feat: Band 3 — Design Quantities and the Simulation MCP Server - #83
Open
tinix84 wants to merge 5 commits into
Open
feat: Band 3 — Design Quantities and the Simulation MCP Server#83tinix84 wants to merge 5 commits into
tinix84 wants to merge 5 commits into
Conversation
ruff I001 on the current lock's ruff; the pre-push gate requires a clean run. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Waveform capture over a deterministic steady-state window (last N complete periods of the declared switching frequency, start interpolated), per-component stress (time-weighted mean/RMS with exact piecewise-linear integration, peak, min/max, peak-to-peak, max slew rate) and a power balance (efficiency, total loss, per-component loss, unattributed remainder) — all pure functions of one Simulation Result and a caller-declared Signal Map. Nothing is inferred from signal names and nothing is written to the Cache Record (ADR-0012). Also records ADR-0011 (Simulation MCP Server as a second server), adds the Design Quantity and Signal Map terms to CONTEXT.md, and folds two README lines to stay under the 150-line cap. Refs: #78, #81, #52, #76 Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…ecycle `pyplecs-mcp-sim` is a second stdio server (ADR-0011) exposing simulation_submit / _status / _wait / _result / _cancel / _list as a thin transport over SimulationOrchestrator. Submission is rejected with an explicit tool error when PLECS is unavailable and no Simulation Task is created; every other tool keeps answering. A failed task is a Simulation Result with success=false, not a tool error. The result payload is the normalized time + named signals shape the REST sync route already returns. The ToolCatalogue learns to await coroutine handlers (dispatch_async, used by the stdio adapter) and to check object/array/number/boolean/enum/range arguments locally; the docs server's eight tools and sync dispatch are unchanged and asserted frozen. Refs: #77, #24, #75 Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…s over MCP/REST - Simulation MCP Server gains cache_lookup / cache_invalidate (the cache's own explain_miss names the differing Cache Key id), models_list (offline, via the Circuit Model parser) and simulation_quantities. - LivePlecsAdapter routes whole Simulation Requests to PLECS one model session per model file with a real XML-RPC reachability probe; the orchestrator now accepts any adapter offering simulate_requests, and pyplecs-mcp-sim is wired to it. Live runs against installed PLECS remain owed to the Band 4 map #56. - POST /simulations/{task_id}/quantities and the MCP tool share one payload helper, so REST, MCP and Python return identical Design Quantities computed on demand; the Cache Record is untouched (ADR-0012). Refs: #79, #80, #82, #24, #52 Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This was referenced Aug 28, 2026
- Name components that cannot be attributed a loss (`unattributable_components`) instead of folding them silently into the unattributed remainder. - `PlecsRequestPort` protocol replaces the hasattr probe; LivePlecsAdapter no longer carries a raising `simulate_batch`. - `models_list`: any per-file failure is reported per file; drop the unrequested `recursive` flag. - Drop the `create_simulation_mcp_server` middle man so importing pyplecs.mcp no longer imports the live adapter; use TIME_COLUMN; "metric" wording. - Restore LF on the pyproject script lines; test the non-completed → 400 path. Refs: #83 Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This was referenced Aug 28, 2026
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
1 participant
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.
Band 3 — the answer is legible
Stacked on #74 (Band 2); merge that first. Specs #75 (Simulation MCP Server, answering #59) and #76 (Design Quantities); tickets #77–#82. Decisions: ADR-0011, ADR-0012.
What changes for an engineer
pyplecs.quantitiesturns a Simulation Result into Design Quantities on demand: named waveforms over the last N complete switching periods, per-component stress (time-weighted mean/RMS with exact piecewise-linear integration, peak, min/max, peak-to-peak, max slew rate) and a power balance (efficiency, total loss, per-component loss, unattributed remainder). Roles come from a caller-declared Signal Map; nothing is inferred from names and nothing is written to the Cache Record. Reachable from Python,POST /simulations/{task_id}/quantities, and thesimulation_quantitiesMCP tool — identical payloads.pyplecs-mcp-simis the Simulation MCP Server:simulation_submit/_status/_wait/_result/_cancel/_list,cache_lookup/_invalidate,models_list,simulation_quantities, over stdio submit/poll with opaque ids. PLECS unreachable → explicit tool error at submission, no task created, every other tool still answers. The docs serverpyplecs-mcpis untouched and asserted frozen.LivePlecsAdapterroutes whole Simulation Requests to PLECS one model session per model file; the orchestrator's seam accepts any adapter offeringsimulate_requests.Commits
style:sort an import block the current ruff flags (pre-push gate).feat(quantities):Design Quantities + CONTEXT terms + ADR-0011/0012 + README rows.feat(mcp):Simulation MCP Server core; async-aware, typedToolCatalogue.feat:cache/model tools, live PLECS adapter, quantities over REST/MCP.Evidence
PlecsServer.Review (two-axis
/code-review, since 4f6e56d)Fixed in commit 5: silent drop of one-legged components from loss attribution (now
unattributable_components),PlecsRequestPortprotocol instead of ahasattrprobe, per-file failure capture inmodels_listand norecursiveflag, middle-mancreate_simulation_mcp_serverremoved,TIME_COLUMN, "metric" wording, pyproject line-ending churn, missing non-completed→400 test. Spec text amended on #76 (exact piecewise-linear integration, non-decreasing time axis, interpolated window start, zero-power error). Kept deliberately: README reflow (the 150-line cap), the third time+signals serializer (unifyingapi/tas.py/simulation_sync.pyis #46's concern),priorityin the task snapshot.Still owed / not in this PR
pyplecs-apistill builds an orchestrator with no PLECS adapter (pre-existing);LivePlecsAdaptercloses it in one line, left to architecture 4/5: concentrate the Simulation Task lifecycle #48.needs-info; Define the Simulation MCP execution contract to verify #59 and Choose the semi-manual .cir and .asc acceptance matrix #61 carry proposed answers awaiting the maintainer's decision.Closes #77, closes #78, closes #79, closes #80, closes #81, closes #82, closes #24, closes #52.
🤖 Generated with Claude Code