[RFC] The AutoMaintainer Enterprise SDK

[purvanshjoshi]

[RFC] The AutoMaintainer Enterprise SDK

The Vision

While the AutoMaintainer UI and CLI are designed for active development, true enterprise automation requires programmatic access. We propose abstracting the LangGraph 5-agent pipeline into an embeddable Software Development Kit (SDK).

By publishing an SDK (@pxa-labs/automaintainer-sdk for Node.js, and automaintainer-sdk for Python), enterprise teams can integrate autonomous software engineering natively into their Continuous Integration and Continuous Deployment (CI/CD) pipelines.

Proposed Architecture

The SDK will serve as a programmatic wrapper around the core FastAPI backend and LangGraph execution engine. It will provide async/await bindings for all agent roles, allowing scripts to dynamically trigger specific agent behaviors.

flowchart TD
    classDef ci fill:#2d3748,stroke:#4a5568,stroke-width:2px,color:#fff,rx:8px
    classDef sdk fill:#2b6cb0,stroke:#2c5282,stroke-width:2px,color:#fff,rx:8px
    classDef backend fill:#276749,stroke:#22543d,stroke-width:2px,color:#fff,rx:8px

    subgraph CI_Pipeline [Enterprise CI/CD Environment]
        Trigger["GitHub Action / Jenkins Pipeline\n(Triggers on PR Open)"]:::ci
        Integration["Custom Python/Node Script\n(Uses automaintainer-sdk)"]:::ci
    end
    
    subgraph Core [AutoMaintainer Engine]
        SDK_Binding["SDK API Client\n(RPC / REST)"]:::sdk
        LangGraph["LangGraph Supervisor\n(5-Agent Pipeline)"]:::backend
    end

    Trigger -->|Executes| Integration
    Integration -->|await client.agents.reviewer.analyze(pr_id)| SDK_Binding
    SDK_Binding <-->|Orchestrates| LangGraph
    LangGraph -->|Pushes Commits to PR| Trigger

Loading

Proposed API Endpoints

The SDK will expose the following programmatic capabilities:

from automaintainer import Client

client = Client(api_key="your_groq_key")

# Example 1: Automated PR Reviews in CI/CD
review_results = await client.agents.reviewer.analyze_pull_request(
    repo="PxA-Labs/my-repo",
    pr_number=42
)
if review_results.has_critical_bugs:
    # Trigger the implementer to fix it automatically!
    await client.agents.implementer.fix_pull_request(pr_number=42)

# Example 2: Weekly Architectural Audits
architecture_report = await client.agents.architect.scan_repository(
    repo="PxA-Labs/my-repo"
)
print(architecture_report.technical_debt_score)

Distribution Strategy

The SDK will be distributed via PyPI for Python environments and NPM for TypeScript/Node.js environments. The core logic will still execute the Python LangGraph engine via sub-processes or REST calls.

Open Questions

1. Should the SDK execute the LangGraph pipeline entirely on the local CI/CD runner machine, or should it send HTTP requests to a managed SaaS endpoint (e.g., our Supabase-backed infrastructure)?
2. How do we securely handle GitHub Tokens when the SDK operates autonomously within a GitHub Action runner?

Please leave your thoughts, feature requests, and architectural feedback below!