PUBLIC ALPHABiztraak is now in public alpha. Share feedback
Biztraak
Agents & Automation
Overview Workflow Examples Extending Agents

Agent Workflow Examples

Biztraak Agents aren’t just chatbots, they’re workflow engines.
They map natural language into structured Biztraak actions while attaching to the object context you’re working on (project, component, deployment, or template).
Workflows record state and tool executions. Human checkpoints and linked deployment results are recorded only when the workflow definition includes those steps.

Available workflow

Deployment Agent (Rules-Based Engine)

  • Purpose: Convert natural language into deterministic infrastructure and code deployments, or scaffold entire projects from scratch.
  • Context: Operates on the deployment engine, validating requirements against Biztraak’s project rules and enabled workflow capabilities.
  • How It Works:
    1. Parse intent into a structured project plan.
    2. Validate the project, component, service-provider, and deployment requirements.
    3. Always pause for human approval after planning, before scaffolding or deployment.
    4. If ProjectId already exists, skip project creation but still require component specifications. Otherwise, create the project from the plan.
    5. Reuse an existing non-build-review service provider when one matches, or create one otherwise.
    6. Wait for missing components or service providers to become available before deployment.
    7. Start Terraform and complete the deployment through the linked AgentWorkflowId callback. The workflow records the linked deployment result when that callback returns one.
  • Extended capabilities:
    • Conditional scaffolding: the agent creates only the project, components, and service-provider configuration that are missing from the generated plan.
    • Template recommendations: not part of the current agent workflow. Use the Marketplace directly.

Inputs →

  • Team, project, and optional project component
  • Optional conversation, requested model, and notes
  • Natural-language project requirements used to generate the plan

Outputs →

  • Project plan and component specification
  • Project, component, and service-provider records
  • Terraform deployment request and linked completion result
  • Workflow, tool-execution, and human-checkpoint events

Examples

  • “Deploy the staging API to AWS us-east-1 with enhanced logging and approval context.”
  • “Scaffold a new project for an e-commerce platform with frontend, API, and database.”
  • “Find templates for a healthcare dashboard with stronger access and audit visibility.” → use the Marketplace directly.
  • “Scale the production DB cluster to 3 nodes.”
  • “Roll back the last infra change in staging.”

Workflow-dependent and forthcoming examples

The following examples describe possible extensions to the platform workflow model. They are not shipped as universal workflows. use the relevant builder or Cloud Dev page to confirm whether the corresponding feature is enabled for your team.

UI Builder assistance

  • Purpose: Automate frontend scaffolding in the UI Builder.
  • Context: Works on the component schema being edited in the builder.
  • How It Works:
    1. Parse input into UI schema objects (rows, inputs, containers).
    2. Update component JSON schema.
    3. Generate Angular/React code and inject into the repository.
  • Inputs: UI specifications described in natural language.
  • Outputs: updated schema + generated frontend code.

Examples

  • “Add a login form with email + password fields.” → generates schema and form code.
  • “Insert a two-column layout with a sidebar and content area.” → updates layout schema.
  • “Connect this form to the Auth federated app.” → wires in dependencies automatically.

API Builder assistance

  • Purpose: Automate backend creation in the API Builder.
  • Context: Attached to your API schema and connected components.
  • How It Works:
    1. Parse intent into entities, fields, and relationships.
    2. Generate models, resolvers, services, and GraphQL mutations.
    3. Insert code into the repository with CI/CD pipelines wired in.
  • Inputs: Entities, fields, relationships, endpoint type.
  • Outputs: GraphQL schema + service code + API endpoints.

Examples

  • “Create a Todo API with fields: title, description, completed.” → scaffolds full CRUD API.
  • “Generate CRUD for a Customer object with RBAC enforcement.” → adds authorization rules.
  • “Add a generateReport mutation that queries Postgres and returns JSON.” → builds custom resolver + service.

Cloud Dev assistance

  • Purpose: Bring agents directly into Biztraak Cloud VS Code environments.
  • Context: Knows the repo, open file, and active component.
  • How It Works:
    1. Reads current file and project metadata.
    2. Parses the request into code or infrastructure actions.
    3. Writes changes directly into the workspace.
    4. Optionally triggers test or deployment pipelines.
  • Inputs: developer commands inside VS Code.
  • Outputs: code snippets, infra modules, test scaffolds, or pipeline configs.

Examples

  • “Generate a Terraform module for an S3 bucket with logging.” → agent writes aws-storage.tf.
  • “Add unit tests for the UserService class.” → scaffolds test file in repo.
  • “Wire this API method to the frontend form.” → updates code across API + UI layers.

Why these workflows matter

  • Context-Aware: actions scoped to the right project, component, or file.
  • Developer-First: agents complement, not replace, native infra, code, and pipelines.
  • Governance-aware: actions can stay logged, reviewable, and easier to trace.
  • Unified: the workflow model can connect UI, API, and Cloud Dev actions where those integrations are enabled.
  • Extensible: future or team-specific extensions can build on the platform workflow model.

Next steps

The available deployment workflow plans project resources, pauses for human review, and can start Terraform deployments with recorded workflow state. Other agent-assisted experiences depend on their respective product rollout and configuration.