Sandwich format wraps seed prompts with KB preamble (axioms framework)
and LEK-1 kernel postfix, then generates responses via local MLX
inference. Outputs chat JSONL compatible with 'core ml train'.
Supports --dry-run mode to output prompts without inference.
Co-Authored-By: Virgil <virgil@lethean.io>
Native MLX LoRA training on Apple Silicon — no Python required.
Reads chat-format JSONL, applies LoRA to target projections,
trains with AdamW + masked cross-entropy loss on assistant tokens.
Usage: core ml train --model-path /path/to/model --data training.jsonl
Co-Authored-By: Virgil <virgil@lethean.io>
Add compiled binary to gitignore. Remove entries for deleted paths
(cmd/bugseti/bugseti, internal/core-ide/core-ide).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Remove internal/bugseti/ (now core/bugseti repo), cmd/bugseti/
(now core/bugseti/cmd/), and internal/tools/ (i18n-validate
moved to core/go). core/cli internal/ is now empty.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
## Summary
- Extract PHP/Laravel commands to `core/php` repo (42 files, standalone module)
- Extract CI/release + SDK commands to `core/ci` repo (10 files)
- Remove `internal/variants/` build tag system entirely
- Move all 30 remaining command packages from `internal/cmd/` to top-level `cmd/`
- Rewrite `main.go` with direct imports — no more variant selection
- PHP and CI are now optional via commented import lines in main.go
Co-authored-by: Claude <developers@lethean.io>
Reviewed-on: #2
Co-authored-by: Charon <charon@lthn.ai>
Co-committed-by: Charon <charon@lthn.ai>
- Change module from forge.lthn.ai/core/go to forge.lthn.ai/core/cli
- Remove pkg/ directory (now served from core/go)
- Add require + replace for forge.lthn.ai/core/go => ../go
- Update go.work to include ../go workspace module
- Fix all internal/cmd/* imports: pkg/ refs → forge.lthn.ai/core/go/pkg/
- Rename internal/cmd/sdk package to sdkcmd (avoids conflict with pkg/sdk)
- Remove SDK library files from internal/cmd/sdk/ (now in core/go/pkg/sdk/)
- Remove duplicate RAG helper functions from internal/cmd/rag/
- Remove stale cmd/core-ide/ (now in core/ide repo)
- Update IDE variant to remove core-ide import
- Fix test assertion for new module name
- Run go mod tidy to sync dependencies
core/cli is now a pure CLI application importing core/go for packages.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: Claude <developers@lethean.io>
Reviewed-on: #1
Port the standalone lab dashboard (lab.lthn.io) into the core CLI as
pkg/lab/ with collectors, handlers, and HTML templates. The dashboard
monitors machines, Docker containers, Forgejo, HuggingFace models,
training runs, and InfluxDB metrics with SSE live updates.
New command: core lab serve --bind :8080
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Ports all remaining LEM pipeline commands from pkg/lem into core ml,
eliminating the standalone LEM CLI dependency. Each command is split
into reusable business logic (pkg/ml/) and a thin cobra wrapper
(internal/cmd/ml/).
New commands: query, inventory, metrics, ingest, normalize, seed-influx,
consolidate, import-all, approve, publish, coverage.
Adds Path(), Exec(), QueryRowScan() convenience methods to DB type.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Tracks model size at load time and checks Metal active memory after
each generation. If usage exceeds 3× model size, forces double GC
and cache clear as a safety net.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The Go wrapper was tracking inter-array references via desc.inputs,
creating chains that kept all intermediate arrays alive across requests.
After 3-4 requests, Metal memory grew to 170GB+ and macOS killed the
process.
Fix: remove desc.inputs/numRefs entirely. MLX-C has its own internal
reference counting — when Go GC finalizes an Array wrapper, it calls
mlx_array_free which decrements the C-side refcount. If the C-side
count reaches 0, Metal memory is freed. Go GC + MLX-C refcounting
together handle all lifecycle management correctly.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Go GC cannot see Metal/C memory pressure, so intermediate arrays from
each forward pass accumulated without bound, causing OOM kills after
3-4 requests. Fix: runtime.SetFinalizer on every Array releases C
handles when GC collects them, and runtime.GC() is forced every 4
tokens during generation. Also adds SetMemoryLimit(24GB) as a hard
Metal ceiling.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add ClearCache() wrapping mlx_clear_cache
- Clear Metal allocator cache every 8 tokens during generation
- Set 16GB cache limit on backend init
- Prevents GPU memory from growing unbounded during inference
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Gemma 3 tokenizer.json uses [["a","b"],...] format for merges
instead of the ["a b",...] format. Support both.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Supports both multimodal (Gemma3ForConditionalGeneration) and
text-only configs. Resolves weights with language_model. prefix
fallback. Computes head_dim from hidden_size when missing.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Use pure C callback instead of //export to avoid const char* vs
GoString type mismatch in cgo-generated headers.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Use _axis/_axes variants for softmax, argmax, topk, sum, mean, squeeze,
concatenate, argpartition
- Fix size_t vs int for count parameters throughout
- Fix int64_t strides in as_strided
- Add mlx_optional_int + mode param to quantized_matmul
- Use mlx_array_new() for null arrays (freqs, key, mask, sinks)
- Fix expand_dims to single-axis signature
- Fix compile callback signature (size_t index)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
CGo wrapper for mlx-c providing zero-Python Metal GPU inference.
Includes Gemma 3 model architecture, BPE tokenizer, KV cache,
composable sampling, and OpenAI-compatible serve command.
Build-tagged (darwin && arm64 && mlx) with stubs for cross-platform.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Port LEM scoring/training pipeline into CoreGo as pkg/ml with:
- Inference abstraction with HTTP, llama-server, and Ollama backends
- 3-tier scoring engine (heuristic, exact, LLM judge)
- Capability and content probes for model evaluation
- GGUF/safetensors format converters, MLX to PEFT adapter conversion
- DuckDB integration for training data pipeline
- InfluxDB metrics for lab dashboard
- Training data export (JSONL + Parquet)
- Expansion generation pipeline with distributed workers
- 10 CLI commands under 'core ml' (score, probe, export, expand, status, gguf, convert, agent, worker)
- 5 MCP tools (ml_generate, ml_score, ml_probe, ml_status, ml_backends)
All 37 ML tests passing. Binary builds at 138MB with all commands.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Move module identity to our own Forgejo instance. All import paths
updated across 434 Go files, sub-module go.mod files, and go.work.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The main.go was removed when Wails3 apps were added to cmd/, breaking
`go build .` for the core CLI. Restore it and update variants/full.go
to include daemon, forge, mcpcmd, prod, and session commands. Drop gitea
(superseded by forge) and unifi (unused).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add HubService to the Wails service list and attempt hub registration
at startup when hubUrl is configured. Drains any pending operations
queued from previous sessions.
Co-Authored-By: Virgil <virgil@lethean.io>
Replace no-op stubs with real implementations for queueOp, drainPendingOps,
savePendingOps, and loadPendingOps. Operations are persisted to hub_pending.json
and replayed on next hub connection — 5xx/transport errors are retried, 4xx
responses are dropped as stale. Adds PendingCount() for queue inspection.
Co-Authored-By: Virgil <virgil@lethean.io>
Add Register, Heartbeat, ClaimIssue, UpdateStatus, ReleaseClaim,
and SyncStats methods for hub coordination. ClaimIssue returns
ConflictError on 409 and calls drainPendingOps before mutating.
Co-Authored-By: Virgil <virgil@lethean.io>
Exchange a Forge API token for a hub API key by POSTing to
/api/bugseti/auth/forge. Skips if hub token already cached.
Adds drainPendingOps() stub for future Task 7 use.
Co-Authored-By: Virgil <virgil@lethean.io>
Add doRequest() and doJSON() methods for hub API communication. doRequest
builds full URLs, sets bearer auth and JSON headers, tracks connected
state. doJSON handles status codes: 401 unauthorised, 409 ConflictError,
404 NotFoundError, and generic errors for other 4xx/5xx responses.
Co-Authored-By: Virgil <virgil@lethean.io>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Introduce HubService struct with types for hub coordination: PendingOp,
HubClaim, LeaderboardEntry, GlobalStats, ConflictError, NotFoundError.
Constructor generates a crypto/rand client ID when none exists. Includes
no-op loadPendingOps/savePendingOps stubs for future persistence.
Co-Authored-By: Virgil <virgil@lethean.io>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add HubURL, HubToken, ClientID, and ClientName fields to Config struct
for agentic portal integration. Include getter/setter methods following
the existing pattern (SetForgeURL, SetForgeToken also added).
Co-Authored-By: Virgil <virgil@lethean.io>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace all exec.Command("gh", ...) calls with the existing pkg/forge
wrapper around the Forgejo Go SDK. BugSETI no longer requires the gh
CLI to be installed.
Changes:
- fetcher: use forge.ListIssues/GetIssue instead of gh issue list/view
- submit: use forge.ForkRepo/CreatePullRequest instead of gh pr create
- seeder: use git clone with forge URL + token auth instead of gh clone
- ghcheck: CheckForge() returns *forge.Client via forge.NewFromConfig()
- config: add ForgeURL/ForgeToken fields (GitHubToken kept for migration)
- pkg/forge: add Token(), GetCurrentUser(), ForkRepo(), CreatePullRequest(),
ListIssueComments(), and label filtering to ListIssuesOpts
Co-Authored-By: Virgil <virgil@lethean.io>
The workspace map previously only cleaned up during Capture() calls,
meaning stale entries would accumulate indefinitely if no new captures
occurred. This adds:
- Background sweeper goroutine (Start/Stop lifecycle) that runs every 5
minutes to evict expired workspaces
- Configurable MaxWorkspaces and WorkspaceTTLMinutes in Config (defaults:
100 entries, 24h TTL) replacing hardcoded constants
- cleanup() now returns eviction count for observability logging
- Nil-config fallback to safe defaults
Fixes#54
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Prevent path traversal in Journal.Append() by validating RepoOwner and
RepoName before using them in file paths. Malicious values like
"../../etc/cron.d" could previously write outside the journal baseDir.
Defence layers:
- Reject inputs containing path separators (/ or \)
- Reject ".." and "." traversal components
- Validate against safe character regex ^[a-zA-Z0-9][a-zA-Z0-9._-]*$
- Verify resolved absolute path stays within baseDir
Closes#46
CVSS 6.3 — OWASP A01:2021-Broken Access Control
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Move shared state initialization (issues, seen) and the load() call
inside the mutex scope in NewQueueService() to eliminate the race
window where concurrent callers could observe partially initialized
state. Remove the redundant heap.Init before the lock since load()
already calls heap.Init when restoring from disk.
Add documentation to save() and load() noting they must be called
with q.mu held.
Fixes#51
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Pass the API key via x-goog-api-key HTTP header instead of the URL
query parameter to prevent credential leakage in proxy logs, web
server access logs, and monitoring systems.
Resolves: #47 (CVSS 5.3, OWASP A09:2021)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Implements quota enforcement for agents including daily token limits,
daily job limits, concurrent job caps, model allowlists, and global
per-model budgets. Quota recovery returns 50% for failed jobs and
100% for cancelled jobs.
Go: AllowanceService with MemoryStore, AllowanceStore interface, and
25 tests covering all enforcement paths.
Laravel: migration for 5 tables (agent_allowances, quota_usage,
model_quotas, usage_reports, repo_limits), Eloquent models,
AllowanceService, QuotaMiddleware, and REST API routes.
Closes#99
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add ForgejoClient and ForgejoService to the Laravel app, providing a
clean service layer for all Forgejo REST API operations the orchestrator
needs. Supports multiple instances (forge, dev, qa) with config-driven
auto-routing, token auth, retry with circuit breaker, and pagination.
Covers issues, PRs, repos, branches, user/token management, and orgs.
Closes#98
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
