[FEATURE] Epic: Enable support for Tencent Cloud Serverless

[Blankll]

Enable Tencent Cloud Serverless Support via a Terraform-like Engine in ServerlessInsight

Summary

This Epic proposes extending ServerlessInsight with a Terraform-like infrastructure engine to enable first-class support for Tencent Cloud serverless resources, while laying a solid foundation for multi-cloud, stateful, and repeatable Infrastructure-as-Code (IaC) workflows.

The goal is not only to add a Tencent Cloud provider, but to introduce core IaC capabilities—state, plan, and deterministic apply—so ServerlessInsight can safely manage serverless infrastructure across providers (Aliyun today, Tencent Cloud next, others later).

Architecture

---

Problem Statement

Currently, ServerlessInsight:

• Supports declarative YAML-based infrastructure definitions
• Can deploy serverless resources via si deploy
• Has an initial provider implementation for Aliyun (based on ROS)

However, it does not:

• Track deployed infrastructure state
• Detect drift between desired and actual resources
• Provide a preview (plan) before applying changes
• Support Tencent Cloud (SCF, API Gateway, COS) as a backend provider

As a result:

• Users cannot safely manage the full lifecycle of serverless infrastructure on Tencent Cloud
• Multi-cloud workflows lack consistency and predictability
• Re-running deployments risks unintended changes

---

Proposed Solution

Introduce a Terraform-like engine inside ServerlessInsight and use it to implement Tencent Cloud serverless support.

This engine will:

• Parse declarative YAML (existing capability - refer parser)
• Track deployed infrastructure using a state store
• Compute diffs between desired and actual infrastructure (plan)
• Apply changes in correct dependency order (reuse si deploy)
• Enable idempotent create / update / delete workflows
• Remain backward-compatible with existing providers (e.g., Aliyun ROS)

---

Core Design Principles

• Declarative-first: YAML remains the source of truth
• Provider-agnostic: No provider-specific syntax leaks into YAML
• Stateful & safe: Always know what exists before changing it
• Incremental adoption: Reuse existing parser and deploy logic
• Extensible: Future Tencent Cloud resources (VPC, DB, MQ) fit naturally

---

Core Components

1. Configuration Parsing (Existing)

• Parse serverlessinsight.yml
• Convert YAML into an internal resource model (functions, buckets, APIs, databases, etc.)
• Already implemented and reused as-is

---

2. Resource Representation (scfStack)

Define TypeScript abstractions for each resource type, such as:

• functions
• events(apigateway should under)
• bucket

Each resource:

• Encapsulates its properties (runtime, memory, handler, bucket name, etc.)
• Implements lifecycle operations:
  • create()
  • read()
  • update()
  • delete()
  • import() (future)
  • refresh()

Tencent Cloud resources will be implemented using:

• [tencentcloud-sdk-nodejs](https://github.com/TencentCloud/tencentcloud-sdk-nodejs)

---

3. Dependency Graph

Build a dependency graph to ensure correct execution order.

Dependencies can be:

• Inferred from YAML references
  • Example: API Gateway event referencing a function
• Explicit, via dependsOn

This ensures:

• Functions exist before API Gateways
• Buckets exist before website configs
• Safe parallelism where possible

---

4. State Management (New)

Introduce a Terraform-like state store:

• Initial implementation: local JSON file
• Designed to support remote backends in the future
• Maps logical resources → physical cloud resource IDs

State responsibilities:

• Track created resources
• Enable drift detection
• Ensure idempotent operations

Before planning or applying, the engine should:

• Refresh state by querying cloud APIs
• Reconcile state with real-world infrastructure

---

5. Planning Phase (si diff) — New Command

Introduce a new command:

si plan

Responsibilities:

• Compare desired state (YAML) vs current state
• Produce a plan without making changes

The plan identifies:

• Resources to create
• Resources to update
• Resources to delete

Planning flow:

1. Load existing state
2. Refresh state from cloud
3. Diff against YAML
4. Produce an ordered execution plan

---

6. Apply / Execution (si deploy) — Reuse Existing Command

Reuse and enhance the existing si deploy command to:

• Consume the generated plan
• Apply actions in dependency order
• Call provider SDKs (Tencent Cloud, Aliyun, etc.)
• Persist state after each successful operation

Execution behavior:

• Create / update resources via provider SDK
• Delete obsolete resources
• Stop dependent operations on failure
• Leave partial state intact for recovery

---

7. State Updates

After apply:

• Persist updated state
• Ensure next si diff starts from a correct baseline
• Enable drift detection on future runs

---

Tencent Cloud Support Scope

Initial Tencent Cloud resources:

• SCF (Serverless Cloud Functions)
• API Gateway
• COS (Cloud Object Storage)

YAML syntax:

• Follows ServerlessInsight’s existing provider-agnostic conventions
• Does not copy Terraform or serverless-framework syntax
• Supports variables, stages, and references

Example YAML (excerpt):

version: 0.0.1
provider:
  name: tencent
  region: cn-hangzhou

vars:
  testv: testVarValue
  handler: index.handler

stages:
  default:
    region: ${vars.region}
    node_env: default
  dev:
    region: ${vars.region}
    node_env: development
  prod:
    region: cn-shanghai

service: insight-poc-api

tags:
  owner: geek-fun

functions:
  insight_poc_fn:
    name: insight-poc-api-fn
    code:
      runtime: nodejs18
      handler: ${vars.handler}
      path: tests/fixtures/artifacts/artifact.zip
    memory: 512
    timeout: 10
    environment:
      NODE_ENV: ${stages.node_env}
      TEST_VAR: ${vars.testv}
      TEST_VAR_EXTRA: abcds-${vars.testv}-andyou

events:
  gateway_event:
    type: API_GATEWAY
    name: insight-poc-gateway
    triggers:
      - method: GET
        path: /api/hello
        backend: ${functions.insight_poc_fn}
    domain:
      domain_name: insight-api.serverlessinsight.com

buckets:
  insight_poc_bucket:
    name: insight-poc-bucket
    website:
      code: dist
      index: index.html
      error_page: 404.html
      error_code: 404

---

Cross-Provider Compatibility

• Aliyun (ROS-based) remains unchanged
• Tencent Cloud uses the new stateful engine
• Engine abstracts differences between providers
• No breaking changes to existing users

---

Implementation Outline (TypeScript)

some coding style guidance:

• Tries to use functional programing style for TypeScript implementation,
• use less keywords function, dfine function as const xxx = xxx or export const xxx = xxx
• use type, enum rather than interface as possible
• create index.ts for a module, manage the exports in index.ts so outeside module can import from index.ts rather than expose the file name inside module

Key modules:

• parser — YAML parsing & validation
• scfStack — CRUD lifecycle
• dependencies — DAG construction
• state — local JSON state (future backend support)
• planner — diff & execution plan
• executor — apply plan using si deploy
• CLI extensions:
  • si plan
  • si deploy

---

Expected Outcomes

By implementing this Epic, ServerlessInsight gains:

• Terraform-like plan/apply safety
• First-class Tencent Cloud serverless support
• Drift detection & idempotent deployments
• A scalable foundation for multi-cloud expansion
• Improved reliability for multi-tenant environments

---

References

• Terraform Dependency Graph
  https://developer.hashicorp.com/terraform/internals/graph
• Terraform State
  https://developer.hashicorp.com/terraform/language/state
• Terraform Plan
  https://developer.hashicorp.com/terraform/cli/commands/plan
• Serverless Cloud Framework (Tencent)
  https://github.com/Serverlesstencent/serverless-cloud-framework