@wundam/orchex 1.0.0-rc.20 → 1.0.0-rc.22

package/README.md

@@ -12,8 +12,8 @@ Your AI assistant does tasks one at a time. Orchex makes it do 10 at once — sa
 - **Ownership Enforcement** — Each stream can only modify files in its `owns` array. No two agents touch the same file. Zero conflicts.
 - **`orchex run`** — Describe what you want, get parallel execution. Auto-generates plans, previews waves, executes with ownership enforcement.
 - **`orchex learn`** — The advanced path. Paste a markdown plan, get executable parallel streams with dependency inference and anti-pattern detection.
-- **Self-Healing** — Categorized error analysis with targeted fix streams. Not blind retry.
-- **Multi-LLM** — OpenAI, Gemini, Claude, DeepSeek, Ollama, AWS Bedrock. Route different orchestrations to different providers.
+- **Self-Healing** — Categorized error analysis with targeted fix streams. Not blind retry. Model validation before execution prevents wasted API calls.
+- **Multi-LLM** — OpenAI, Gemini, Claude, DeepSeek, Ollama, AWS Bedrock. Dynamic model registry auto-discovers available models. Key-aware routing prevents "model not found" errors.
 - **BYOK** — Bring your own API key from any supported provider. You control costs.
 
 ## Prerequisites

package/dist/index.js

@@ -140,6 +140,14 @@ async function handleLoginCommand(args) {
                 catch {
                     // Non-fatal — entitlement sync is best-effort
                 }
+                // Also cache model registry for offline use
+                try {
+                    const { fetchAndCacheModels } = await import('./model-cache.js');
+                    await fetchAndCacheModels(apiUrl);
+                }
+                catch {
+                    // Non-fatal — model registry can be fetched on-demand later
+                }
                 console.log('');
                 return;
             }
@@ -537,7 +545,27 @@ async function handleRunCommand(args) {
     }
     const waveList = waves.map(w => ({ number: w.number, streams: w.streams.map(s => s.id) }));
     const allWarnings = [...result.warnings, ...diagnostics.warnings];
-    const preview = formatPlanPreview(initStreams, waveList, allWarnings, seqDiag);
+    // Validate models for each stream and build model decisions map
+    const { resolveModel } = await import('./model-validator.js');
+    const { resolveApiUrl, DEFAULT_MODELS } = await import('./config.js');
+    const { estimatePlannedCost } = await import('./intelligence/index.js');
+    const modelApiUrl = config.mode === 'cloud' && config.apiKey ? await resolveApiUrl() : undefined;
+    const modelDecisions = {};
+    for (const [id, stream] of Object.entries(initStreams)) {
+        const suggestedProvider = stream.suggestedProvider ?? resolvedProvider;
+        const suggestedModel = DEFAULT_MODELS[suggestedProvider] ?? modelFlag;
+        const validation = await resolveModel(suggestedProvider, suggestedModel, modelApiUrl);
+        // Estimate cost using cost-tracker
+        const ownedLines = (stream.owns ?? []).length * 100; // rough estimate
+        const costEstimate = estimatePlannedCost(ownedLines * 4, validation.resolvedModel); // ~4 tokens/line
+        modelDecisions[id] = {
+            provider: suggestedProvider,
+            model: validation.resolvedModel,
+            estimatedCostUsd: costEstimate.finalCost,
+            fallback: validation.wasFallback ? `${validation.originalModel} → ${validation.resolvedModel}` : undefined,
+        };
+    }
+    const preview = formatPlanPreview(initStreams, waveList, allWarnings, seqDiag, modelDecisions);
     // Show preview
     console.log(formatPlanPreviewText(preview));
     if (dryRun) {

package/dist/intelligence/index.d.ts

@@ -4,7 +4,7 @@ export { extractDeliverables, processDeliverables, formatDeliverablesReport } fr
 export type { Deliverable } from './deliverable-extractor.js';
 export { buildDependencyGraph, formatDependencyReport } from './dependency-inferrer.js';
 export { generateStreams, formatStreamsForReview, toInitFormat, extractPrerequisites } from './stream-generator.js';
-export { formatPlanPreview, formatPlanPreviewText } from './plan-preview.js';
+export { formatPlanPreview, formatPlanPreviewText, type ModelDecision } from './plan-preview.js';
 export { detectSequentialEdits, autoFixSequentialEdits } from './sequential-diagnostics.js';
 export type { SequentialEditDiagnostic } from './sequential-diagnostics.js';
 export { createDiagnostics, detectOwnershipConflicts } from './diagnostics.js';
@@ -30,14 +30,14 @@ export { applyPartialApproval, rollbackStream } from './interactive-approval.js'
 export { generateFixStream } from './self-healer.js';
 export { cleanupOrphanFixStreams, getFixChainInfo, isFixStream, getOriginalStreamId, onStreamComplete } from './fix-stream-manager.js';
 export { routeStream, loadRoutingRules } from './smart-router.js';
-export type { RouteDecision } from './smart-router.js';
+export type { RouteDecision, RouteOptions } from './smart-router.js';
 export { detectPatterns, formatPatterns, patternsToPromptHints, savePatterns, loadPatterns, adaptReport } from './pattern-detector.js';
 export type { DetectorInput } from './pattern-detector.js';
 export { pruneUnusedFiles, estimateTokens, generateCachingHints } from './context-optimizer.js';
 export { checkBudget, createBudgetConfig, ContextBudgetExceededError } from './budget-enforcer.js';
 export type { BudgetCheckResult } from './budget-enforcer.js';
 export { extractRelevantChunks } from './file-chunker.js';
-export { getModelCosts } from './cost-tracker.js';
+export { getModelCosts, estimatePlannedCost } from './cost-tracker.js';
 export { extractImports, resolveImportPath } from './heuristics.js';
 export { findMatchingTemplate, applyTemplate, suggestSplit } from './slicing-templates.js';
 export { optimizeTopology } from './topology-optimizer.js';