The bare catch {} swallowed the ApiError thrown by the private-IP
checks, defeating the entire DNS-based SSRF protection. Now catches
and re-throws ApiError so security rejections propagate correctly;
only DNS-lookup failures fall through.
- Add google-generative-ai to CustomProvider.apiType union type
- Update assertApiType to accept google-generative-ai in create/update
- Fix createCustomProvider mapping in legacy.ts for Google type
- Fix fetchCustomProviders mapping for Google type
- Add google-generative-ai to CustomProvidersSection API_TYPES
- Add SSRF protection to probeProviderModels (block private/loopback)
- Add body validation to probe-models route handler
- Update stale JSDoc in legacy.ts
Adds an "advisory" `gateMode` option for workflow steps: steps in advisory mode log findings but no longer block merge completion, while steps that must block can still be configured as blocking. The change covers the workflow step types and schema, executor gating logic, dashboard UI in WorkflowSte
Fusion-Task-Id: FN-4468
Implements the `experiment-finalize` CLI command (FN-4222) and its companion dashboard API route, wiring the feature through the pi extension as a new callable tool. Includes the command implementation, extension tooling, API integration, and corresponding test coverage.
Fusion-Task-Id: FN-4222
Add 'Detect Models' button to custom provider forms that calls the
provider's /models endpoint to discover available models automatically.
- Supports OpenAI-compatible, Anthropic-compatible, and Google
Generative AI providers
- Auto-fills context window and max tokens from provider response
- Filters out embedding, reranking, and non-text models
- Removes empty default model rows after detection
- Added comprehensive backend and frontend test coverage
Adds a tracking issue mechanism for GitHub-imported tasks, creating an internal Fusion task to track the import lifecycle. The TaskCard component gains visibility into tracking status, backed by new tests in both TaskCard and GitHub tracking modules.
Fusion-Task-Id: FN-4271
The rule flagged every committed reservation pointing at an existing task ID
as an anomaly, but that's the happy-path steady state — reservations
transition to `committed` immediately after the task row is inserted, so a
committed reservation is always expected to reference an existing task. On
any node with task history, the dashboard banner fired with hundreds of
"affected" IDs and the store emitted a spurious `[task-id-integrity] anomaly
detected` error log.
Removes the rule, its type/label/reader, and updates tests (core regression
guard now asserts committed reservations don't trigger anomalies; dashboard
server-test fixtures use a still-valid anomaly kind).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Implements automatic closing of linked GitHub issues when a task is deleted (FN-4253), adding deletion logic to the GitHub tracking state module with corresponding tests for both state management and deletion behavior, plus a patch changeset for the `@runfusion/fusion` CLI.
Fusion-Task-Id: FN-4253
- Add dashboard API routes and mission interview store support to list and inspect draft missions
- Update Mission Manager UI to surface draft interview sessions with styling and regression coverage
- Extend CLI mission commands and extension tools to expose draft listings consistently
- Document the draft surfacing behavior and include a changeset for the published CLI package
Fusion-Task-Id: FN-4124
Implements task ID integrity detection across the stack (FN-4192): a new `task-id-integrity` module in core provides detector logic wired into the store, integrity health is surfaced via the dashboard API and legacy endpoint, and a banner component surfaces alerts in the UI — backed by comprehensive
Fusion-Task-Id: FN-4192
