Skip to content

LLM-Driven Extraction of Unstructured Data — Built for API Deployments & ETL Pipeline Workflows

4.9 KB
Raw Blame
# Unstract critical paths registry.
#
# A "critical path" is an end-to-end user or system flow whose failure would
# constitute a production incident. The rig reports:
#   ✅ covered    — at least one group in `covered_by` ran green this build
#   ⚠️ gap        — no covering group ran green this build
#   ❌ regression — a path that was ✅ on the cached main baseline is now not ✅
#
# Only one kind of gap gates --fail-on-critical-gap:
#   • in-scope gap     — a covering group ran in this tier but not green; fails.
#   • out-of-scope gap — covered only by an unrun tier, or no group declared;
#                        warn-only (a tier can't fail for coverage it can't run).
#
# Only wire `covered_by` to a group that really exercises the path — a bogus
# mapping fails the build when that group breaks, for the wrong reason.
#
# `proof` (default: group):
#   • group  — covered when any covered_by group runs green. Coarse: survives
#              the covering test being deleted.
#   • marker — additionally requires ≥1 passing test carrying
#              @pytest.mark.critical_path("<id>") in this build. Flip each
#              path to marker once its tests are marked.
#
# We intentionally do NOT chase 100% coverage. Focus on filling these gaps first.
version: 1

paths:
  - id: auth-login
    description: "User can log in and obtain a session cookie."
    covered_by: [e2e-login]
    proof: marker

  - id: adapter-register-llm
    description: "Register and validate an LLM adapter."
    covered_by: [integration-backend]
    proof: marker

  - id: workflow-author
    description: "Create a workflow; its source+destination endpoints materialise and are configurable."
    covered_by: [integration-backend]
    proof: marker

  - id: co-owner-manage
    description: "Add/remove co-owners of a shared resource; enforce the last-owner guard."
    entry: "POST /api/v1/workflow/{id}/owners/"
    covered_by: [integration-backend, e2e-coowners]
    proof: marker

  - id: workflow-create-execute
    description: "Create a workflow, configure source+destination, execute, poll, fetch result."
    covered_by: [e2e-workflow]
    proof: marker

  - id: api-deployment-provision
    description: "Deploying a workflow as an API mints a usable key and a resolvable endpoint."
    covered_by: [integration-backend]
    proof: marker

  - id: api-deployment-auth
    description: "Unauthenticated or mis-scoped API-deployment calls are rejected before dispatch."
    covered_by: [integration-backend]
    proof: marker

  - id: api-deployment-run
    description: "Deploy a workflow as an API, POST a document, receive structured JSON."
    covered_by: [e2e-api-deployment]
    proof: marker

  - id: mcp-server-auth
    description: "Unauthenticated or mis-scoped hosted-MCP calls are rejected before any tool runs."
    covered_by: [integration-backend]
    proof: marker

  - id: mcp-platform-auth
    description: "The org-scoped MCP endpoint stays behind the platform-API-key middleware; unauthenticated or mis-scoped calls reach no tool."
    covered_by: [integration-backend]
    proof: marker

  - id: platform-key-whoami
    description: "A platform API key resolves its own organisation over the org-less whoami endpoint; the org comes from the key row, not the URL."
    covered_by: [integration-backend]
    proof: marker

  - id: prompt-studio-author
    description: "Create a Prompt Studio project and add a prompt to it."
    covered_by: [integration-backend]
    proof: marker

  - id: prompt-studio-fetch-response
    description: "Prompt Studio: create project, add prompt, run a prompt, get response."
    covered_by: [e2e-prompt-studio]
    proof: marker

  - id: connector-register-test
    description: "Connector credentials are validated against the live system and stored encrypted."
    covered_by: [integration-backend]
    proof: marker

  - id: pipeline-etl-execute
    description: "Run an ETL pipeline from source connector to destination."
    covered_by: [e2e-etl]
    proof: marker

  - id: usage-aggregate-read
    description: "Per-run token usage aggregates correctly and stays scoped to its organization."
    covered_by: [integration-backend]
    proof: marker

  - id: usage-token-tracking
    description: "Per-execution token usage is recorded and retrievable."
    covered_by: [e2e-api-deployment]
    proof: marker

  # Deliberate gap. The rejoin half is covered by e2e-api-deployment, but the
  # fan-out half has no observable: no batch or task id is persisted, and timing
  # can't stand in for one (see the test's module docstring). Closing it needs
  # batch identity on WorkflowFileExecution, not another test.
  - id: workflow-execution-fan-out
    description: "Multi-file workflow execution fans out to file-processing workers and rejoins."
    covered_by: []

  - id: callback-result-delivery
    description: "Async results are posted back via the callback worker."
    covered_by: [e2e-api-deployment]
    proof: marker