Add superml CLI: low-token terminal commands with automatic experiment journaling#1
Open
BlackhatShiftey wants to merge 3 commits into
Open
Add superml CLI: low-token terminal commands with automatic experiment journaling#1BlackhatShiftey wants to merge 3 commits into
BlackhatShiftey wants to merge 3 commits into
Conversation
Adds cli/ — a standalone bash CLI that runs the same ML engineering workflows (plan, debug, research, verify, iterate, experiment) as direct terminal commands without the plugin session overhead. - Uses claude -p --system-prompt with compact ~300-token prompts instead of injecting the full SKILL.md via session-start hooks - No leeroopedia auth required; uses Claude Code OAuth (existing claude.ai subscription) - ~5-10x fewer tokens per call vs plugin mode - Defaults to haiku, overridable with --model or SUPERML_MODEL Install: cd cli && bash install.sh
When --ground is passed, the CLI fetches 2-3 official doc pages for the frameworks in the task before writing any content. Cites exact API names, config keys, and version-specific behavior from fetched pages inline as [Label](URL). URL registry covers: Transformers, PEFT, TRL, Axolotl, vLLM, TGI, SGLang, DeepSpeed, FSDP, LangChain, LangGraph, lm-eval-harness. Usage: superml plan --ground "...", superml debug -g "..."
Every call now automatically: - Reads MEMORY.md, experiments/lessons.md, experiments/journal.md (last 150 lines) and injects them into context before responding - Captures output and appends a timestamped entry to experiments/journal.md after each successful call This happens at the bash level for every skill, every call — no invocation or instruction required. Use --no-log to skip journaling for one-off queries.
Contributor
|
Perfect, I'll run checks and come back to you! |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds
cli/— a standalone bash CLI that runs the same ML engineering workflows as direct terminal commands, without plugin session overhead or leeroopedia auth.experiments/journal.mdat the bash level, no invocation neededMEMORY.md,experiments/lessons.md, and recent journal entries before every call--groundflag — fetches official docs via WebFetch before responding (same URL registry as the plugin's web mode)--no-logflag — skip journaling for throwaway queriesUsage
Relation to the plugin
The CLI and plugin are complementary. The plugin's KB mode (with a leeroopedia key) gives richer grounded responses. The CLI is the faster path when you want a quick answer, are iterating rapidly, or don't have a key. The
--groundflag closes most of the gap using WebFetch from official docs.Files added