DriftQ logo
DriftQ

Roadmap

Where the DriftQ project is going next (v2 and beyond).

Note: this roadmap is directional and subject to change.

DriftQ is the overall project/org. DriftQ-Core is the main GitHub repo that contains the v1 broker implementation.

v1: DriftQ-Core (MVP reliability foundation)

Completed

v1 is the reliability layer: the broker semantics you need before any workflow system can be trusted.

  • Leasing — safe ownership + crash recovery
  • Retries — backoff + max attempts
  • DLQ — strict dead-letter routing
  • Idempotency — prevent duplicate side effects under redelivery
  • Observability — inflight, lag, DLQ totals, backpressure rejects

Status: v1 MVP is complete. Next is building “workflow-level” capability on top of it.

v2: Replayable Workflow Runtime — Time-travel debugging and safe retries for AI workflows

In progress

v2 turns the reliability foundation into a practical workflow engine focused on AI pipelines and RAG.

  • Packaging & docs — polished local-first run path + clean docs
  • Replayable runs — append-only run history + deterministic replay / re-drive from step N
  • Workflow primitives — timers/delay, cancel, join/barrier, timeouts
  • Budgets & throttles — attempt/time/cost limits, concurrency caps, explicit backpressure
  • Developer tooling — Run Inspector, attempt diffs, DLQ drilldown + fingerprints
  • Observability — trace/run correlation + step metrics to answer “slow/stuck/costing” fast
  • Templates & demos — starter workflows (including a one-command RAG Index Builder demo)

Intent: make AI workflows reproducible, debuggable, and safe to retry — not "best-effort scripts."

Future (planned / TBD)

  • v3: Multi-Agent Runtime & Real-Time AI

    Planned

  • v4: DriftQ Cloud (Managed AI Orchestration Platform)

    Planned

  • v5: AI Nervous System (Long-Term Vision)

    Planned

Expectation: details for future versions stay intentionally high-level until they’re actively built.