Core Products·Kaviya Sree P (checked by Ancy)·4 min read

TeksBuild: supported stacks, sandbox preview, code review, Vercel deploys

Comprehensive guide to TeksBuild: supported tech stacks, isolated sandbox previews, line-by-line human-in-the-loop code review, Vercel deployment, and triage runbooks for build failures.

TeksBuild turns an approved Plan Bundle into working, production-grade software. It operates through an autonomous execution loop: decomposing tasks, generating code, running unit and contract tests inside an isolated sandbox container, and self-healing when tests break.

Once all tests pass, TeksBuild packages a live Sandbox Preview and requests your line-by-line Code Review before anything is deployed.


1. Supported technology stacks

TeksBuild supports modern, type-safe frameworks and battle-tested databases. Every stack template comes pre-configured with linting, testing, and CI/CD out of the box:

Stack / Layer Primary Technologies Capabilities
Web & Frontend Astro 5, React 19, Tailwind CSS v4, TypeScript Server-islands, static prerendering, responsive UI, zero-runtime CSS.
Backend & APIs Node.js (TypeScript) / Python 3.12 (FastAPI) OpenAPI 3.1 typed schemas, REST endpoints, streaming responses.
Data & ORM PostgreSQL, Drizzle ORM, SQLite, pgvector Type-safe migrations, connection pooling, typed schema validation.
Data Platform Snowflake, Databricks, dbt Core, Apache Airflow Warehouses, data modeling, automated ETL/ELT test pipelines.
Deployment Vercel Edge/Serverless, Docker Runbox, AWS ECS Automatic previews, branch staging, custom domain routing.

[!TIP] If you select “Let Factory decide” during project setup, TeksBuild chooses the optimal stack based on your problem requirements (defaulting to Astro + React + TypeScript + PostgreSQL for high performance and low operational complexity).


2. The Cloud Sandbox Preview

Before any code lands on your Git main branch or production environment, TeksBuild executes the application inside an isolated Runbox Sandbox Container:

Key Preview Features

  • Ephemeral Sandbox URL: A dedicated, secure preview URL is generated for every build run (e.g. https://preview-run-8492.tekspaz.dev).
  • Interactive Testing: Click buttons, fill out forms, and trigger API requests in real time to verify that the app behaves as requested.
  • Console & Network Inspector: View browser console logs, HTTP request/response payloads, and database queries executed during your preview session.
  • Strict Network Isolation: Sandboxes cannot access your host machine, internal AWS metadata services (169.254.169.254), or unapproved external URLs.

3. Human-in-the-Loop Code Review

TeksBuild never commits code silently. Once the build loop completes, your project advances to the Review Stage (/projects/[id]/review):

flowchart TD
    A["Approved Plan"] --> B["Task Decomposition"]
    B --> C["Code Generation"]
    C --> D["Automated Test Execution"]
    D -- "Tests Fail" --> E["Diagnostic Agent Self-Healing"]
    E --> C
    D -- "All Tests Pass" --> F["Live Sandbox Preview"]
    F --> G{"Human Code Review"}
    G -- "Approve" --> H["Deploy to Vercel / Target"]
    G -- "Request Fix" --> E

Reviewing the changes

  1. Diff Inspector: Inspect side-by-side git diffs of all added, modified, or deleted files.
  2. Automated Check Badges: Verify the green badges for:
    • TypeScript compilation (astro check / tsc --noEmit)
    • Unit tests (vitest run)
    • Contract tests (vitest run --mode contract)
    • Security scan (secret leakage checks & dependency audits)
  3. Commit Summary: Read the AI-generated commit message explaining architectural decisions and implementation rationale.
  4. Click Approve Code: Once satisfied, click Approve Code to unlock deployment.

4. Deploying to Vercel

Tekspaz features deep first-class integration with Vercel for zero-config global deployment:

Deployment Steps

  1. Ensure your Vercel account or team is connected in Connect GitHub & Vercel.
  2. From the review screen, click Deploy to Vercel.
  3. TeksBuild packages the production build:
    • Generates production serverless functions and static assets via @astrojs/vercel.
    • Propagates environment variables securely.
    • Pushes deployment triggers via Vercel’s Deployment API.
  4. Monitor the deployment progress bar directly in your project’s Deployments tab.
  5. Once live, the production URL (e.g. https://my-app.tekspaz.ai) and deployment health checks are displayed.

5. What to do when a build fails

If an unexpected build error occurs or tests cannot be automatically resolved, TeksBuild stops and marks the run as build_failed. Follow this triage runbook:

1. Read the Failure Diagnostic

Click on the failed build card to inspect the diagnostic report:

  • Phase: Indicates whether failure occurred during dependency installation (pnpm install), compilation (tsc), or test suite execution (vitest).
  • Error Stack: Pinpoints the exact file and line number that failed.
  • Diagnostic Agent Explanation: Summarizes root cause (e.g. “Missing environment variable DATABASE_URL in preview container” or “Type mismatch in API response schema”).

2. Common failure scenarios and fixes

Issue: TypeScript Type Mismatch

  • Symptoms: error TS2322: Type 'string' is not assignable to type 'number'.
  • Resolution: Click Retry with Prompt, and tell the builder:
    Fix type error in src/lib/api.ts: ensure the id field is parsed with Number() before passing to db.

Issue: Database Migration Timeout

  • Symptoms: Connection to PostgreSQL timed out after 10000ms.
  • Resolution: Verify your database connection string in Data Sources (/data-sources). Ensure your database security groups allow inbound connections from Tekspaz IPs.

Issue: Missing Environment Variable

  • Symptoms: ReferenceError: process.env.STRIPE_SECRET_KEY is not defined.
  • Resolution: Add the missing secret in Project Settings › Environment Variables, then click Re-run Build.

3. Triggering a fresh build

If the failure was transient (e.g. temporary network blip or npm registry timeout), click Re-run Build from the top right action menu. TeksBuild will re-provision an isolated container and rerun from the last good checkpoint.


What to explore next

Ask the DocsAI Assistant

Instant answers grounded in our guides with direct links to sources.

Suggestions:

Have questions about this guide?

Our founding engineering team is ready to help you get started.

Contact Support