| #!/bin/bash |
| set -e |
|
|
| cd /home/user/webapp |
|
|
| git config user.email "pyaesonegtckglay-dotcom@users.noreply.github.com" |
| git config user.name "pyaesonegtckglay-dotcom" |
|
|
| echo "=== Git Status ===" |
| git status --short |
|
|
| echo "=== Staging all changes ===" |
| git add -A |
|
|
| echo "=== Committing ===" |
| git commit -m "feat: Workflow Factor OS v2.0 — Full Production Upgrade |
| |
| BREAKING CHANGES & MAJOR FEATURES: |
| |
| ## LLM Gateway (Provider-Agnostic) |
| - REMOVED: direct OpenAI SDK dependency from all agents |
| - ADDED: integrations/llm-providers/index.ts — pure fetch-based LLM gateway |
| - SUPPORTS: openai | anthropic | openai-compatible | custom endpoints |
| - FEATURES: retry with exponential backoff, fallback provider, timeout handling |
| - CONFIG: LLM_API_KEY, LLM_GATEWAY_URL, LLM_MODEL, LLM_PROVIDER env vars |
| - BACKWARD COMPAT: OPENAI_API_KEY still supported as legacy fallback |
| |
| ## Node Registry System (Strict — No Hallucinated Nodes) |
| - REBUILT: apps/simulator/src/knowledge/nodeRegistry.ts — 35+ real n8n nodes |
| - ADDED: NodeRegistryEntry interface with full schema per node type |
| - ADDED: isValidNodeType(), getNodeDef(), getTriggerNodeTypes() helpers |
| - ENFORCED: unknown node types REJECTED at graph engine + validator stages |
| - COVERS: Triggers, AI/LLM, Transform, HTTP, Messaging, Productivity, Database, CRM, Utility |
| |
| ## Swarm Multi-Agent Orchestration |
| - NEW: apps/simulator/src/services/swarmOrchestrator.ts |
| - RUNS: 1-5 parallel workflow design candidates simultaneously |
| - STRATEGIES: Reliability-First, Simplicity-First, Observability-First, AI-Optimized, Pipeline-Optimized |
| - SELECTS: best design via quality scoring (validation + registry + expression + dataflow) |
| - CONFIG: swarmSize option (default: 3) |
| |
| ## Self-Healing System |
| - NEW: apps/simulator/src/services/selfHealing.ts |
| - PHASE 1: Rule-based deterministic fixes (active:false, IDs, positions, retry, onError, unknown nodes) |
| - PHASE 2: AI-assisted healing for complex expression/logic issues |
| - MEMORY: records failure patterns, applies historical fixes |
| - TRIGGERS: automatically on validation failure before re-validation |
| |
| ## Memory / Learning System |
| - NEW: apps/simulator/src/services/memorySystem.ts |
| - STORES: successful patterns, failed patterns, fix strategies, architecture templates |
| - BUILT-IN: 4 high-performing templates (Telegram Bot, Webhook Pipeline, Data Pipeline, Email-CRM) |
| - INFLUENCES: planning (memory context injection), architecture selection |
| - STATS: getStats() for reporting |
| |
| ## Webhook Auto-Bind System |
| - NEW: apps/simulator/src/services/webhookAutoBind.ts |
| - AUTO-DETECTS: trigger type from user request + intent (Telegram, Slack, GitHub, Email, Schedule, Webhook, Manual) |
| - AUTO-GENERATES: webhook paths, trigger settings, credential stubs |
| - INJECTS: trigger node into graph automatically |
| - USER: never needs to manually define triggers |
| |
| ## Validation Engine (UPGRADED — 7 stages) |
| - ADDED: Stage F — NodeRegistry validation (rejects unknown/hallucinated node types) |
| - ADDED: Stage G — DataFlow validation (SET empty, CODE placeholder, IF no conditions) |
| - UPGRADED: Expression validation (static nodes, placeholder detection, unsafe access) |
| - SCORE: 7-dimension weighted scoring |
| |
| ## All Agents Upgraded (Provider-Agnostic) |
| - intentInterpreter.ts: uses LLMGateway, detects triggerType for auto-bind |
| - workflowPlanner.ts: uses LLMGateway |
| - graphEngine.ts: uses LLMGateway + validates all node types against registry |
| - compiler.ts: uses LLMGateway + strict expression rules + node schema injection |
| - validator.ts: uses LLMGateway + 7-stage validation |
| - simulator.ts: uses LLMGateway + chaos testing + deployment blockers |
| - credentialIntelligence.ts: uses registry for credential mapping |
| |
| ## All Prompts Upgraded |
| - intentInterpreter: triggerType detection added |
| - workflowPlanner: strict node type rules, expression examples, anti-placeholder rules |
| - graphEngine: registry enforcement, expression requirements, data contract rules |
| - compiler: full expression guide (correct/wrong examples), node-specific rules |
| - simulator: scoring formula, chaos scenarios, deployment blockers |
| |
| ## wrangler.toml (Added/Upgraded) |
| - LLM_PROVIDER, LLM_GATEWAY_URL, LLM_MODEL vars |
| - Development environment overrides |
| - Production environment section |
| - Secrets documentation updated |
| |
| ## Worker (Upgraded) |
| - types/env.ts: LLM_GATEWAY_URL, LLM_MODEL, LLM_PROVIDER, LLM_FALLBACK_* added |
| - types/workflow.ts: SwarmResultSummary, TriggerDetectionResult, HealingResultSummary types added |
| - routes/generate.ts: passes LLM config to simulator, stores swarm/healing/trigger results |
| - routes/health.ts: capabilities info, safety policy documentation |
| - routes/validate.ts, simulate.ts: llmApiKey support (legacy openaiApiKey still works) |
| |
| ## Safety Rules (Maintained) |
| - NEVER auto-activate workflows |
| - ALWAYS validate + simulate before deploy |
| - ALWAYS require human approval before activation |
| - NEVER deploy partially valid workflows |
| - NEVER allow unknown node types |
| - active: false enforced at compiler + self-healing stages" |
|
|
| echo "=== Pushing to main ===" |
| git push origin main |
|
|
| echo "=== Done! ===" |
|
|
