feat(typesafe): add Jev decision integration

[BillLeoutsakosvl346]

Summary

Add a TypeSafe (Jev) block and four Agent tools: Choose an Option, Assign a Score, Estimate Yes/No Probability (Noul), and Evaluate Questions. They call TypeSafe’s System One endpoint with a user-supplied API key and use Sim’s existing request transport, retries, and model-input projection.

Preserve structured inputs, nullable question descriptions, fractional scores, complete probability maps, and named mixed-question answers. Validate choices, probability keys, score ranges, and structured legends against the request actually sent after model-input projection. Include seven catalog templates and seven suggested skills, documentation, and regenerated tool metadata, catalog, navigation, icon mappings, block-name metadata, and Copilot docs manifest. No SDK dependency, provider implementation, API route, OAuth, billing, or migrations.

Use the user-supplied standalone TypeSafe mark (typesafe-ai-dark.svg) for the block, catalog templates, and documentation. Preserve its exact path and viewBox; use currentColor for light/dark theme support.

Type of Change

• Bug fix
• New feature
• Breaking change
• Documentation
• Other

Testing

Used Bun 1.4.1 and installed existing locked dependencies with --frozen-lockfile.

All PR checks passed on 812299aeb4: app build, lint/type checks and repository audits, test shards, both PostgreSQL integration jobs, companion, security scan, docs preview, and automated review checks. Greptile reports 5/5; both reviewers have zero open threads.

Passed:

• Workspace formatting, lint, and type checks.
• All 47 repository audits, plus block-registry and canvas-sentence checks.
• Tool request-boundary, registry-boundary, parameter-reachability, and strict API-validation checks.
• Tool metadata, integration catalog, deployment configuration, documentation, and docs-manifest drift checks.
• Bare-icon, icon-path, and icon-precision checks with the supplied TypeSafe icon.
• 368 existing tests across executor, request transport, model-input adapters, block definitions, and parameter handling.

Manual verification:

• All four block forms and Agent operations appear in the local UI. Criteria examples and required state match each operation; the advanced model combobox offers the three initial suggestions and accepts a custom ID.
• Request/response probes covered text, JSON strings, resolved objects/arrays, shared reference navigation, fractional scores, zero/one Noul values, mixed answers, invalid JSON/criteria, missing or mismatched answers, and low confidence as success.
• Nested model-input projection preserves authentication, model selection, and question-type controls, and rejects ambiguous question-ID projection instead of reassigning content.
• Live HTTP 200 calls passed for all four operations using the configured key. Score returned 0.67 with a structured legend; mixed answers preserved question IDs and zero probabilities. Live negative calls confirmed HTTP 401 for invalid credentials and HTTP 400 for an invalid question type.

Live throttling/overload and cancellation were not induced; retry/error behavior uses the existing executor and its existing tests. No complete Agent model-driven conversation was executed.

No tests were added or modified, as requested. Existing CI remains enabled. The API key is excluded from the commit.

The final response-validation change also passed all 47 audits, workspace type checks, 228 existing executor/transport/projection tests, manual malformed-response and projection probes, and four successful live API calls. Probes included undeclared choices, missing/extra outcomes, out-of-range scores, mismatched legends, and comparison against projected option names and descriptions.

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing — no test changes requested; existing tests passed
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

Screenshots/Videos

Local UI verification completed. The final React icon was visually checked in a browser at 16, 24, 32, and 48 px on light, dark, and block-tile backgrounds. Its original SVG path and viewBox are preserved exactly.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated
docs	Skipped		Sep 21, 2026 9:21pm UTC

[cubic-dev-ai]

No issues found across 24 files

Confidence score: 5/5

• Automated review surfaced no issues in the provided summaries.
• No files require special attention.

_{Re-trigger cubic}

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge, with the previously identified response-validation gap addressed and no new actionable issue found.

Summary

Adds a TypeSafe (Jev) workflow integration with four operations, Agent-tool support, structured input projection, response validation, documentation, catalog metadata, and generated registry artifacts. The latest revision completes criteria-relative validation using the request body produced after model-input projection.

• Adds Choice, Score, Noul, and mixed-question Evaluate operations.
• Validates answer IDs, types, option/rubric keys, score bounds, and score legends against the projected request.
• Registers the integration across workflow blocks, Agent tools, documentation, icons, deployment metadata, and generated catalogs.
• Both previous review threads are resolved: criteria validation was implemented, while additional test files were intentionally excluded from the agreed scope.

Diagram

sequenceDiagram
  participant User
  participant Block as TypeSafe Block / Agent Tool
  participant Projector as Model-input projection
  participant Executor as Sim request executor
  participant Jev as TypeSafe System One
  participant Validator as Response validator

  User->>Block: State, question criteria, API key, model
  Block->>Projector: Select model-visible input
  Projector->>Executor: Projected parameters
  Executor->>Jev: POST state, model, and questions
  Jev-->>Executor: Typed answers, model, and usage
  Executor->>Validator: Response and serialized request body
  Validator->>Validator: Check IDs, types, criteria, bounds, and legends
  Validator-->>User: Operation-specific workflow output

Loading

_{Reviews (2) · Last reviewed commit: "fix(typesafe): validate answers against ..."}

[BillLeoutsakosvl346]

@greptile

[BillLeoutsakosvl346]

@cubic-dev-ai review this PR

[cubic-dev-ai]

@cubic-dev-ai review this PR

@BillLeoutsakosvl346 I have started the AI code review. It will take a few minutes to complete.

[cubic-dev-ai]

No issues found across 26 files

Confidence score: 5/5

• Automated review surfaced no issues in the provided summaries.
• No files require special attention.

_{Re-trigger cubic}

[waleedlatif1]

superseded by #8103