@rulemetric/cli 0.18.1 → 0.19.1

package/oclif.manifest.json

@@ -937,6 +937,120 @@
         "list.js"
       ]
     },
+    "experiments:enroll": {
+      "aliases": [],
+      "args": {},
+      "description": "Submit a fresh work request to an eligible project experiment",
+      "examples": [
+        "<%= config.bin %> experiments enroll --project /repo --family parser-repair --prompt-file task.txt"
+      ],
+      "flags": {
+        "api-url": {
+          "description": "Override API base URL",
+          "env": "RULEMETRIC_API_URL",
+          "hidden": true,
+          "name": "api-url",
+          "hasDynamicHelp": false,
+          "multiple": false,
+          "type": "option"
+        },
+        "project": {
+          "char": "p",
+          "description": "Project path (defaults to current directory)",
+          "name": "project",
+          "hasDynamicHelp": false,
+          "multiple": false,
+          "type": "option"
+        },
+        "family": {
+          "description": "Task family shown by the waiting experiment",
+          "name": "family",
+          "required": true,
+          "hasDynamicHelp": false,
+          "multiple": false,
+          "type": "option"
+        },
+        "prompt-file": {
+          "description": "UTF-8 file containing a real, fresh work request",
+          "name": "prompt-file",
+          "required": true,
+          "hasDynamicHelp": false,
+          "multiple": false,
+          "type": "option"
+        },
+        "model": {
+          "description": "Optional model constraint; must match the experiment pin",
+          "name": "model",
+          "hasDynamicHelp": false,
+          "multiple": false,
+          "type": "option"
+        }
+      },
+      "hasDynamicHelp": false,
+      "hiddenAliases": [],
+      "id": "experiments:enroll",
+      "pluginAlias": "@rulemetric/cli",
+      "pluginName": "@rulemetric/cli",
+      "pluginType": "core",
+      "strict": true,
+      "enableJsonFlag": false,
+      "isESM": true,
+      "relativePath": [
+        "dist",
+        "commands",
+        "experiments",
+        "enroll.js"
+      ]
+    },
+    "experiments:list": {
+      "aliases": [],
+      "args": {},
+      "description": "List experiments with execution, metric finding, and application separately",
+      "examples": [
+        "<%= config.bin %> experiments list",
+        "<%= config.bin %> experiments list --all"
+      ],
+      "flags": {
+        "api-url": {
+          "description": "Override API base URL",
+          "env": "RULEMETRIC_API_URL",
+          "hidden": true,
+          "name": "api-url",
+          "hasDynamicHelp": false,
+          "multiple": false,
+          "type": "option"
+        },
+        "project": {
+          "char": "p",
+          "description": "Project path (defaults to current directory)",
+          "name": "project",
+          "hasDynamicHelp": false,
+          "multiple": false,
+          "type": "option"
+        },
+        "all": {
+          "description": "Every project",
+          "name": "all",
+          "allowNo": false,
+          "type": "boolean"
+        }
+      },
+      "hasDynamicHelp": false,
+      "hiddenAliases": [],
+      "id": "experiments:list",
+      "pluginAlias": "@rulemetric/cli",
+      "pluginName": "@rulemetric/cli",
+      "pluginType": "core",
+      "strict": true,
+      "enableJsonFlag": false,
+      "isESM": true,
+      "relativePath": [
+        "dist",
+        "commands",
+        "experiments",
+        "list.js"
+      ]
+    },
     "evals:agent": {
       "aliases": [],
       "args": {},
@@ -980,11 +1094,9 @@
           "required": true
         }
       },
-      "description": "Run evals in a loop until the effect verdict is statistically powered (§6f auto-run-to-power)",
+      "description": "Retired adaptive loop; use evals run for one bounded experiment",
       "examples": [
-        "<%= config.bin %> evals auto-run <target-id>            # preview (cost + plan)",
-        "<%= config.bin %> evals auto-run <target-id> --yes      # run until powered / budget",
-        "<%= config.bin %> evals auto-run <target-id> --yes --max-runs 120 --batch-parallel 4"
+        "<%= config.bin %> evals run <target-id>"
       ],
       "flags": {
         "api-url": {
@@ -1127,9 +1239,9 @@
           "required": true
         }
       },
-      "description": "Blind A/B comparison between two eval target versions",
+      "description": "Retired blind comparison; use evals run for recorded task outcomes",
       "examples": [
-        "<%= config.bin %> evals compare <target-a> <target-b>"
+        "<%= config.bin %> evals run <target-id>"
       ],
       "flags": {
         "api-url": {
@@ -1254,10 +1366,9 @@
           "required": true
         }
       },
-      "description": "Optimize eval target description for better triggering accuracy",
+      "description": "Retired optimization workflow; use canonical experiments",
       "examples": [
-        "<%= config.bin %> evals optimize-description <target-id>",
-        "<%= config.bin %> evals optimize-description <target-id> --queries 10"
+        "<%= config.bin %> evals run <target-id>"
       ],
       "flags": {
         "api-url": {
@@ -1342,10 +1453,9 @@
           "required": true
         }
       },
-      "description": "Iteratively optimize eval target content using eval results as feedback",
+      "description": "Retired optimization workflow; use canonical experiments",
       "examples": [
-        "<%= config.bin %> evals optimize <target-id>",
-        "<%= config.bin %> evals optimize <target-id> --auto --max-iterations 3"
+        "<%= config.bin %> evals run <target-id>"
       ],
       "flags": {
         "api-url": {
@@ -1482,7 +1592,7 @@
           "required": true
         }
       },
-      "description": "Run evals against a target, optionally grade results, and post to API",
+      "description": "Queue an experiment against a target using the configured worker",
       "examples": [
         "<%= config.bin %> evals run <target-id>",
         "<%= config.bin %> evals run <target-id> --parallel 3 --runs-per-eval 2",
@@ -1671,56 +1781,6 @@
         "validate-judge.js"
       ]
     },
-    "experiments:list": {
-      "aliases": [],
-      "args": {},
-      "description": "Show instructions under randomised test and what each sample supports",
-      "examples": [
-        "<%= config.bin %> experiments list",
-        "<%= config.bin %> experiments list --project /path/to/project",
-        "<%= config.bin %> experiments list --all"
-      ],
-      "flags": {
-        "api-url": {
-          "description": "Override API base URL",
-          "env": "RULEMETRIC_API_URL",
-          "hidden": true,
-          "name": "api-url",
-          "hasDynamicHelp": false,
-          "multiple": false,
-          "type": "option"
-        },
-        "project": {
-          "char": "p",
-          "description": "Project path (defaults to current directory)",
-          "name": "project",
-          "hasDynamicHelp": false,
-          "multiple": false,
-          "type": "option"
-        },
-        "all": {
-          "description": "Every project, not just this one",
-          "name": "all",
-          "allowNo": false,
-          "type": "boolean"
-        }
-      },
-      "hasDynamicHelp": false,
-      "hiddenAliases": [],
-      "id": "experiments:list",
-      "pluginAlias": "@rulemetric/cli",
-      "pluginName": "@rulemetric/cli",
-      "pluginType": "core",
-      "strict": true,
-      "enableJsonFlag": false,
-      "isESM": true,
-      "relativePath": [
-        "dist",
-        "commands",
-        "experiments",
-        "list.js"
-      ]
-    },
     "gateway:ensure": {
       "aliases": [],
       "args": {},
@@ -2833,6 +2893,141 @@
         "switch.js"
       ]
     },
+    "research:backfill": {
+      "aliases": [],
+      "args": {},
+      "description": "Scan every ~/.claude/projects transcript for past rate-limit (429) events and POST them as observations. Opt-in seed for the usage-research dataset — forward-only tail is the default capture path.",
+      "examples": [
+        "<%= config.bin %> research backfill",
+        "<%= config.bin %> research backfill --claude-dir /custom/.claude --dry-run"
+      ],
+      "flags": {
+        "api-url": {
+          "description": "Override API base URL",
+          "env": "RULEMETRIC_API_URL",
+          "hidden": true,
+          "name": "api-url",
+          "hasDynamicHelp": false,
+          "multiple": false,
+          "type": "option"
+        },
+        "claude-dir": {
+          "description": "Override the Claude Code projects directory (defaults to ~/.claude)",
+          "name": "claude-dir",
+          "noCacheDefault": true,
+          "hasDynamicHelp": false,
+          "multiple": false,
+          "type": "option"
+        },
+        "dry-run": {
+          "description": "Print observation count without POSTing",
+          "name": "dry-run",
+          "allowNo": false,
+          "type": "boolean"
+        }
+      },
+      "hasDynamicHelp": false,
+      "hiddenAliases": [],
+      "id": "research:backfill",
+      "pluginAlias": "@rulemetric/cli",
+      "pluginName": "@rulemetric/cli",
+      "pluginType": "core",
+      "strict": true,
+      "enableJsonFlag": false,
+      "isESM": true,
+      "relativePath": [
+        "dist",
+        "commands",
+        "research",
+        "backfill.js"
+      ]
+    },
+    "research:resume-scoped-outcomes": {
+      "aliases": [],
+      "args": {},
+      "description": "Finish uploading scoped headless outcomes whose receipt was preserved locally after the actor exited but whose capture or verification upload failed. Never re-runs the actor or re-measures the repository; ships the stored receipt and exact transcript only.",
+      "examples": [
+        "<%= config.bin %> research resume-scoped-outcomes",
+        "<%= config.bin %> research resume-scoped-outcomes --list"
+      ],
+      "flags": {
+        "api-url": {
+          "description": "Override API base URL",
+          "env": "RULEMETRIC_API_URL",
+          "hidden": true,
+          "name": "api-url",
+          "hasDynamicHelp": false,
+          "multiple": false,
+          "type": "option"
+        },
+        "list": {
+          "description": "Show the outcome store without uploading anything",
+          "name": "list",
+          "allowNo": false,
+          "type": "boolean"
+        }
+      },
+      "hasDynamicHelp": false,
+      "hidden": true,
+      "hiddenAliases": [],
+      "id": "research:resume-scoped-outcomes",
+      "pluginAlias": "@rulemetric/cli",
+      "pluginName": "@rulemetric/cli",
+      "pluginType": "core",
+      "strict": true,
+      "enableJsonFlag": false,
+      "isESM": true,
+      "relativePath": [
+        "dist",
+        "commands",
+        "research",
+        "resume-scoped-outcomes.js"
+      ]
+    },
+    "research:tail-transcript": {
+      "aliases": [],
+      "args": {
+        "transcript": {
+          "description": "Absolute path to a Claude Code .jsonl transcript",
+          "name": "transcript",
+          "required": true
+        }
+      },
+      "description": "Read new lines from a Claude Code transcript since the last cached offset, POST any 429 events, and persist the new offset. Invoked by the session-end hook — safe to re-run.",
+      "flags": {
+        "api-url": {
+          "description": "Override API base URL",
+          "env": "RULEMETRIC_API_URL",
+          "hidden": true,
+          "name": "api-url",
+          "hasDynamicHelp": false,
+          "multiple": false,
+          "type": "option"
+        },
+        "no-post": {
+          "description": "Parse and advance the offset without POSTing (debug)",
+          "name": "no-post",
+          "allowNo": false,
+          "type": "boolean"
+        }
+      },
+      "hasDynamicHelp": false,
+      "hidden": true,
+      "hiddenAliases": [],
+      "id": "research:tail-transcript",
+      "pluginAlias": "@rulemetric/cli",
+      "pluginName": "@rulemetric/cli",
+      "pluginType": "core",
+      "strict": true,
+      "enableJsonFlag": false,
+      "isESM": true,
+      "relativePath": [
+        "dist",
+        "commands",
+        "research",
+        "tail-transcript.js"
+      ]
+    },
     "packs:add-instruction": {
       "aliases": [],
       "args": {
@@ -3560,141 +3755,6 @@
         "stop.js"
       ]
     },
-    "research:backfill": {
-      "aliases": [],
-      "args": {},
-      "description": "Scan every ~/.claude/projects transcript for past rate-limit (429) events and POST them as observations. Opt-in seed for the usage-research dataset — forward-only tail is the default capture path.",
-      "examples": [
-        "<%= config.bin %> research backfill",
-        "<%= config.bin %> research backfill --claude-dir /custom/.claude --dry-run"
-      ],
-      "flags": {
-        "api-url": {
-          "description": "Override API base URL",
-          "env": "RULEMETRIC_API_URL",
-          "hidden": true,
-          "name": "api-url",
-          "hasDynamicHelp": false,
-          "multiple": false,
-          "type": "option"
-        },
-        "claude-dir": {
-          "description": "Override the Claude Code projects directory (defaults to ~/.claude)",
-          "name": "claude-dir",
-          "noCacheDefault": true,
-          "hasDynamicHelp": false,
-          "multiple": false,
-          "type": "option"
-        },
-        "dry-run": {
-          "description": "Print observation count without POSTing",
-          "name": "dry-run",
-          "allowNo": false,
-          "type": "boolean"
-        }
-      },
-      "hasDynamicHelp": false,
-      "hiddenAliases": [],
-      "id": "research:backfill",
-      "pluginAlias": "@rulemetric/cli",
-      "pluginName": "@rulemetric/cli",
-      "pluginType": "core",
-      "strict": true,
-      "enableJsonFlag": false,
-      "isESM": true,
-      "relativePath": [
-        "dist",
-        "commands",
-        "research",
-        "backfill.js"
-      ]
-    },
-    "research:resume-scoped-outcomes": {
-      "aliases": [],
-      "args": {},
-      "description": "Finish uploading scoped headless outcomes whose receipt was preserved locally after the actor exited but whose capture or verification upload failed. Never re-runs the actor or re-measures the repository; ships the stored receipt and exact transcript only.",
-      "examples": [
-        "<%= config.bin %> research resume-scoped-outcomes",
-        "<%= config.bin %> research resume-scoped-outcomes --list"
-      ],
-      "flags": {
-        "api-url": {
-          "description": "Override API base URL",
-          "env": "RULEMETRIC_API_URL",
-          "hidden": true,
-          "name": "api-url",
-          "hasDynamicHelp": false,
-          "multiple": false,
-          "type": "option"
-        },
-        "list": {
-          "description": "Show the outcome store without uploading anything",
-          "name": "list",
-          "allowNo": false,
-          "type": "boolean"
-        }
-      },
-      "hasDynamicHelp": false,
-      "hidden": true,
-      "hiddenAliases": [],
-      "id": "research:resume-scoped-outcomes",
-      "pluginAlias": "@rulemetric/cli",
-      "pluginName": "@rulemetric/cli",
-      "pluginType": "core",
-      "strict": true,
-      "enableJsonFlag": false,
-      "isESM": true,
-      "relativePath": [
-        "dist",
-        "commands",
-        "research",
-        "resume-scoped-outcomes.js"
-      ]
-    },
-    "research:tail-transcript": {
-      "aliases": [],
-      "args": {
-        "transcript": {
-          "description": "Absolute path to a Claude Code .jsonl transcript",
-          "name": "transcript",
-          "required": true
-        }
-      },
-      "description": "Read new lines from a Claude Code transcript since the last cached offset, POST any 429 events, and persist the new offset. Invoked by the session-end hook — safe to re-run.",
-      "flags": {
-        "api-url": {
-          "description": "Override API base URL",
-          "env": "RULEMETRIC_API_URL",
-          "hidden": true,
-          "name": "api-url",
-          "hasDynamicHelp": false,
-          "multiple": false,
-          "type": "option"
-        },
-        "no-post": {
-          "description": "Parse and advance the offset without POSTing (debug)",
-          "name": "no-post",
-          "allowNo": false,
-          "type": "boolean"
-        }
-      },
-      "hasDynamicHelp": false,
-      "hidden": true,
-      "hiddenAliases": [],
-      "id": "research:tail-transcript",
-      "pluginAlias": "@rulemetric/cli",
-      "pluginName": "@rulemetric/cli",
-      "pluginType": "core",
-      "strict": true,
-      "enableJsonFlag": false,
-      "isESM": true,
-      "relativePath": [
-        "dist",
-        "commands",
-        "research",
-        "tail-transcript.js"
-      ]
-    },
     "service:install": {
       "aliases": [],
       "args": {},
@@ -3744,6 +3804,15 @@
           "allowNo": false,
           "type": "boolean"
         },
+        "auto-update": {
+          "dependsOn": [
+            "worker-only"
+          ],
+          "description": "Opt this machine into automatic stable CLI updates after active worker jobs finish (requires --worker-only)",
+          "name": "auto-update",
+          "allowNo": false,
+          "type": "boolean"
+        },
         "worker-only": {
           "description": "Install ONLY the job worker as a persistent service, targeting the globally-installed CLI (no repo checkout, no build, no API/gateway/proxy). Sources auth from ~/.config/rulemetric/env. macOS (launchd) + Linux (systemd --user).",
           "name": "worker-only",
@@ -3829,6 +3898,60 @@
         "uninstall.js"
       ]
     },
+    "service:updates": {
+      "aliases": [],
+      "args": {},
+      "description": "Show or change automatic worker updates on this machine (off by default)",
+      "examples": [
+        "<%= config.bin %> service updates",
+        "<%= config.bin %> service updates --enable",
+        "<%= config.bin %> service updates --disable"
+      ],
+      "flags": {
+        "api-url": {
+          "description": "Override API base URL",
+          "env": "RULEMETRIC_API_URL",
+          "hidden": true,
+          "name": "api-url",
+          "hasDynamicHelp": false,
+          "multiple": false,
+          "type": "option"
+        },
+        "enable": {
+          "description": "Install stable npm releases automatically after active jobs finish",
+          "exclusive": [
+            "disable"
+          ],
+          "name": "enable",
+          "allowNo": false,
+          "type": "boolean"
+        },
+        "disable": {
+          "description": "Stop future automatic updates",
+          "exclusive": [
+            "enable"
+          ],
+          "name": "disable",
+          "allowNo": false,
+          "type": "boolean"
+        }
+      },
+      "hasDynamicHelp": false,
+      "hiddenAliases": [],
+      "id": "service:updates",
+      "pluginAlias": "@rulemetric/cli",
+      "pluginName": "@rulemetric/cli",
+      "pluginType": "core",
+      "strict": true,
+      "enableJsonFlag": false,
+      "isESM": true,
+      "relativePath": [
+        "dist",
+        "commands",
+        "service",
+        "updates.js"
+      ]
+    },
     "sessions:analyze": {
       "aliases": [],
       "args": {
@@ -4815,5 +4938,5 @@
       ]
     }
   },
-  "version": "0.18.1"
+  "version": "0.19.1"
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@rulemetric/cli",
-  "version": "0.18.1",
+  "version": "0.19.1",
   "publishConfig": {
     "access": "public"
   },
@@ -80,9 +80,9 @@
     "postgres": "^3.4.0",
     "react": "^19.0.0",
     "zod": "^3.24.0",
-    "@rulemetric/hooks": "0.18.1",
-    "@rulemetric/proxy": "0.18.1",
-    "@rulemetric/skills-registry": "0.18.1"
+    "@rulemetric/hooks": "0.19.1",
+    "@rulemetric/skills-registry": "0.19.1",
+    "@rulemetric/proxy": "0.19.1"
   },
   "//deps": "Worker (`evals agent`) runtime deps: the @opentelemetry/* set, @sentry/node, and postgres are STATIC top-level imports loaded at worker startup (via bundled @rulemetric/{telemetry,db}) — they MUST be runtime deps or a clean `npm i -g` crashes with ModuleLoadError. `better-sqlite3` is optionalDependencies (below): it's a native module, lazy `require()`'d inside a try/catch (packages/session providers cursor/opencode) so a missing binary degrades gracefully — making it a hard dep would break `npm i -g` on platforms without prebuilt binaries. `resend` stays a devDep: lazy `await import('resend')` guarded by RESEND_API_KEY, which end-user workers never set (provider falls back to noop). Classification verified via import-graph scan of dist/commands/evals/agent.js + per-importer static/lazy check.",
   "optionalDependencies": {
@@ -102,15 +102,15 @@
     "tsx": "^4.19.0",
     "typescript": "^5.9.0",
     "vitest": "^3.2.0",
-    "@rulemetric/contracts": "0.0.1",
-    "@rulemetric/api": "0.0.1",
     "@rulemetric/core": "0.0.1",
+    "@rulemetric/contracts": "0.0.1",
+    "@rulemetric/db": "0.0.1",
     "@rulemetric/effectiveness": "0.0.1",
     "@rulemetric/email": "0.0.1",
     "@rulemetric/session": "0.2.1",
     "@rulemetric/telemetry": "0.0.1",
     "@rulemetric/typescript-config": "0.0.0",
-    "@rulemetric/db": "0.0.1"
+    "@rulemetric/api": "0.0.1"
   },
   "scripts": {
     "build": "node scripts/build.mjs",

package/dist/chunk-2C7CFCMM.js

@@ -1 +0,0 @@
-import{c as X}from"./chunk-HOQPVOMF.js";import{b as D}from"./chunk-GIRPEMTW.js";import{a as x}from"./chunk-MODXEGCU.js";import{e as J,g as B}from"./chunk-GWQ24FQP.js";import{e as w,f as R}from"./chunk-5VWVQCNR.js";import{c as H,d as U}from"./chunk-MHLHLJ5T.js";import{a as u}from"./chunk-ZLLER3HP.js";import{readFileSync as q,realpathSync as V}from"node:fs";import y from"node:path";import A from"node:os";function Q(s){let n=y.resolve(s);if(n===y.resolve(A.homedir()))return!1;let a=y.parse(n).root;if(n===a)return!1;try{let o=V(A.tmpdir());if(V(n)===o)return!1}catch{}return!0}u(Q,"isScorableProjectPath");var Z=5,ee=Number(process.env.RULEMETRIC_MAX_SUGGESTION_PROJECTS??10),te=48,ne=.5,O=400,oe=3,re=.6,se=1;function ae(s){let n=s.bloated??[],a=new Set([...s.harmful.map(o=>o.id),...s.candidates.map(o=>o.id)]);return n.filter(o=>!a.has(o.id)).slice(0,se).map(o=>({instructionId:o.id,estimatedTokens:o.estimatedTokens,topProjectId:o.topProjectId??null,score:.55,reason:`Costs ~${o.estimatedTokens} tokens every session and measures neutral across ${o.exposures} exposed session(s) \u2014 enough data to detect an effect, and none found. Consider shortening "${o.name}" rather than removing it.`+(o.containsSafetyLanguage?" CAUTION: it contains guardrail language (destructive-command or credential rules) \u2014 preserve those lines when shortening.":"")}))}u(ae,"buildTrimProposals");async function ie(s,n,a){let o=[],i=0;for(let d of s)try{await a("/api/instruction-experiments",{instructionId:d.instructionId,projectPath:n,source:"groom"}),i++}catch(g){let m=g instanceof Error?g.message:String(g);o.push({...d,reason:`${d.reason} (not measurable by withholding: ${m})`})}return{nominated:i,unmeasurable:o}}u(ie,"nominateRemovals");function ce(s){let n=[];for(let o of s.harmful){let i=o.verdict.delta;n.push({instructionId:o.id,score:Math.min(.95,.75+Math.abs(i)),topProjectId:o.topProjectId??null,reason:`Measured harmful: sessions using "${o.name}" scored ${(i*100).toFixed(1)}pp worse across ${o.exposures} exposed session(s). Observational \u2014 review before archiving.`})}let a=new Set(s.harmful.map(o=>o.id));for(let o of s.candidates)a.has(o.id)||n.push({instructionId:o.id,score:re,topProjectId:o.topProjectId??null,reason:`Unused: "${o.name}" is ${o.reason} (${o.totalExposures} lifetime exposure(s)) but still costs context budget every session.`});return n.slice(0,oe)}u(ce,"buildRemovalProposals");function le(){let s=y.join(A.homedir(),".config","rulemetric","skills-registry.json");try{let n=q(s,"utf-8"),a=JSON.parse(n),o=new Date(a.generatedAt).getTime();return Number.isNaN(o)?"unknown":Date.now()-o>te*60*60*1e3?"stale":"fresh"}catch{return"unknown"}}u(le,"observeSkillsRegistry");var Re=u(async(s,n)=>{let a=J(),o=a!=="none";o?n.logger.info(`Scoring backend: ${a}`):n.logger.warn("No LLM backend available (claude CLI missing and ANTHROPIC_API_KEY unset); skipping relevance scoring. Removal/trim proposals still run \u2014 they need no LLM.");let i;try{i=(await w("/api/sessions/project-paths?limit="+ee)).projectPaths}catch(e){n.logger.warn(`Failed to fetch project paths: ${e instanceof Error?e.message:String(e)}`);return}if(i.length===0){n.logger.info("No recent project paths; nothing to score");return}let d=new Map;try{let e=await w("/api/projects?limit=200");for(let t of e.projects??[])t.primaryCheckoutPath&&d.set(x(t.primaryCheckoutPath),t.id)}catch(e){n.logger.warn(`Failed to fetch projects for id resolution: ${e instanceof Error?e.message:String(e)}`)}let g=new Map;try{let e=await w(`/api/catalog?limit=200&proposable=1&contentPreview=${O}`);for(let t of e)g.set(t.id,{id:t.id,name:t.name,description:t.description??"",tags:t.tags??[],contentPreview:(t.contentPreview??"").slice(0,O)})}catch(e){n.logger.warn(`Failed to fetch catalog: ${e instanceof Error?e.message:String(e)}`)}try{let e=await w("/api/instructions?limit=100");for(let t of e.instructions??[])g.has(t.id)||g.set(t.id,{id:t.id,name:t.name,description:t.description??"",tags:t.tags??[],contentPreview:(t.content??"").slice(0,O)})}catch(e){n.logger.warn(`Failed to fetch personal instructions: ${e instanceof Error?e.message:String(e)}`)}let m=[],p=[];try{let e=await w("/api/insights/groom-candidates");if(!e.captureHealthy)n.logger.info(`Grooming paused: ${e.reason??"capture unhealthy"}`);else{m=ce(e),p=ae(e);let t=e.safetyGuarded+e.harmfulSafetyGuarded;n.logger.info(`Removal proposals: ${m.length} (${e.harmful.length} measured-harmful, ${e.candidates.length} never-used, ${t} withheld by the safety allowlist); trim proposals: ${p.length} (${(e.bloated??[]).length} measured-neutral and expensive, ${e.trimSafetyFlagged??0} flagged as containing guardrail language)`)}}catch(e){n.logger.warn(`Failed to fetch groom candidates: ${e instanceof Error?e.message:String(e)}`)}let b=[...g.values()];if(b.length===0&&m.length===0&&p.length===0){n.logger.info("No candidate instructions and nothing to retire; nothing to score");return}n.logger.info(`Scoring ${b.length} candidates across ${i.length} project(s)`);let M=0,l=null,I=new Set;for(let e of i)try{let t=x(e);if(!Q(t)){n.logger.info(`Skipping ${t}: not a project directory (home/root)`);continue}let f=D(t),k=o&&f.length>0;f.length===0&&n.logger.info(`${t}: insufficient language signals; removal/trim only`);let v=X(t),z={name:y.basename(t),description:v.description,languages:f,topics:v.topics,dependencies:v.dependencies},E=k?U(b):[];k&&n.logger.info(`Scoring ${b.length} candidates for ${t} in ${E.length} batch(es)`);let C=[],$=0;for(let r of E)try{let c=await B(H(z,r),"",4096);if(!c){$++;continue}let T=c.recommendations??[];for(let P of T)typeof P.instructionId=="string"&&typeof P.score=="number"&&typeof P.reason=="string"&&C.push({instructionId:P.instructionId,score:P.score,reason:P.reason})}catch(c){$++,n.logger.warn(`Batch failed for ${t}: ${c instanceof Error?c.message:String(c)}`)}$>0&&n.logger.warn(`${$}/${E.length} batch(es) failed for ${t} (backend ${a}); suggestions are partial`),k&&C.length===0&&n.logger.warn(`No usable LLM output for ${t} (backend ${a})`);let K=C.filter(r=>r.score>=ne).sort((r,c)=>c.score-r.score).slice(0,Z),_=d.get(t),W=l===null||l===t;l===null&&(l=t);let N=u(r=>r.topProjectId?r.topProjectId===_:W,"placeHere"),F=m.filter(N),{nominated:L,unmeasurable:G}=await ie(F,t,R);L>0&&n.logger.info(`${t}: ${L} removal candidate(s) nominated for randomised withholding instead of the feed`);let h=[...K.map(r=>({instructionId:r.instructionId,score:r.score,reason:r.reason,kind:"add"})),...G.map(({topProjectId:r,...c})=>({...c,kind:"remove"})),...p.filter(N).map(({estimatedTokens:r,topProjectId:c,...T})=>({...T,kind:"trim"}))];if(h.length===0){n.logger.info(`Nothing to suggest for ${t}`);continue}await R("/api/instruction-suggestions/batch-upsert",{projectPath:t,..._?{projectId:_}:{},suggestions:h}),n.logger.info(`${t}: ${h.filter(r=>r.kind==="add").length} add + ${h.filter(r=>r.kind==="remove").length} remove + ${h.filter(r=>r.kind==="trim").length} trim`);for(let r of h)r.kind!=="add"&&I.add(r.instructionId);let Y=new Set(G.map(r=>r.instructionId));for(let r of F)Y.has(r.instructionId)||I.add(r.instructionId);M++}catch(t){n.logger.warn(`Failed to score ${e}: ${t instanceof Error?t.message:String(t)}`)}let j=[...m.filter(e=>!I.has(e.instructionId)),...p.filter(e=>!I.has(e.instructionId)).map(({estimatedTokens:e,...t})=>t)];if(j.length>0&&l)try{let e=new Set(p.map(t=>t.instructionId));await R("/api/instruction-suggestions/batch-upsert",{projectPath:l,...d.get(l)?{projectId:d.get(l)}:{},suggestions:j.map(({topProjectId:t,...f})=>({...f,kind:e.has(f.instructionId)?"trim":"remove"}))}),n.logger.info(`Orphan sweep: ${j.length} proposal(s) whose project was not scored this run placed on ${l}`)}catch(e){n.logger.warn(`Orphan sweep failed: ${e instanceof Error?e.message:String(e)}`)}let S=le();if(S==="unknown")n.logger.warn("Skills registry unreadable or has no usable generatedAt; leaving suggestion staleness unchanged");else try{await R("/api/instruction-suggestions/mark-stale",{stale:S==="stale"}),S==="stale"?n.logger.warn("Skills registry is stale (>48h); marked catalog-skill suggestions stale"):n.logger.info("Skills registry is fresh; cleared staleness on catalog-skill suggestions")}catch{}n.logger.info(`Done: scored ${M}/${i.length} projects`)},"cronSuggestInstructions");export{Q as a,ae as b,ie as c,ce as d,Re as e};

package/dist/chunk-647M2I2Y.js

@@ -1 +0,0 @@
-import{a as s}from"./chunk-ZLLER3HP.js";function w(r){if(typeof r!="string")return null;let n=r.toUpperCase();return n==="A"||n==="B"||n==="TIE"?n:null}s(w,"normalizeWinner");function v(r){return Array.isArray(r)?r.filter(n=>typeof n=="string"):[]}s(v,"stringList");function d(r){let n=r.trim(),a=n.match(/```(?:json)?\s*([\s\S]*?)```/);a&&(n=a[1].trim());let e=n.match(/\{[\s\S]*\}/);e&&(n=e[0]);let t;try{t=JSON.parse(n)}catch{return null}if(typeof t!="object"||t===null)return null;let{per_eval:l,overall:g}=t,i=typeof g=="object"&&g!==null?g:{},u=w(i.winner);return u?{perEval:(Array.isArray(l)?l:[]).flatMap(p=>{if(typeof p!="object"||p===null)return[];let o=p,c=w(o.winner);return!c||typeof o.eval_name!="string"?[]:[{evalName:o.eval_name,winner:c,reasoning:typeof o.reasoning=="string"?o.reasoning:""}]}),overall:{winner:u,reasoning:typeof i.reasoning=="string"?i.reasoning:"",strengthsWinner:v(i.strengths_winner),weaknessesLoser:v(i.weaknesses_loser)}}:null}s(d,"parseComparatorOutput");function m(r,n){let a=s(e=>{if(e==="TIE")return null;let t=n.isSwapped?n.targetBId:n.targetAId,l=n.isSwapped?n.targetAId:n.targetBId;return e==="A"?t:l},"resolve");return{winnerTargetId:a(r.overall.winner),perEval:r.perEval.map(e=>({evalName:e.evalName,winnerTargetId:a(e.winner),reasoning:e.reasoning})),overall:{reasoning:r.overall.reasoning,strengthsWinner:r.overall.strengthsWinner,weaknessesLoser:r.overall.weaknessesLoser}}}s(m,"resolveComparison");export{d as a,m as b};