analyzthis_design 2.5.1 → 2.6.0

package/dist/agents/session-schema.json

@@ -30,8 +30,9 @@
       "experts": ["string — persona ids"],
       "reason": "string"
     },
+    "active_run": "object | null — { id, mode, at }; feedback only accepts outputs carrying this run id",
     "persona_outputs": {
-      "<persona_id>": "object — structured output block per agents/manifests/<persona_id>.json"
+      "<persona_id>": "object — output plus run_id, attribution { provider, model, model_id, effort, policy_version, prompt_hash, validator_version }, quality, and accepted state"
     },
     "full_prompts": {
       "<persona_id>": "object — { system: string, user: string } exact prompts sent to LLM for training"
@@ -73,6 +74,8 @@
         "correction": "string — how the user corrected or what they wanted instead",
         "tags": ["string — e.g. wrong_hierarchy, invented_tokens, off_brief"],
         "original_output": "string — snapshot of persona output at feedback time",
+        "run_id": "string | null — exact run that produced the rated output",
+        "attribution": "object — provider/model/effort and policy metadata copied from the output",
         "submitted_at": "ISO 8601 | omitted — set when feedback submit succeeds",
         "context": "object — task_map_summary, problem_type, mode"
       }
@@ -83,6 +86,13 @@
     },
     "vault_path": "string — Obsidian vault written by Kavi collect (empty until first collect)",
     "last_collect_at": "ISO 8601 timestamp — last successful collect run",
+    "installed_ui_libraries": {
+      "installed": [{ "library": "string", "job": "string", "how": "package.json | folder | package+folder" }],
+      "missing_defaults": [{ "job": "string", "library": "string", "install": "string", "docs": "string" }],
+      "full_kits": ["string — Chakra, Mantine, HeroUI, MUI, Ant Design, Ark"],
+      "mixed_kits": "boolean — more than one full kit, or a full kit plus Shadcn",
+      "folders": ["string — components/ui, components/magicui, ..."]
+    },
     "deliberation": {
       "round": "number — current deliberation round",
       "max_rounds": "number",
@@ -141,7 +151,7 @@
       "output_tokens_est": "number — estimated output tokens across the last run",
       "cache_hits": "number — retrieve/cache hits during the last run",
       "mode": "lite | full",
-      "effort_log": [{ "persona": "string", "effort": "trivial|standard|hard", "model": "string|null", "scoped_mode": "string|null" }],
+      "effort_log": [{ "persona": "string", "effort": "trivial|standard|hard", "provider": "string|null", "model": "string|null", "model_id": "string|null", "input_tokens": "number", "output_tokens": "number", "run_id": "string", "scoped_mode": "string|null" }],
       "cost_usd": "number — estimated $ cost of the run from config.pricing (0 when pricing missing)",
       "deliberation_rounds": "number — adversarial rounds executed in last run",
       "objections_raised": "number",

package/dist/bin/cli.js

@@ -19,6 +19,7 @@ const designSpec = require('../lib/design-spec');
 const { run: orchestratorRun } = require('../lib/orchestrator/run');
 const hostLlm = require('../lib/host-llm');
 const moodboard = require('../lib/moodboard');
+const calibrator = require('../lib/calibrate');
 
 const HELP = `
 Analyzthis_Design — 8 AI design personas with a personal knowledge bank
@@ -79,6 +80,9 @@ Usage:
 ── Efficiency / cost commands ────────────────────────────────
   metrics             Print the last run's cost metrics (--all for every project)
   cost                Print the last run's $ cost from config.pricing (--all for every project)
+  calibrate           Compare available models on the fixed persona benchmark
+                      (--max-usd 5, --max-calls 500, --runs 3, --judge, --apply)
+                      Preview only with --dry-run; share text-free results with --share
   receipt             Inferred token receipt for host (slash/MCP) turns — not a bill
   export-training     Write accepted-output JSONL pairs for LoRA prep
                       (--persona <id>, --all, --output <path>)
@@ -98,12 +102,26 @@ Usage:
                             full: all 19 named tools. Or ANALYZTHIS_MCP_CATALOG=full
 
 ── Mood board commands ─────────────────────────────────────
-  moodboard create --task <text>   Build a mood board from web + DS references
-                                    [--url <url> ...] [--auto]
-  moodboard critique --board <id>  Run team deliberation over a board
-  moodboard add --board <id>       Add a user reference and rerun critique
-                                    [--url <url>] [--title <t>] [--desc <d>]
-                                    [--tags a,b] [--note <text>]
+  moodboard create --task <text>   Start a visual mood board (Zara scouts element crops)
+                                    [--adjectives a,b,c] [--not "..."] [--sections components,typography]
+                                    [--kind mood] for the older link-and-tag board
+  moodboard brief --board <id>     Edit the brief [--adjectives] [--not] [--sections] [--yes]
+  moodboard capture-script          Print the page script that reads real colours/fonts
+                                    [--selector ".pricing-card"]
+  moodboard tile --board <id>      Add one element crop: --page <url> --image <file|url>
+                                    --element "<what>" --section <id> [--steal] [--refuse]
+                                    [--peak] [--colors #hex,..] [--fonts a,b] [--concept]
+                                    [--video film.webm] [--stills a.png,b.png]
+  moodboard film --board <id>      Record real sites scrolling as motion tiles: --url <url>
+                                    (repeatable) [--section motion] [--steal] [--refuse]
+                                    Needs playwright-core + Google Chrome
+  moodboard critique --board <id>  Team votes on each tile (advice only)
+  moodboard votes --board <id>     Save team votes from a JSON file --votes votes.json
+  moodboard show --board <id>      Open the clickable board on 127.0.0.1 [--port 4317]
+  moodboard decide --board <id>    Your choice: --tile <n> --decision keep|maybe|reject|clear
+  moodboard rescout --board <id>   Send Zara back with your note [--note "..."]
+  moodboard lock --board <id>      Lock your direction for every persona [--note "..."]
+  moodboard add --board <id>       Add a reference to a link board and rerun critique
   moodboard list                   List mood boards for this project
 
 ── Reference data (retrieve-on-demand) ───────────────────────
@@ -156,6 +174,14 @@ Usage:
   --budget   free | cheap | auto — chunk model budget (run command; default: auto)
   --sequential  Run chunks sequentially (run command; default on)
   --max-chunks  Cap number of planner chunks (run command; default 6)
+  --max-usd  Calibration hard spending limit in USD (default 5)
+  --max-calls Calibration hard model-call limit (default 500)
+  --runs     Calibration repeats per fixed task/model (default 3)
+  --delta    Non-inferiority margin on the 0-1 score scale (default 0.05)
+  --judge    Enable blind cross-family calibration judges
+  --apply    Write qualified calibration winners to the local policy
+  --resume   Resume a calibration checkpoint path
+  --share    Opt-in submission of text-free fixed-benchmark aggregates
   --unchunked   Force legacy single-pass orchestrator (run command)
   --output   Write final session-state.json to this path (run / run-unchunked command); or JSONL path (export-training)
   --persona  Persona id (session accept; export-training; feedback commands)
@@ -210,7 +236,10 @@ Usage:
   npx analyzthis_design research --query "EY design system tokens"
 
   npx analyzthis_design moodboard create --task "B2B fintech dashboard, trustworthy, high-contrast" --auto
+  npx analyzthis_design moodboard create --task "Invoice approval, partner signs" --adjectives calm,precise,trusted --not "AI purple hero"
+  npx analyzthis_design moodboard tile --board <id> --page https://linear.app/pricing --image ./crop.png --element "Pricing card" --section components --steal "one quiet CTA"
   npx analyzthis_design moodboard critique --board <boardId>
+  npx analyzthis_design moodboard show --board <boardId>
   npx analyzthis_design moodboard add --board <boardId> --url https://dribbble.com/shots/example --title "Alt hero" --tags "landing,trust"
 
   npx analyzthis_design retrieve --file colors.csv --column "Product Type" --keywords saas,dashboard
@@ -326,6 +355,10 @@ const boardVal     = getFlag('board');
 const titleVal     = getFlag('title');
 const descVal      = getFlag('desc');
 const autoFlag     = flags.includes('--auto');
+const kindVal      = getFlag('kind');
+const stealVal     = getFlag('steal');
+const refuseVal    = getFlag('refuse');
+const peakVal      = getFlag('peak');
 const budgetVal    = getFlag('budget');
 const sequentialFlag = flags.includes('--sequential');
 const maxChunksVal = getFlag('max-chunks');
@@ -335,6 +368,32 @@ const fixFlag = flags.includes('--fix') || flags.includes('--no');
 const likedFlag = flags.includes('--liked');
 const becauseVal = getFlag('because');
 const sendFlag = flags.includes('--send');
+const applyFlag = flags.includes('--apply');
+const shareFlag = flags.includes('--share');
+const judgeFlag = flags.includes('--judge');
+const maxUsdVal = getFlag('max-usd');
+const maxCallsVal = getFlag('max-calls');
+const runsVal = getFlag('runs');
+const deltaVal = getFlag('delta');
+const resumeVal = getFlag('resume');
+const sectionVal = getFlag('section');
+const sectionsVal = getFlag('sections');
+const elementVal = getFlag('element');
+const imageVal = getFlag('image');
+const tileVal = getFlag('tile');
+const decisionVal = getFlag('decision');
+const noteVal = getFlag('note');
+const portVal = getFlag('port');
+const selectorVal = getFlag('selector');
+const adjectivesVal = getFlag('adjectives');
+const notThisVal = getFlag('not');
+const colorsVal = getFlag('colors');
+const fontsVal = getFlag('fonts');
+const votesVal = getFlag('votes');
+const conceptFlag = flags.includes('--concept');
+const videoVal = getFlag('video');
+const stillsVal = getFlag('stills');
+const splitList = (raw) => (raw ? raw.split(',').map((s) => s.trim()).filter(Boolean) : undefined);
 
 // ─── Commands ────────────────────────────────────────────────────────────────
 
@@ -660,7 +719,7 @@ switch (cmd) {
     const msub = flags.find((f) => !f.startsWith('--')) || 'create';
     if (msub === 'create') {
       if (!taskVal) {
-        console.error('\n  ✗  --task is required.  Example: npx analyzthis_design moodboard create --task "B2B fintech dashboard" --auto\n');
+        console.error('\n  ✗  --task is required.  Example: npx analyzthis_design moodboard create --task "B2B fintech dashboard" --adjectives calm,dense,trusted\n');
         process.exit(1);
       }
       (async () => {
@@ -675,11 +734,23 @@ switch (cmd) {
             project: projectVal,
             task: taskVal,
             urls: urls,
+            kind: kindVal || 'ui',
             discover: autoFlag || urls.length > 0,
             dryRun: dryRunFlag,
+            brief: { adjectives: splitList(adjectivesVal), not_this: notThisVal || '', sections: splitList(sectionsVal) },
           });
-          console.log(`\n✅ Mood board created: ${result.board.board_id}`);
-          console.log(`   References: ${result.board.references.length}`);
+          console.log(`\n✅ ${result.board.kind === 'ui' ? 'Mood board (visual)' : 'Link mood board'} created: ${result.board.board_id}`);
+          if (result.board.kind === 'ui') {
+            console.log(`   Sections: ${result.board.brief.sections.join(', ')}`);
+            console.log(`   Scout:    Zara round ${result.board.scout.round} — ${result.board.scout.status}`);
+            console.log(`   Next:     Zara crops elements, then moodboard tile --board ${result.board.board_id} --page <url> --image <file> --element "<what>" --section components`);
+            console.log(`   Then:     moodboard show --board ${result.board.board_id}  (you decide in the browser)`);
+            console.log('');
+            console.log(result.board.scout.brief);
+            console.log('');
+          } else {
+            console.log(`   References: ${result.board.references.length}`);
+          }
           console.log(`   Board dir:  ${result.boardDir}`);
           console.log(`   Workspace:  ${result.workspacePath}\n`);
           if (dryRunFlag) {
@@ -712,6 +783,10 @@ switch (cmd) {
           });
           console.log(`\n✅ Mood board critique complete: ${result.board.board_id}`);
           console.log(`   Verdict: ${result.board.synthesis?.verdict || 'n/a'}`);
+          if (result.board.kind === 'ui' && result.board.scout) {
+            console.log(`   Scout:   round ${result.board.scout.round} — ${result.board.scout.status}`);
+            console.log(`   Team votes are advice. You decide: moodboard show --board ${result.board.board_id}`);
+          }
           console.log(`   Rounds:  ${result.board.deliberation?.round || 0}`);
           console.log(`   Board:   ${result.boardDir || path.join(session.sessionDir(projectVal || session.getProjectId()), 'moodboard', boardVal)}`);
           console.log(`   Workspace: ${result.workspacePath}\n`);
@@ -726,6 +801,149 @@ switch (cmd) {
       })();
       break;
     }
+    if (msub === 'film') {
+      const urls = flags.filter((f, i) => i > 0 && flags[i - 1] === '--url' && !f.startsWith('--'))
+        .concat(flags.filter((f) => f.startsWith('--url=')).map((f) => f.slice(6)));
+      if (!boardVal || !urls.length) {
+        console.error('\n  ✗  --board and --url are required.  Example: npx analyzthis_design moodboard film --board <id> --url https://osmo.supply --steal "easing taste"\n');
+        process.exit(1);
+      }
+      (async () => {
+        try {
+          console.log(`\n  🎬 Filming ${urls.length} site${urls.length === 1 ? '' : 's'} (about 15 seconds each)…`);
+          const out = await require('../lib/moodboard-film').filmTiles({
+            project: projectVal, boardId: boardVal, urls,
+            section: sectionVal, element: elementVal, steal: stealVal, refuse: refuseVal, peak: peakVal,
+          });
+          out.failed.forEach((f) => console.error(`  ✗  ${f.url}: ${f.reason}`));
+          const added = out.saved ? out.saved.added : [];
+          (out.saved ? out.saved.skipped : []).forEach((s) => console.error(`  ✗  ${s.element}: ${s.reason}`));
+          if (!added.length) process.exit(1);
+          console.log(`\n✅ ${added.length} film tile${added.length === 1 ? '' : 's'} added: ${added.map((t) => '#' + t.n + ' ' + t.site).join(', ')}`);
+          console.log(`   Next: moodboard show --board ${boardVal}\n`);
+        } catch (err) {
+          console.error(`\n  ✗  ${err.message}\n`);
+          process.exit(1);
+        }
+      })();
+      break;
+    }
+    if (msub === 'tile' || msub === 'frames') {
+      const pageUrl = getFlag('page') || urlVal;
+      if (!boardVal || !pageUrl) {
+        console.error('\n  ✗  --board and --page are required.  Example: npx analyzthis_design moodboard tile --board <id> --page https://linear.app/pricing --image ./crop.png --element "Pricing card" --section components --steal "one quiet CTA"\n');
+        process.exit(1);
+      }
+      try {
+        const result = moodboard.addTiles({
+          project: projectVal,
+          boardId: boardVal,
+          tile: {
+            page_url: pageUrl,
+            element: elementVal || titleVal || '',
+            section: sectionVal || 'components',
+            image_path: imageVal && !/^https?:\/\//i.test(imageVal) ? imageVal : undefined,
+            image_url: imageVal && /^https?:\/\//i.test(imageVal) ? imageVal : (msub === 'frames' ? urlVal : undefined),
+            video_path: videoVal,
+            stills: splitList(stillsVal),
+            tokens: { colors: splitList(colorsVal), fonts: splitList(fontsVal) },
+            steal: stealVal || '',
+            refuse: refuseVal || '',
+            peak: peakVal || '',
+            concept: conceptFlag,
+          },
+        });
+        if (result.skipped.length) {
+          console.error(`\n  ✗  Not added: ${result.skipped.map((s) => s.reason).join('; ')}\n`);
+          process.exit(1);
+        }
+        console.log(`\n✅ Tile #${result.added[0].n} added (${result.board.tiles.length} on the board)`);
+        console.log(`   Next: more tiles, then moodboard critique or moodboard show --board ${result.board.board_id}\n`);
+      } catch (err) {
+        console.error(`\n  ✗  ${err.message}\n`);
+        process.exit(1);
+      }
+      break;
+    }
+    if (msub === 'brief') {
+      try {
+        const result = moodboard.setBrief({
+          project: projectVal, boardId: boardVal,
+          adjectives: splitList(adjectivesVal), not_this: notThisVal, sections: splitList(sectionsVal),
+          confirmed_by_user: yesFlag ? true : undefined,
+        });
+        console.log(`\n✅ Brief saved: ${JSON.stringify(result.board.brief)}\n`);
+      } catch (err) {
+        console.error(`\n  ✗  ${err.message}\n`);
+        process.exit(1);
+      }
+      break;
+    }
+    if (msub === 'votes') {
+      try {
+        const raw = JSON.parse(fs.readFileSync(votesVal, 'utf8'));
+        const result = moodboard.recordPersonaVotes({ project: projectVal, boardId: boardVal, votes: raw.votes || [], verdict: raw.verdict, summary: raw.summary });
+        console.log(`\n✅ Team advice saved. ${result.ignored.length} ignored. Next: moodboard show --board ${result.board.board_id}\n`);
+      } catch (err) {
+        console.error(`\n  ✗  ${err.message}  (usage: moodboard votes --board <id> --votes votes.json)\n`);
+        process.exit(1);
+      }
+      break;
+    }
+    if (msub === 'decide') {
+      try {
+        moodboard.recordUserDecision({ project: projectVal, boardId: boardVal, tile: tileVal, decision: decisionVal, note: noteVal || '' });
+        console.log(`\n✅ Tile ${tileVal}: ${decisionVal}. Your choice overrides every persona.\n`);
+      } catch (err) {
+        console.error(`\n  ✗  ${err.message}  (usage: moodboard decide --board <id> --tile 3 --decision keep|maybe|reject|clear --note "...")\n`);
+        process.exit(1);
+      }
+      break;
+    }
+    if (msub === 'rescout') {
+      try {
+        const result = moodboard.requestRescout({ project: projectVal, boardId: boardVal, note: noteVal || '' });
+        console.log(`\n✅ Zara goes back out (round ${result.board.scout.round}).\n\n${result.board.scout.brief}\n`);
+      } catch (err) {
+        console.error(`\n  ✗  ${err.message}\n`);
+        process.exit(1);
+      }
+      break;
+    }
+    if (msub === 'lock') {
+      try {
+        const result = moodboard.lockDirection({ project: projectVal, boardId: boardVal, note: noteVal || '' });
+        const d = result.board.direction;
+        console.log(`\n✅ Direction locked: ${d.kept.length} kept, ${d.rejected.length} rejected. Saved to session as design_direction.`);
+        if (d.palette.length) console.log(`   Palette: ${d.palette.join(' ')}`);
+        if (d.fonts.length) console.log(`   Fonts:   ${d.fonts.join(', ')}`);
+        console.log('');
+      } catch (err) {
+        console.error(`\n  ✗  ${err.message}\n`);
+        process.exit(1);
+      }
+      break;
+    }
+    if (msub === 'show') {
+      if (!boardVal) {
+        console.error('\n  ✗  --board is required.  Example: npx analyzthis_design moodboard show --board <id>\n');
+        process.exit(1);
+      }
+      require('../lib/moodboard-view').serveBoard({ project: projectVal, boardId: boardVal, port: portVal != null ? portVal : undefined, onChange: (what, b) => {
+        console.log(`   ${new Date().toLocaleTimeString()}  ${what}  — scout: ${b.scout && b.scout.status}`);
+      } }).then((live) => {
+        console.log(`\n✅ Mood board live at ${live.url}`);
+        console.log('   Open it in the browser. Keep / maybe / reject saves instantly. Ctrl+C to stop.\n');
+      }).catch((err) => {
+        console.error(`\n  ✗  ${err.message}\n`);
+        process.exit(1);
+      });
+      break;
+    }
+    if (msub === 'capture-script') {
+      console.log(moodboard.captureScript(selectorVal || 'body', getFlag('text') || ''));
+      break;
+    }
     if (msub === 'add') {
       if (!boardVal) {
         console.error('\n  ✗  --board is required.  Example: npx analyzthis_design moodboard add --board <id> --url <url>\n');
@@ -776,7 +994,7 @@ switch (cmd) {
       }
       break;
     }
-    console.error('\n  Unknown moodboard subcommand. Use: create | critique | add | list\n');
+    console.error('\n  Unknown moodboard subcommand. Use: create | brief | tile | critique | votes | show | decide | rescout | lock | capture-script | add | list\n');
     process.exit(1);
   }
 
@@ -1000,6 +1218,86 @@ switch (cmd) {
     break;
   }
 
+  case 'calibrate': {
+    (async () => {
+      try {
+        if (shareFlag && dryRunFlag && resumeVal) {
+          const saved = JSON.parse(fs.readFileSync(path.resolve(resumeVal), 'utf8'));
+          const shared = await feedbackSubmit.submitCalibration(saved, { dryRun: true, yes: yesFlag });
+          console.log(`\n── Calibration share preview (${shared.rows.length} text-free rows) ──`);
+          console.log(JSON.stringify(shared.rows[0], null, 2));
+          console.log('');
+          return;
+        }
+
+        const options = {
+          maxUsd: maxUsdVal == null ? 5 : Number(maxUsdVal),
+          maxCalls: maxCallsVal == null ? 500 : Number(maxCallsVal),
+          runs: runsVal == null ? 3 : Number(runsVal),
+          delta: deltaVal == null ? 0.05 : Number(deltaVal),
+          judge: judgeFlag,
+          apply: applyFlag,
+          resume: resumeVal || null,
+        };
+        const preview = await calibrator.preview(options);
+        console.log('\n── Calibration preview ─────────────────────────────');
+        console.log(`  Models:          ${preview.models}`);
+        console.log(`  Fixed tasks:     ${preview.tasks}`);
+        console.log(`  Planned calls:   ${preview.estimated_calls}/${options.maxCalls}`);
+        console.log(`  Worst-case cost: $${preview.estimated_usd.toFixed(4)}/$${options.maxUsd.toFixed(2)}`);
+        if (!preview.allowed) {
+          console.error(`  Blocked by:      ${preview.reasons.join(', ')}\n`);
+          process.exitCode = 1;
+          return;
+        }
+        if (dryRunFlag) {
+          console.log('  Preview only: no model calls or policy writes.\n');
+          return;
+        }
+
+        if (!yesFlag) {
+          if (!process.stdin.isTTY) {
+            throw new Error('Non-interactive terminal. Pass --yes after reviewing the cost preview.');
+          }
+          const readline = require('readline');
+          const approved = await new Promise((resolve) => {
+            const rl = readline.createInterface({ input: process.stdin, output: process.stdout });
+            rl.question('Run this calibration? [y/N] ', (answer) => {
+              rl.close();
+              resolve(/^y(es)?$/i.test(String(answer || '').trim()));
+            });
+          });
+          if (!approved) {
+            console.log('\n  Calibration cancelled.\n');
+            return;
+          }
+        }
+
+        const result = await calibrator.calibrate(options);
+        console.log(`\n── Calibration ${result.status} ─────────────────────`);
+        console.log(`  Calls:      ${result.calls}`);
+        console.log(`  Cost:       $${Number(result.spent_usd || 0).toFixed(4)}`);
+        console.log(`  Checkpoint: ${result.checkpoint}`);
+        Object.entries(result.selections || {}).forEach(([persona, efforts]) => {
+          Object.entries(efforts).forEach(([effort, selection]) => {
+            console.log(`  ${persona}/${effort}: ${selection.winner || 'keep current'} (${selection.reason})`);
+          });
+        });
+        if (result.policy_file) console.log(`  Local policy: ${result.policy_file}`);
+
+        if (shareFlag) {
+          const shared = await feedbackSubmit.submitCalibration(result, { dryRun: false, yes: yesFlag });
+          console.log(`  Shared:      ${shared.submitted || 0} fixed-benchmark rows`);
+        }
+        console.log('');
+      } catch (error) {
+        console.error(`\n  ✗  ${error.message}\n`);
+        process.exitCode = 1;
+      }
+    })();
+    break;
+  }
+
   case 'scores': {
     const em = require('../lib/evolution-metrics');
     if (flags.includes('--submit') || flags.includes('--submit-dry-run')) {