onto-mcp 0.4.13 → 0.4.14

package/dist/mcp/tool-schemas.js

@@ -1,5 +1,6 @@
 import { z } from "zod";
 import { RECONSTRUCT_DOMAIN_ID_GRAMMAR_DESCRIPTION, RECONSTRUCT_DOMAIN_ID_PATTERN, } from "../core-runtime/reconstruct/domain-id.js";
+import { PerCallLlmOverrideSchema } from "../core-runtime/discovery/settings-chain.js";
 const ReviewModeSchema = z.enum(["core-axis", "full"]);
 const ReviewTargetScopeKindSchema = z.enum(["file", "directory", "bundle"]);
 const ReviewExecutionRouteSchema = z.enum([
@@ -29,10 +30,21 @@ const OntoReviewToolInputBaseSchema = z.object({
     confirmValueAlignment: z.boolean().optional(),
     prepareOnly: z.boolean().optional(),
     returnRunningAfterMs: z.number().int().min(0).optional(),
+    // Per-call LLM override: ephemeral settings-`llm` overlay applied to every
+    // review dispatch seat (all actors + units) for this invocation. Omit → settings
+    // unchanged (default-off, byte-identical). See tool-schemas import.
+    llmOverride: PerCallLlmOverrideSchema.optional(),
 }).strict();
-export const OntoReviewToolInputSchema = OntoReviewToolInputBaseSchema.refine((input) => !(input.domain && input.noDomain), {
+// A provider switch needs an explicit model (the switched-in provider has no
+// default model to inherit); `auth` stays optional — the model-switcher
+// normalizer defaults it per provider, matching settings behavior.
+const requireModelWhenOverrideProvider = (input) => !(input.llmOverride?.provider && !input.llmOverride.model);
+const overrideModelRefineMessage = "llmOverride.model is required when llmOverride.provider is set (a provider switch needs an explicit model).";
+export const OntoReviewToolInputSchema = OntoReviewToolInputBaseSchema
+    .refine((input) => !(input.domain && input.noDomain), {
     message: "Use either domain or noDomain, not both.",
-});
+})
+    .refine(requireModelWhenOverrideProvider, { message: overrideModelRefineMessage });
 export const OntoPrepareReviewToolInputSchema = OntoReviewToolInputBaseSchema.extend({
     prepareOnly: z.literal(true).default(true),
 }).refine((input) => !(input.domain && input.noDomain), {
@@ -90,15 +102,20 @@ export const OntoReconstructToolInputSchema = OntoObserveSourceToolInputSchema.e
     resumeMode: z.enum(["fresh", "reuse_existing_authored_artifacts"]).optional(),
     semanticAuthorRealization: z.enum(["direct_call"]).default("direct_call"),
     confirmationProviderRealization: z.enum(["direct_call"]).default("direct_call"),
-    // Optional reasoning-effort pin applied to both reconstruct actors (live only).
-    llmEffort: z.string().min(1).optional(),
     // Opt-in per-stage answer-support JUDGE overrides (live only). judgeLlmEffort
     // runs the judge at a different effort; judgeModel swaps the judge MODEL on the
     // semantic author's provider. An unsupported judgeModel degrades to the author
     // model (INV-MODEL-1); the judge otherwise inherits the semantic-author config.
     judgeLlmEffort: z.string().min(1).optional(),
     judgeModel: z.string().min(1).optional(),
-}).strict();
+    // Per-call LLM override: ephemeral settings-`llm` overlay applied to the
+    // reconstruct actor seats (semantic_author, confirmation_provider,
+    // semantic_map_synthesize, dispatch_fallback) for this invocation. The judge
+    // keeps its own judgeModel/judgeLlmEffort knobs. Omit → settings unchanged.
+    llmOverride: PerCallLlmOverrideSchema.optional(),
+}).strict().refine(requireModelWhenOverrideProvider, {
+    message: overrideModelRefineMessage,
+});
 export const OntoReconstructSessionInputSchema = z.object({
     sessionRoot: z.string().min(1),
     projectRoot: z.string().min(1).optional(),

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "onto-mcp",
-  "version": "0.4.13",
+  "version": "0.4.14",
   "description": "MCP-native ontology review runtime with context-isolated lenses and controlled deliberation",
   "homepage": "https://github.com/kangminlee-maker/onto-mcp#readme",
   "bugs": {