agentic-api 2.0.641 → 2.0.642

package/README.md

@@ -324,30 +324,66 @@ console.log('Messages:', result.messages.length);
 
 ## MapLLM Document Processing
 
-Map-reduce pattern for large documents:
+True map-reduce pattern for large documents with structured outputs and callback-driven accumulation:
 
 ```typescript
-import { MapLLM, FileNativeLoader } from '@agentic-api';
+import { MapLLM, FileNativeLoader, StringNativeLoader } from '@agentic-api';
 
+// Chunking strategies: lines, pages, paragraphs, overlap
 const loader = new FileNativeLoader('document.pdf', { type: 'pages', size: 1 });
 const mapper = new MapLLM(loader);
 
 const config = {
-  digestPrompt: "Analyze this chunk for key information.",
-  reducePrompt: "Merge analysis with previous results."
+  digestPrompt: "Analyze this chunk and extract key points.",
+  reducePrompt: "Merge this analysis with previous results.",
+  reduceModulo: 5  // Optional: reduce every N chunks
 };
 
+// Callback controls accumulation and termination
 const result = await mapper.reduce(config, (result, currentValue) => {
-  result.acc = result.acc + `\n${currentValue}\n`;
+  // Structured output support
+  result.format = { name: "analysis", schema: myJsonSchema, strict: true };
+  result.acc = { ...result.acc, ...currentValue };
+  
+  // Stop conditions
   if (result.metadata.iterations > 20) result.maxIterations = true;
   return result;
-}, { acc: "", model: 'LOW-fast', verbose: true });
+}, { acc: {}, model: 'LOW', verbose: true });
 ```
 
 **Documentation:** [MapLLM](./docs/10.AGENTS.MAPLLM.md)
 
 ---
 
+## JobRunner (Plan → Execute → Reduce)
+
+Sequential task execution engine for complex multi-objective requests:
+
+```typescript
+import { JobRunner, jobPlannerPrompt } from '@agentic-api';
+
+// JobRunner workflow:
+// 1. Planner: userRequest → JobPlan (list of TaskSpec)
+// 2. Executor: TaskSpec + memory → TaskResult
+// 3. Reducer: (prevMemory, task, result) → ReducedJobMemory (via MapLLM)
+
+// Features:
+// - Strict contracts (structured outputs / schema validation)
+// - Context reduction after each task via MapLLM reducer
+// - Error policy: 2 attempts max per task, then thrown → synthesis
+// - Sequential deterministic execution (V1)
+```
+
+**Key Contracts:**
+- `JobPlan`: jobId, goal, tasks[]
+- `TaskSpec`: id, title, type, dependsOn[], acceptance[]
+- `TaskResult`: taskId, ok, summary, data, artifacts[], error
+- `ReducedJobMemory`: memory (short canonical), index (stable refs)
+
+**Documentation:** [JobRunner](./docs/10.AGENTS.JOB.md)
+
+---
+
 ## Rules Management System
 
 Git-based workflow for business documents:
@@ -409,7 +445,7 @@ npm test
 
 ## License
 
-MIT License - Copyright (c) 2024-2026 Pilet & Renaud SA
+MIT License - Copyright (c) 2024-2026 olivier@evaletolab.ch
 
 Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agentic-api",
-  "version": "2.0.641",
+  "version": "2.0.642",
   "description": "API pour l'orchestration d'agents intelligents avec séquences et escalades automatiques",
   "main": "dist/src/index.js",
   "types": "dist/src/index.d.ts",