ML inference backends, scoring engine, agent orchestrator
  • Go 96.9%
  • JavaScript 1.8%
  • Shell 1.3%
Find a file
Snider 087a470136
Some checks failed
security-scan.yml / wip: local inference improvements (Codex sandbox handoff) (push) Failing after 0s
test.yml / wip: local inference improvements (Codex sandbox handoff) (push) Failing after 0s
wip: local inference improvements (Codex sandbox handoff)
Bundling in-flight work in go/ + go/cmd so go-mlx can pull this
repo in as an external/ submodule for the auto-tuning Codex
sandbox to see.

New surfaces:
- go/lql.go (+ example + test) — LARQL primitive
- go/capability_example_test.go

Refactor / iteration touches most of go/ + go/cmd (backend,
adapter, agent, train, serve, expand, ingest, gguf, parquet,
ollama, publish, status, worker, workflow, etc.) plus the
serve_backend_{default, mlx} surfaces and the tui_train shell.

WIP — passes local build but no test sweep before this commit;
revisit when Codex returns its sandbox findings.

Co-Authored-By: Virgil <virgil@lethean.io>
2026-05-16 17:57:05 +01:00
.core chore: add .core/ build and release configs 2026-03-06 18:52:36 +00:00
.forgejo/workflows ci: add Forgejo Actions test and security scan workflows 2026-02-23 03:29:04 +00:00
.idea fix: align dependency versions for forge resolution 2026-03-14 14:47:08 +00:00
docs refactor(ml): coreerr.E → core.E sweep + workflow surface + per-file docs 2026-05-11 10:47:37 +01:00
external chore(deps): update inference openai adapter 2026-05-08 16:34:48 +01:00
go wip: local inference improvements (Codex sandbox handoff) 2026-05-16 17:57:05 +01:00
.editorconfig chore: add Go repo norms (badges, contributing, lint, taskfile, editorconfig) 2026-02-23 06:45:41 +00:00
.gitignore chore(repo): cleanup tracked-artifacts + replace-directives (audit dims) 2026-05-01 10:39:38 +01:00
.gitmodules refactor(go): restructure to /go/ subtree (Phase 1, Mantis #1240) 2026-05-01 06:48:41 +01:00
.golangci.yml chore: add Go repo norms (badges, contributing, lint, taskfile, editorconfig) 2026-02-23 06:45:41 +00:00
AGENTS.md refactor(core): align go-ml with v0.9.1 hardened core/go reference shape 2026-04-29 04:51:37 +01:00
CLAUDE.md merge: resolve main→dev conflicts — migrate imports to dappco.re 2026-03-23 20:51:08 +00:00
CONTRIBUTING.md chore: add Go repo norms (badges, contributing, lint, taskfile, editorconfig) 2026-02-23 06:45:41 +00:00
go.work chore(ml): Phase 2 cascade — refresh submodules + go.work hygiene 2026-05-01 09:26:11 +01:00
go.work.sum refactor(ml): coreerr.E → core.E sweep + workflow surface + per-file docs 2026-05-11 10:47:37 +01:00
LICENCE chore: add EUPL-1.2 LICENCE file (UK English canonical) 2026-05-01 08:34:32 +01:00
README.md refactor: migrate module path to dappco.re/go/core/ml 2026-03-22 02:16:43 +00:00

Go Reference License: EUPL-1.2 Go Version

go-ml

ML inference backends, multi-suite scoring engine, and agent orchestrator for the Lethean AI stack. The package provides pluggable backends (Apple Metal via go-mlx, managed llama-server subprocesses, and OpenAI-compatible HTTP APIs), a concurrent scoring engine that evaluates model responses across heuristic, semantic, content, and standard benchmark suites, 23 capability probes, GGUF model management, and an SSH-based agent orchestrator that streams checkpoint evaluation results to InfluxDB and DuckDB.

Module: dappco.re/go/core/ml Licence: EUPL-1.2 Language: Go 1.25

Quick Start

import "dappco.re/go/core/ml"

// HTTP backend (Ollama, LM Studio, any OpenAI-compatible endpoint)
backend := ml.NewHTTPBackend("http://localhost:11434", "qwen3:8b")
resp, err := backend.Generate(ctx, "Hello", ml.GenOpts{MaxTokens: 256})

// Scoring engine
engine := ml.NewEngine(backend, ml.Options{Suites: "heuristic,semantic", Concurrency: 4})
scores := engine.ScoreAll(responses)

Documentation

Build & Test

go test ./...
go test -race ./...
go test -bench=. ./...
go build ./...

Licence

European Union Public Licence 1.2 — see LICENCE for details.