Hanzo Research · /v1/research

R&D Ops Board

connecting…
every completed run = versioned evidence
local durable · cloud mirroring rolling out
Program progress — runs by status
shipped running planned

Research projects — enso · hanzo-engine · hanzo-ml

What's left — the critical path

Ordered, evidence-gated — what the team ships next

  1. Native Research SDK (Python first, then Rust · Go · C++) — one fluent call records a run with markdown notes + issue/PR/commit/benchmark links to /v1/research.
  2. Git-driven runs.github/workflows sync to .hanzo/workflows; every prod update runs the full benchmark suite as a regression + error gate on Hanzo CI/CD.
  3. Self-service billing — signup → own org → starter credit → scoped key → enso → metered.
  4. Enso Critic (candidate selector) → ScoutVOI Controller → typed Conductor — the adaptive-inference program (HIP-0512).
  5. hanzo MCP query — read historic research + benchmark evidence over the API, always-correct and queryable.

Native cloud plane — how a run reaches the board

Research SDK (py · rust · go · c++) or .hanzo/workflows CI run
  → POST /v1/research/experiments (idempotent by content hash · markdown notes + links)
  → per-org SQLite (versioned, append-only) → hanzoai/datastore OLAP (aggregate)
  → this board (GET /v1/research/projects · /totals · /experiments) + polling analytics
  → hanzo MCP query · informs enso config / preset / training via the Genome archive