Add CLI v2, human docs, and Skill 2.0 agent package#2
Open
webwebb56 wants to merge 17 commits into
Open
Conversation
Exposes the Mass Dynamics API as MCP tools usable from Claude desktop (Chat and Cowork tabs). Covers experiments, datasets, and all pipeline types (normalisation/imputation, pairwise comparison, dose response).
- Add batch tool to collapse multiple sequential tool calls into one MCP call, with name→UUID chaining support for get_upload - Add files tools: read_csv_preview, load_metadata_from_csv, plan_wide_to_md_format; enforces entity-data boundary (never reads intensity files in full) - Add validate_upload_inputs tool to catch sample_name mismatches before create_upload - Improve create_upload docstring to clarify ExperimentDesign column handling - Switch _client.py to import MDClientV2 directly (v2 only, no factory)
…g analyses Claude was guessing parameter values (e.g. normalise method names) instead of looking them up. describe_pipeline(job_slug) returns the full parameter schema with valid_values for each field. All run_* docstrings now mandate calling it first.
create_upload, update_sample_metadata, run_dose_response, run_pairwise_comparison, and generate_pairwise_comparisons now all explicitly forbid manual construction of experiment_design/sample_metadata arrays. Docstrings mandate calling load_metadata_from_csv first and passing its output verbatim.
Three outputs from one codebase: - cli/: Click-based CLI importing from md-python SDK (v2 API) - batch command for multi-step operations (24x faster, 33% fewer tokens) - v2 routing (/uploads, flat payloads, v2 accept headers) - docs/: Human-readable install guide, command reference, workflow examples - skill/: Skill 2.0 format for Claude Cowork/Code - Pure orchestrator SKILL.md (no rules, just workflow routing) - instructions/, examples/good+bad/, eval/checklist - scripts/: build-skill.py (packages .skill), build-docs.py (auto-generates docs) Co-authored-by: Claude <noreply@anthropic.com>
Fix: batch dispatcher now correctly handles multi-word names with --by-name flag. Previously 'uploads get uPhos Tissue DIA-NN PTMs --by-name' only passed 'uPhos' to the API. Now it joins all tokens between 'get' and '--by-name' as the name. Add: dist/md-platform.skill — the packaged Skill 2.0 artifact. Contains the skill orchestrator, instructions, examples, eval checklist, references, and the CLI source bundled in scripts/md-cli/. Ready for Claude Cowork/Code install. Co-authored-by: Claude <noreply@anthropic.com>
Fix: _dispatch_batch_command in main.py now joins multi-word names correctly when --by-name flag is present. Same fix was already in batch.py (fork CLI). Both dispatchers now handle 'experiments get Name With Spaces --by-name'. Also adds 'uploads' as accepted command in standalone dispatcher for v2 compat. Rebuilt dist/md-platform.skill with the fix included. Tested: 4/4 batch commands pass (health, auth, get-by-id, get-by-name). Co-authored-by: Claude <noreply@anthropic.com>
…e auth UX Three targeted friction fixes based on blind-user simulation analysis: 1. Design file elimination: --conditions inline flag, md design infer command, parse_conditions_string/resolve_design helpers 2. Table/column discovery: md tables list command, column reference docs in SKILL.md 3. Auth & output UX: 3-path auth guidance, --format ids-only, output_result helper, enterprise deployment docs Simulation results: first-try success 80% → 95%, retries 4-8 → 2-4, blockers 3 → 1. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…rebuild skill Implements the two missing CLI modules: - api.py: MDClient facade wrapping md-python SDK (V2 primary, V1 fallback) for all 40+ operations the CLI needs (experiments, datasets, analysis, tables, workspaces, viz, enrichment, entities) - config.py: persistent credential storage (~/.md-cli/config.json) with env var override support New CLI commands leveraging Aaron's PR MassDynamics#3 V2 endpoints: - md experiments query (search/filter uploads) - md experiments delete (V2) - md experiments metadata (sample metadata retrieval) - md datasets query (cross-upload dataset search) - md datasets download-url (presigned S3 URLs) - md entities query (cross-dataset protein/gene search) Updated SKILL.md with V2 command documentation. Rebuilt dist/md-platform.skill (40KB, self-contained). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Adds CLI command and api.py method for updating sample metadata on an existing upload. Accepts a CSV file (sample_name, condition, ...). Useful for correcting conditions, adding covariates, or relabelling samples post-upload without re-uploading data. Updated SKILL.md and rebuilt dist/md-platform.skill. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The batch dispatcher only handled basic V1 commands (get, list). Added routing for: experiments query/metadata/delete, datasets query/ download-url, and jobs list. Also added _parse_opts() helper for extracting --key value pairs from batch command args. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…infrastructure SKILL.md fixes (from eval benchmark iteration-4b): - Enrichment: add hard rule to always use `md` CLI, never HTTP notation - Entities: constrain --keyword to gene/protein names (not qualitative terms); document that --intersect does not exist; add BRD4 example - Entities: add `md datasets query --state COMPLETED` as canonical discovery command - Entities: promote entity_mapping_search Flipper flag to prominent 403 warning - batch-usage.md: clarify `jobs list` vs `jobs` (bare command returns nothing useful) Pairwise dataset builder: - Add log_intensities and use_imputed_intensities fields to PairwiseComparisonDataset so callers can explicitly control limma log-space and imputation behaviour Fixture infrastructure (eval harness, not production): - Add src/md_python/fixtures.py: HTTP record/replay activated only when MD_RECORD_FIXTURES or MD_REPLAY_FIXTURES env vars are set; inert otherwise - Add src/md_python/__init__.py hook to install fixtures when env vars present - Add tests/test_fixtures.py - Add skill/references/md-format-upload.md reference document Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
webwebb56
force-pushed
the
feature/cli-and-skill
branch
from
86c3ff1 to
3a4ef50
Compare
infusini
force-pushed
the
feature/mcp-server
branch
3 times, most recently
from
09717e8 to
e4c5bad
Compare
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
3 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
Test plan
🤖 Generated with Claude Code
Co-authored-by: Claude noreply@anthropic.com