0.20.0

package/README.md

@@ -96,7 +96,8 @@ npm link
 `tunedtensor.json` is the source of truth for behavior and training settings.
 Use `/spec`, `/spec diff`, `/spec validate`, and `/spec history` in the shell to
 review it without a model call. Ask the agent for changes, inspect its proposed
-diff, then `/approve` to save a validated edit with revision history. Pipeline
+diff, then press `y` when tt asks (or `/approve` later) to save a validated edit
+with revision history. Pipeline
 runs use the saved spec. See [the spec workflow](docs/spec-workflow.md).
 
 ## Conversational terminal
@@ -173,7 +174,9 @@ Ordinary sentences go through the locally orchestrated model session. The
 model has no shell or general filesystem tool. It can read and propose reviewed
 edits to an existing spec, or prepare one new folder directly beneath the shell's current working directory with a validated
 `tunedtensor.json`, or prepare a local pipeline dry-run sealed to the reviewed
-workspace and spec contents. Writes and pipeline previews wait for `/approve`. The tools refuse path
+workspace and spec contents. Writes and pipeline previews wait for your approval: after the
+answer, tt shows the proposal and asks inline, where `y` approves, `n` rejects and `Esc`
+leaves it pending for a later `/approve` or `/reject`. Any other key is ignored. The tools refuse path
 traversal, symlinked workspace roots, overwriting a new-project target, unsupported spec
 fields, cloud pipeline targets, and spec changes after review. Known CLI
 commands such as `runs list`, `doctor`, and `models list` still execute directly.

package/dist/{chunk-U5HNS367.js → chunk-4Q54NHKR.js}

@@ -1227,6 +1227,116 @@ import { readFile as readFile2, stat as stat2, writeFile as writeFile2 } from "f
 import { dirname as dirname2, isAbsolute as isAbsolute3, relative as relative2, resolve as resolve4, sep as sep2 } from "path";
 import { homedir as homedir2 } from "os";
 import { mkdir as mkdir2 } from "fs/promises";
+
+// src/eval/rules.ts
+function validateEvalCases(spec) {
+  const errors = [];
+  const cases = spec.eval_cases ?? [];
+  for (const [caseIndex, evalCase] of cases.entries()) {
+    const label = `eval_cases[${caseIndex}]`;
+    if (!evalCase || typeof evalCase !== "object") {
+      errors.push(`${label} must be an object`);
+      continue;
+    }
+    if (typeof evalCase.input !== "string" || evalCase.input.trim().length === 0) {
+      errors.push(`${label}.input must be a non-empty string`);
+    }
+    if (evalCase.runtime !== "python") {
+      errors.push(`${label}.runtime must be "python"`);
+    }
+    if (!Array.isArray(evalCase.tests) || evalCase.tests.length === 0) {
+      errors.push(`${label}.tests must contain at least one test`);
+      continue;
+    }
+    for (const [testIndex, test] of evalCase.tests.entries()) {
+      const testLabel = `${label}.tests[${testIndex}]`;
+      if (!test || typeof test !== "object") {
+        errors.push(`${testLabel} must be an object`);
+        continue;
+      }
+      if (test.name !== void 0 && (typeof test.name !== "string" || test.name.trim().length === 0)) {
+        errors.push(`${testLabel}.name must be a non-empty string when provided`);
+      }
+      if (test.args !== void 0 && (!Array.isArray(test.args) || !test.args.every((arg) => typeof arg === "string"))) {
+        errors.push(`${testLabel}.args must be an array of strings`);
+      }
+      if (test.stdin !== void 0 && typeof test.stdin !== "string") {
+        errors.push(`${testLabel}.stdin must be a string`);
+      }
+      if (test.expected_exit_code !== void 0 && (typeof test.expected_exit_code !== "number" || !Number.isInteger(test.expected_exit_code) || test.expected_exit_code < 0 || test.expected_exit_code > 255)) {
+        errors.push(`${testLabel}.expected_exit_code must be an integer between 0 and 255`);
+      }
+      if (test.expected_stdout !== void 0 && typeof test.expected_stdout !== "string") {
+        errors.push(`${testLabel}.expected_stdout must be a string`);
+      }
+      validateFiles(test.files, `${testLabel}.files`, errors);
+      validateFiles(test.expected_files, `${testLabel}.expected_files`, errors);
+    }
+  }
+  return errors;
+}
+function validateFiles(files, label, errors) {
+  if (files === void 0) return;
+  if (!Array.isArray(files)) {
+    errors.push(`${label} must be an array`);
+    return;
+  }
+  for (const [index, file] of files.entries()) {
+    const fileLabel = `${label}[${index}]`;
+    if (!file || typeof file !== "object") {
+      errors.push(`${fileLabel} must be an object`);
+      continue;
+    }
+    const row = file;
+    if (typeof row.path !== "string" || !isSafeRelativePath(row.path)) {
+      errors.push(`${fileLabel}.path must be a safe relative path`);
+    }
+    if (row.content !== void 0 && typeof row.content !== "string") {
+      errors.push(`${fileLabel}.content must be a string`);
+    }
+  }
+}
+function isSafeRelativePath(value) {
+  if (value.length === 0 || value.length > 240) return false;
+  if (value.includes("\0") || value.startsWith("/")) return false;
+  const normalized = value.replace(/\\/g, "/");
+  return !normalized.split("/").some((part) => part === "" || part === "." || part === "..");
+}
+function evaluateConstraint(text, constraint) {
+  const lower = constraint.toLowerCase();
+  const neverMatch = lower.match(/^never (?:share|mention|include|reveal|use|say|output|return|give|provide|show|disclose|expose)\s+(.+)/);
+  if (neverMatch) {
+    const forbidden = neverMatch[1].replace(/['"]/g, "").trim();
+    if (text.toLowerCase().includes(forbidden.toLowerCase())) {
+      return { passed: false, message: `Output contains "${forbidden}"` };
+    }
+    return { passed: true };
+  }
+  const alwaysMatch = lower.match(/^always (?:include|use|start with|end with|contain|respond in|reply in)\s+(.+)/);
+  if (alwaysMatch) {
+    const required = alwaysMatch[1].replace(/['"]/g, "").trim();
+    if (!text.toLowerCase().includes(required.toLowerCase())) {
+      return { passed: false, message: `Output missing "${required}"` };
+    }
+    return { passed: true };
+  }
+  return {
+    passed: true,
+    message: `Constraint not enforceable by rules: "${constraint}" (will be checked by LLM judge if a provider is configured)`
+  };
+}
+function checkConstraints(text, constraints) {
+  return constraints.map((constraint) => {
+    const result = evaluateConstraint(text, constraint);
+    return {
+      assertion: `constraint: ${constraint}`,
+      passed: result.passed,
+      message: result.message
+    };
+  });
+}
+
+// src/local-runtime/local-project.ts
 var DEFAULT_LOCAL_SPEC_PATH = "tunedtensor.json";
 function resolveLocalReference(value, baseDirectory) {
   if (typeof value !== "string" || !value || isAbsolute3(value) || /^[a-z][a-z0-9+.-]*:/i.test(value)) {
@@ -1516,6 +1626,11 @@ function validateBehaviorSpec(input) {
     }
     seen.add(key);
   }
+  for (const [index, example] of spec.examples.entries()) {
+    for (const result of checkConstraints(example.output, spec.constraints ?? [])) {
+      if (!result.passed) warnings.push(`examples[${index}] may violate ${result.assertion} (${result.message})`);
+    }
+  }
   if (!spec.guidelines.length) {
     warnings.push("Add guidelines to make expected behavior explicit.");
   }
@@ -6382,6 +6497,7 @@ export {
   localAdapterSpecFileSchema,
   localBehaviorSpecFileSchema,
   isFoundationSpecFile,
+  validateEvalCases,
   DEFAULT_LOCAL_SPEC_PATH,
   localPathsOverlap,
   resolveLocalRunInputPaths,
@@ -6457,4 +6573,4 @@ export {
   formatAgentHostBlock,
   assessHardware
 };
-//# sourceMappingURL=chunk-U5HNS367.js.map
+//# sourceMappingURL=chunk-4Q54NHKR.js.map