Blueprint Run

A blueprint_run records a single execution of the Blueprint Analyzer — Well’s multi-phase web-navigation agent (Claude Opus) that autonomously browses a target URL to accomplish a natural-language goal. Each run belongs to exactly one workspace and progresses through a lifecycle (queued → running → persisting → completed | failed | abandoned). The run is the audit spine for the analysis session: it captures the high-level intent (goal, target_url), the agent’s final synthesized output (result_blueprint), and — on failure — structured error diagnostics (error_type, error_message, failed_at_step, error_context). Individual navigation actions are stored as child BlueprintStep records linked by a one-to-many relationship.

Naming	Value
Object	Blueprint Run
Resource type (JSON:API `type`)	`blueprint_run`
Collection / records root	`blueprint_runs`
REST base	`/v1/blueprint-runs`
Entity class	`BlueprintRun`

API operations

Operation	Method & path	Status
List	`GET /v1/blueprint-runs`	✅ Implemented
Retrieve	`GET /v1/blueprint-runs/{id}`	✅ Implemented
Create	`POST /v1/blueprint-runs`	🟡 Planned
Update	`PATCH /v1/blueprint-runs/{id}`	🟡 Planned
Delete	`DELETE /v1/blueprint-runs/{id}`	🟡 Planned

Data model

Attributes

Field	Type	Required	Constraints	Allowed values	Description
blueprint_run_id	string, UUID, 🔒 system	✅ Yes	unique; generated via gen_random_uuid() on insert	—	Public immutable identifier for the run. Used in all API surfaces; the internal `pk` is never exposed.
goal	string	✅ Yes	text; no max length	—	Natural-language objective passed to the Blueprint Analyzer agent at run creation. Describes what the agent should accomplish on the target website.
target_url	string	✅ Yes	text; no max length; must be a valid URL reachable by the agent	—	The URL the agent navigates as the starting point of the run.
firebase_uid	string	⚪ No	text; nullable	—	Firebase UID of the authenticated user who initiated the run. Written at run creation from the Firebase auth context. Used to associate the run back to a specific user session within the workspace.
status	enum (BlueprintRunStatusEnum)	✅ Yes	default ‘running’; not null; native Postgres enum `blueprint_run_status_enum`	queued, running, persisting, completed, failed, abandoned	Lifecycle state of the run. Transitions: queued (created but not yet started) → running (agent executing steps) → persisting (BlueprintStorageService writing results) → completed (success) \| failed (agent error) \| abandoned (timed out or manually cancelled). The `queued` and `persisting` states were added in Migration20260429110000.
step_count	integer	✅ Yes	default 0; not null; incremented atomically by BlueprintStorageService.persistStep	—	Running count of steps persisted for this run. Incremented atomically (not derived from the steps collection count) so it reflects the true number of completed steps even if child rows are soft-deleted.
result_blueprint	jsonb	⚪ No	nullable; free-form JSON	—	The structured output produced by the agent upon successful completion. Written by BlueprintStorageService from the AI’s final `aiResponse.blueprint` payload. Shape is agent-defined and varies by goal type (e.g. pricing tables, API schemas, form fields). Null until the run reaches `completed` status.
error_type	string	⚪ No	text; nullable	—	Machine-readable error category written when the run transitions to `failed`. Examples: navigation_error, timeout, parse_error. Used to group failures in monitoring dashboards.
error_message	string	⚪ No	text; nullable	—	Human-readable error description accompanying `error_type` on failure. Written by BlueprintStorageService when the agent reports a terminal error.
failed_at_step	integer	⚪ No	nullable; references the 1-based step_number of the failing BlueprintStep	—	The step_number at which the run failed. Written alongside `error_type` and `error_message` by BlueprintStorageService. Null when the run completed successfully or was abandoned before any step executed.
error_context	jsonb	⚪ No	nullable; free-form JSON	—	Full diagnostic payload for the failure event. Written by BlueprintStorageService as `{ aiResponse: <raw agent response> }` at the point of failure. Provides the raw AI output that caused or accompanied the error, for debugging and replay.
created_at	timestamptz, 🔒 system	✅ Yes	set by MikroORM onCreate lifecycle hook; not null	—	Timestamp when the run record was created. Indexed DESC for recency queries (partial index on deleted_at IS NULL).
updated_at	timestamptz, 🔒 system	⚪ No	set by MikroORM onCreate + onUpdate lifecycle hooks; nullable	—	Timestamp of the most recent update to any field on the run. Refreshed automatically on every flush that modifies the record.
completed_at	timestamptz	⚪ No	nullable; written by BlueprintStorageService when status transitions to completed or failed	—	Timestamp when the run reached its terminal state (completed or failed). Combined with `created_at`, this gives the wall-clock duration of the agent execution. Null while the run is in-flight (queued/running/persisting) or abandoned.
deleted_at	timestamptz	⚪ No	nullable; soft-delete sentinel	—	Soft-delete timestamp. When non-null the run is considered deleted and excluded from all standard queries and both partial indexes (which filter WHERE deleted_at IS NULL). Cascade delete from the workspace also hard-deletes via the FK constraint.

Relationships

Name	Type	Required	Description
workspace	to-one (workspace)	✅ Yes	The workspace that owns this blueprint run. FK: `workspace_pk → workspaces.pk`. Cascade: ON DELETE CASCADE — deleting a workspace hard-deletes all its runs. Used to scope the run to a tenant and enforce multi-tenant isolation in Hasura RLS.
steps	to-many (blueprint_step)	—	Ordered collection of `BlueprintStep` records representing the individual navigation actions the agent took during this run. Related via `BlueprintStep.blueprint_run_pk → BlueprintRun.pk`. Cascade: ON DELETE CASCADE — deleting a run hard-deletes all its steps. The composite `composite_steps_list` in composites.yml surfaces step_number, action_type, and action_success in the records table view.

System-computed

blueprint_run_id is generated by Postgres gen_random_uuid() on insert; unique constraint enforced at the DB level.
created_at is set by the MikroORM onCreate lifecycle hook on first flush; never updated.
updated_at is set by both onCreate and onUpdate lifecycle hooks; refreshed on every write to the record.
status defaults to running at the DB level; the controller may set it to queued when enqueuing before agent dispatch. BlueprintStorageService drives transitions: running → persisting at start of persist, then persisting → completed on success or persisting → failed on error.
step_count is incremented atomically by BlueprintStorageService.persistStep (not derived from the child collection) so it is consistent even when child rows are soft-deleted.
completed_at is written by BlueprintStorageService when the run reaches completed or failed status. It is never set for abandoned runs.
result_blueprint is written from aiResponse.blueprint by BlueprintStorageService only on the final successful step (terminal completed transition).
failed_at_step and error_context are written together by BlueprintStorageService when the agent reports a terminal error; error_context captures the raw aiResponse payload for replay.
Both partial indexes (idx_blueprint_runs_workspace_status on (workspace_pk, status) and idx_blueprint_runs_created_at_desc on (created_at DESC)) filter WHERE deleted_at IS NULL, so they apply only to active rows.
Cascade deletes flow from workspace → blueprint_runs → blueprint_steps via FK ON DELETE CASCADE chains.

Example

{
  "data": {
    "type": "blueprint_run",
    "id": "c3f7a2b1-8e4d-4f91-b035-2a6c1d9e0f84",
    "attributes": {
      "goal": "Extract the current pricing plans and feature comparison table from the Stripe pricing page",
      "target_url": "https://stripe.com/pricing",
      "firebase_uid": "XkJ7Qm2pL9nRvBwdTzAs8oY3eH5fCi",
      "status": "completed",
      "step_count": 7,
      "result_blueprint": {
        "pricing_plans": [
          { "name": "Starter", "price_usd": 0, "features": ["Payments", "Dashboard"] },
          { "name": "Growth", "price_usd": 49, "features": ["Payments", "Dashboard", "Radar", "Billing"] }
        ],
        "summary": "Three tiers identified: Starter (free), Growth ($49/mo), Enterprise (custom)."
      },
      "error_type": null,
      "error_message": null,
      "failed_at_step": null,
      "error_context": null,
      "created_at": "2026-05-14T09:12:03.000Z",
      "updated_at": "2026-05-14T09:14:47.000Z",
      "completed_at": "2026-05-14T09:14:47.000Z",
      "deleted_at": null
    },
    "relationships": {
      "workspace": {
        "data": { "type": "workspace", "id": "b91e4d02-7c3a-41f6-9d28-5f0a3bc87e14" }
      },
      "steps": {
        "data": [
          { "type": "blueprint_step", "id": "a1b2c3d4-0001-4e5f-8abc-def012345678" },
          { "type": "blueprint_step", "id": "a1b2c3d4-0002-4e5f-8abc-def012345679" }
        ]
      }
    }
  }
}

_{Source: apps/api/src/database/entities/BlueprintRun.ts · domain: automation · tier: Activity}

Main objects

Supporting objects

Platform objects

Activity objects

Infrastructure objects

API operations

Data model

Attributes

Relationships

System-computed

Example

​API operations

​Data model

​Attributes

​Relationships

​System-computed

​Example

API operations

Data model

Attributes

Relationships

System-computed

Example