zooqueenandHanzo Dev
c8e51e390b
hanzoai/ci keeps the reusable at .github/workflows/build.yml — the one path both planes resolve, since github.com reads a reusable only from there and the forge reads it too. The v1 tag moved with it, so a caller still naming .hanzo/workflows/build.yml asks for a blob that is not in the tree: InsertRun: read hanzoai/ci@v1:.hanzo/workflows/build.yml: object does not exist The caller itself stays at .hanzo/workflows/cicd.yml on purpose: github.com has no runner for these labels, so a caller placed there could never be scheduled. Co-authored-by: Hanzo Dev <[email protected]>
Hanzo Research (Go)
The Go client for /v1/research: record an experiment as a falsifiable test — hypothesis,
attempts, verdict — and query the evidence back.
Stdlib only (net/http + encoding/json), no external modules, so any service can import
it without dragging in a dependency graph.
Install
go get github.com/hanzoai/research
Use it
c := research.New(research.Config{Project: "enso-bench"})
exp, _ := c.Experiment(kind, subject, task)
exp.Record(item, model, research.Result{ /* one attempt */ })
exp.Log("what happened")
exp.Conclude(research.Proven, because, value)
Finish(value) seals a run without an explicit verdict; a hypothesis with no verdict
records as Inconclusive. Record is idempotent, so a retried attempt does not double
count.
Authentication is your per-org API key, read from the environment like every other Hanzo client.
One record, every language
The server keys each experiment on (project, id = kind:subject:task), so the Python,
Rust and C++ clients upsert the same row this one does. JSON spacing differs between them
and is immaterial to the store. If you already know one of those clients, the verbs here
are the same.
Docs
LLM.md is the full surface and the semantics of each verb. example_test.go
is a runnable walkthrough.
License
See LICENSE.