@barivia/barmesh-mcp 0.3.1 → 0.4.0

package/README.md

@@ -13,10 +13,14 @@ form on a shared self-organizing map (SOM)**:
 - **`barmesh_mesh_convergence`** — trains one SOM on all meshes (joint-normalized), projects
   each mesh to a volume-weighted fingerprint, and computes **symmetric KL** and
   **Wasserstein-1 (EMD)** distances stepwise and against a reference mesh, with publication
-  figures and an advisory convergence reading. Submit enqueues **`cfd_prepare`** on worker-io
-  when needed; the proxy auto-polls `prepare_job_id` before the mesh job runs. Default
-  **`defer_figures=true`** → **`cfd_finalize`** on worker-io; **`barmesh_jobs(status)`** auto-polls
-  **`finalize_job_id`** when figures are deferred.
+  figures and an advisory convergence reading. The SOM features are preprocessed by the
+  same staged pipeline as barsom training, so any dataset (small or large, CSV/gzip/Parquet)
+  is handled out-of-core by default; optional `transforms`, `normalize`,
+  `normalization_methods`, and `row_range` give the same preprocessing controls. Submit
+  enqueues **`prepare_training_matrix`** on worker-io when the dataset is staged; the proxy
+  auto-polls `prepare_job_id` before the mesh job runs. Default **`defer_figures=true`** →
+  **`cfd_finalize`** on worker-io; **`barmesh_jobs(status)`** auto-polls **`finalize_job_id`**
+  when figures are deferred.
 - **`barmesh_richardson`** — classical three-level Richardson extrapolation / Grid
   Convergence Index (GCI) on scalar quantities of interest.
 

package/dist/cfd_prepare.js

@@ -1,6 +1,8 @@
 import { pollUntilComplete } from "./shared.js";
 /**
- * When the API enqueues cfd_prepare for mesh-convergence, poll it before the mesh job runs.
+ * Mesh-convergence on a staged dataset enqueues a preprocessing job
+ * (prepare_training_matrix) first; the submit response carries its id as
+ * `prepare_job_id`. Poll it to completion before the mesh job runs.
  */
 export async function pollCfdPrepareIfPresent(data, label, timeoutMs = 600_000) {
     const prepareJobId = data.prepare_job_id;
@@ -8,10 +10,10 @@ export async function pollCfdPrepareIfPresent(data, label, timeoutMs = 600_000)
         return null;
     const poll = await pollUntilComplete(prepareJobId, timeoutMs);
     if (poll.status === "failed") {
-        throw new Error(`${label}: cfd_prepare job ${prepareJobId} failed: ${poll.error ?? "unknown error"}`);
+        throw new Error(`${label}: data preparation job ${prepareJobId} failed: ${poll.error ?? "unknown error"}`);
     }
     if (poll.status !== "completed") {
-        throw new Error(`${label}: cfd_prepare job ${prepareJobId} did not complete (status=${poll.status})`);
+        throw new Error(`${label}: data preparation job ${prepareJobId} did not complete (status=${poll.status})`);
     }
     return prepareJobId;
 }

package/dist/shared.js

@@ -20,7 +20,7 @@ export const FETCH_TIMEOUT_MS = parseInt(process.env.BARIVIA_FETCH_TIMEOUT_MS ??
 export const MAX_RETRIES = 2;
 export const RETRYABLE_STATUS = new Set([502, 503, 504]);
 /** Single source of truth for the proxy version. Keep in sync with package.json on bump. */
-export const CLIENT_VERSION = "0.3.1";
+export const CLIENT_VERSION = "0.4.0";
 export const PUBLIC_SITE_ORIGIN = "https://barivia.se";
 /** Large per-cell CSV uploads may exceed the default fetch timeout. */
 export const UPLOAD_DATASET_TIMEOUT_MS = 180_000;

package/dist/tools/cfd.js

@@ -27,6 +27,10 @@ COMMON MISTAKES: omitting feature_columns (required); choosing a reference_mesh
         emd_method: z.enum(["exact", "sinkhorn"]).optional().describe("EMD solver: exact LP (default) or sinkhorn (fast approximation for large grids)"),
         component_planes_physical: z.boolean().optional().describe("Physical-scale component-plane colorbars (default true)"),
         figures: z.boolean().optional().describe("Generate publication figures (default true)"),
+        transforms: z.record(z.enum(["log", "log1p", "log10", "sqrt", "square", "abs", "invert", "none"])).optional().describe("Per-feature transform applied before normalization (e.g. log1p to compress k/epsilon/omega). Same preprocessing engine as barsom training."),
+        normalize: z.union([z.enum(["all", "auto", "mad", "sigmoidal", "sepd"]), z.array(z.string())]).optional().describe("Normalization mode for SOM features (default auto = joint z-score of non-cyclic features). Use mad for heavy-tailed channels."),
+        normalization_methods: z.record(z.enum(["zscore", "mad", "sigmoidal", "sepd", "none"])).optional().describe("Per-feature normalization override; keys must be in feature_columns."),
+        row_range: z.tuple([z.number().int().min(1), z.number().int().min(1)]).optional().describe("1-based inclusive [start, end] row slice applied during preprocessing (and to mesh labels / cell volumes)."),
         label: z.string().optional().describe("Optional job label"),
     }, async (args) => {
         const { dataset_id, label, ...rest } = args;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@barivia/barmesh-mcp",
-  "version": "0.3.1",
+  "version": "0.4.0",
   "description": "barmesh MCP proxy — SOM-based CFD mesh-convergence and Richardson/GCI analysis on the Barivia cloud API",
   "keywords": [
     "mcp",