Grilling Chipmunks：统一技能与智能体的 Salesforce 领域解决方案

Grilling Chipmunks is an open-source project addressing enterprise AI application development challenges by unifying skills, agents, and project code under a monorepo architecture. It focuses on Salesforce domain solutions, leveraging the gstack workflow to provide a complete intelligent automation system. Key components include a skills system, autonomous agents, and integrated project management, aiming to balance flexibility, scalability, and enterprise-grade stability.

The core challenge in enterprise AI app development is managing skills, agents, and project code within a unified architecture. Grilling Chipmunks aims to solve this by adopting a monorepo approach to centralize component management, with dedicated support for the Salesforce ecosystem. It introduces the gstack workflow to enable complex enterprise automation scenarios, serving as a bridge between traditional business systems and AI capabilities.

The monorepo design offers three key benefits:

1. Code Sharing & Reuse: Simplifies internal dependencies (no intermediate package releases), reduces refactoring costs (visibility of all affected areas), and ensures version consistency.
2. Atomic Changes: Allows cross-module changes in one commit (atomic success/rollback), complete code review context, and coordinated deployment to avoid version mismatches.
3. Unified Toolchain: Ensures consistent development experience, centralized config (ESLint, Prettier, TypeScript), and a single CI/CD workflow for the entire project.

gstack is the project's core workflow engine, following the 'workflow as code'理念:

• Version Control: Workflow definitions are Git-managed for traceability.
• Testability: Logic can be unit-tested like regular code.
• Balance of Declarative & Command式: Declarative parts define structure/dependencies/triggers; command式 parts handle business logic (data processing, API calls).
• State Management: Supports persistence (database storage for recovery), distributed execution, and monitoring (metrics, logs) for enterprise-grade reliability.

Grilling Chipmunks provides deep Salesforce integration (not just API封装):

• Domain Modeling: Object model mapping (type-safe access), business logic orchestration (Apex triggers, approval flows), and security model adaptation.
• Typical Scenarios:
  • Data Sync: Listen to Salesforce changes, clean/transform data, push to other systems (ERP, data warehouses) with error recovery.
  • Process Automation: Centralize scattered logic, orchestrate cross-system workflows, and enable visual monitoring.
  • Intelligent Decision Support: Sales opportunity scoring, case classification/routing, and predictive suggestions based on historical data.

Skills System:

• Definition: Each skill includes metadata (name, version), input/output contracts, implementation logic, and test cases.
• Discovery: A registry allows browsing, documentation access, and dependency understanding.
• Combination: Supports sequence, parallel execution, conditional branches, and loops.
• Versioning: SemVer compliance, backward compatibility, and multi-version isolation.

Agents Architecture:

• Autonomous Decision: Goal-oriented, task planning, tool (skill) usage, and feedback-based adjustment.
• Multi-agent Collaboration: Role division, message communication, task delegation, and conflict resolution.
• Human-Machine Collaboration: Human-in-loop (key decisions), transparent decision processes, and intervention capabilities.

Key technical features:

• Type Safety: Extensive TypeScript use for compile-time error checks, IDE support (autocomplete, refactoring), and self-documenting interfaces.
• Testing Strategy: Multi-layered tests (unit, integration, E2E, contract) to ensure reliability.
• Observability: Structured logs, metrics collection, distributed tracing, and health checks for operational visibility.

Grilling Chipmunks represents a new paradigm for enterprise AI apps—unifying skills, agents, and workflows in a monorepo with deep domain integration (Salesforce). It balances flexibility/scalability with enterprise stability. For AI automation-focused enterprises, it provides a reference architecture. Future trends: More such unified platforms will emerge as bridges between traditional systems and AI, driving wider adoption of intelligent automation.