SWEny, YAML workflows for AI agents I'm running in prod (triage, E2E)

Minor Changes

• 0a59479: Add mcpServers to TriageConfig, ImplementConfig, and SharedNodeConfig. When provided, MCP server configs are forwarded to every codingAgent.run() call (investigate, implement-fix, create-pr, notify), enabling the coding agent to use MCP tools throughout the full recipe execution.

• 556a53d: Add optional provider field to StateDefinition for declaring which provider category a state uses. Updates the built-in triageDefinition and implementDefinition with full provider annotations and rich description text on every state. Pure metadata — no runtime behaviour change.

• 4465923: Add deterministic idempotency to triage recipe via content fingerprinting.

  • fingerprintEvent() — SHA-256 content hash of stable event fields (16-char hex)
  • inMemoryDedupStore() — Map-backed store with configurable TTL (default 24h)
  • TriageConfig.dedupStore — optional; new dedup-check DAG step short-circuits
    to notify before any LLM or provider calls when fingerprint already seen
  • Exports DedupStore, inMemoryDedupStore, fingerprintEvent from engine index

• 4b4b29f: Updated triage build-context step to use renamed provider interfaces:
  canListTriageHistory → canSearchIssuesByLabel, listTriageHistory → searchIssuesByLabel.

Patch Changes

• 130138e: Narrowed provider type hints in cross-repo-check and implement verify-access steps
  to use RepoProvider instead of the full SourceControlProvider where only remote
  API operations are needed.

• 42f6e95: Remove dead re-export shims from triage/steps. The create-pr, implement-fix, and notify step files were one-line re-exports pointing to nodes/ — no architectural purpose since nothing consumed them. Tests and the recipe index now import directly from nodes/.

• 010b6d7: Code review fixes for MCP adapters and triage dedup step.

  providers:

  • MCPClient: clear connectPromise on error to prevent stuck reconnections; use local variable before callTool to guard against concurrent disconnect
  • linearMCP: throw on empty id/identifier in toIssue(); expose limit option in searchIssuesByLabel() (default 100)
  • githubMCP: validate repo config as owner/repo format; guard against PR number=0; validate targetRepo in dispatchWorkflow
  • slackMCP: clarify channel config must be a Slack channel ID (e.g. C123456), not a name
  • LabelHistoryCapable: add limit?: number to searchIssuesByLabel opts

  engine:

  • dedup-check step: rename outcome from "notify" → "duplicate" so routing key is distinct from target node name
  • triage definition: update on: { duplicate: "notify" } to match

• Updated dependencies [2f1a424]

• Updated dependencies [68780d5]

• Updated dependencies [207a317]

• Updated dependencies [9313ff9]

• Updated dependencies [556a53d]

• Updated dependencies [ebbb5a7]

• Updated dependencies [010b6d7]

• Updated dependencies [f33c74d]

• Updated dependencies [1df08e0]

  • @sweny-ai/[email protected]