Skip to content

armgit / mso

Public Imported from GitHub
0

Browser-based graphical shell and control plane for a Linux server you own — real terminal, file manager, live metrics and BYOK AI in one mobile-first workspace. Next.js 16 · React 19 · node-pty · Tailwind v4 · shadcn/ui. Public alpha, self-hosted.

fix(jev): validate decision release contract

Rahmane5a1315df0f4
diff --git a/docs/CHANGELOG.md b/docs/CHANGELOG.md
index 356da8a8..efce35fb 100644
--- a/docs/CHANGELOG.md
+++ b/docs/CHANGELOG.md
@@ -8,10 +8,12 @@ this is the *what*, and it is what Settings → Account → About shows as “Wh
**Added**
+- `jev` persist reusable session capsules
- `settings` nest version and updates under About
**Fixed**
+- `jev` validate decision release contract
- `jev` open responsive session details in release journey
- `settings` align MCP release journey with breadcrumbs
- `settings` scroll MCP pages without reading a ref during render
@@ -20,16 +22,11 @@ this is the *what*, and it is what Settings → Account → About shows as “Wh
- `release` record Windows taskbar fidelity
-## 2026-09-27
-
-**Added**
-
-- `jev` persist reusable session capsules
-
## 2026-09-26
**Added**
+- `jev` productionize decision plane
- `settings` add an Account section for profile, password, devices, and about
- `jev` add policy-bound MCP actions
- `workflows` isolate source-changing agents
@@ -41,9 +38,9 @@ this is the *what*, and it is what Settings → Account → About shows as “Wh
**Fixed**
+- `jev` close release contract gaps
- `auth` rebuild the account profile before writing it
- `windows` match native taskbar proportions
-- `jev` close release contract gaps
- `worktrees` keep runtime paths out of build tracing
- `memory` keep runtime backup paths out of build tracing
- `tests` make recovery fixtures portable
diff --git a/docs/PROGRESS.md b/docs/PROGRESS.md
index 355ffbee..13f7bb74 100644
--- a/docs/PROGRESS.md
+++ b/docs/PROGRESS.md
@@ -21,6 +21,10 @@ separately from estimated avoided LLM impact, calibration/decision metadata, pre
coverage and an owner-only bounded JEV Lab. Workflow Sessions explicitly distinguish JEV
DECISION, MCP ACTION and VERIFIED RESULT and show the continuation capsule value summary.
+The public MCP action contract advances to toolset `2026.09.27.1`. External clients that
+cache MCP descriptors must refresh or rescan actions before using the revised `jev_action`
+schema and read scope.
+
## 2026-09-27 — Windows taskbar fidelity release
The Windows desktop shell taskbar now follows the supplied Windows 11 reference
diff --git a/docs/generated/MCP-CATALOG.md b/docs/generated/MCP-CATALOG.md
index a64a3de3..0f2a63cf 100644
--- a/docs/generated/MCP-CATALOG.md
+++ b/docs/generated/MCP-CATALOG.md
@@ -4,13 +4,13 @@
## Full MSO catalog
-<!-- mcp-toolset: server=1.16.0 version=2026.09.25.1 tools=115 read=58 write=34 exec=23 -->
+<!-- mcp-toolset: server=1.16.0 version=2026.09.27.1 tools=115 read=58 write=34 exec=23 -->
| Fact | Current source value |
|---|---:|
| MCP server | `1.16.0` |
-| Toolset | `2026.09.25.1` |
-| Toolset changed at | `2026-09-25T17:39:19+00:00` |
+| Toolset | `2026.09.27.1` |
+| Toolset changed at | `2026-09-27T05:05:41+00:00` |
| Transport tools | **117** |
| Model/operator tools | **115** |
| Read | **58** |
@@ -149,7 +149,7 @@
## ChatGPT model profile
-<!-- mcp-chatgpt-profile: server=1.16.0 version=2026.09.25.1 tools=115 read=58 write=34 exec=23 app-only=2 total=117 -->
+<!-- mcp-chatgpt-profile: server=1.16.0 version=2026.09.27.1 tools=115 read=58 write=34 exec=23 app-only=2 total=117 -->
The ChatGPT profile automatically projects the complete MSO-owned generic model/operator catalog. OAuth scope is still enforced independently; app-only compatibility bridges stay app-only, and project-owned MCP tool names remain dynamic data behind the generic project bridge.
diff --git a/lib/contracts/audit.ts b/lib/contracts/audit.ts
index 2d09a223..123aa7c6 100644
--- a/lib/contracts/audit.ts
+++ b/lib/contracts/audit.ts
@@ -29,7 +29,6 @@ export type AuditAction =
| "agent.message"
| "agent.subagent"
| "agent.memory"
- | "jev.decision"
| "tool.forge.propose"
| "tool.forge.evaluate"
| "tool.forge.promote"
diff --git a/lib/mcp/tool-contract.ts b/lib/mcp/tool-contract.ts
index aa7177f7..8aa8fd07 100644
--- a/lib/mcp/tool-contract.ts
+++ b/lib/mcp/tool-contract.ts
@@ -135,7 +135,7 @@ export function toolDescriptor(tool: McpTool, profile: McpToolProfile = "full")
return {
name: tool.name,
title: tool.title ?? toolTitle(tool.name),
- description: compact ? compactText(tool.chatgptDescription ?? tool.description, 44) : tool.description,
+ description: compact ? compactText(tool.chatgptDescription ?? tool.description, 48) : tool.description,
inputSchema: compact ? compactSchema(tool.inputSchema) : tool.inputSchema,
...(outputSchema ? { outputSchema: compact ? compactSchema(outputSchema) : outputSchema } : {}),
securitySchemes,
diff --git a/lib/mcp/toolset.ts b/lib/mcp/toolset.ts
index c146cc9f..a41c3582 100644
--- a/lib/mcp/toolset.ts
+++ b/lib/mcp/toolset.ts
@@ -4,8 +4,8 @@ import type { McpTool } from "./tool-kit";
import { toolDescriptor, type McpToolProfile } from "./tool-contract";
export const MCP_SERVER_VERSION = "1.16.0";
-export const MCP_TOOLSET_VERSION = "2026.09.25.1";
-export const MCP_TOOLSET_CHANGED_AT = "2026-09-25T17:39:19+00:00";
+export const MCP_TOOLSET_VERSION = "2026.09.27.1";
+export const MCP_TOOLSET_CHANGED_AT = "2026-09-27T05:05:41+00:00";
export type McpToolsetInfo = {
serverVersion: string;
diff --git a/lib/workflow/jev-decision-plane.test.ts b/lib/workflow/jev-decision-plane.test.ts
index 0c35e904..5305de16 100644
--- a/lib/workflow/jev-decision-plane.test.ts
+++ b/lib/workflow/jev-decision-plane.test.ts
@@ -1,6 +1,12 @@
-import { describe, expect, it } from "vitest";
+import { beforeEach, describe, expect, it, vi } from "vitest";
+const mocks=vi.hoisted(()=>({evaluate:vi.fn(),record:vi.fn()}));
+vi.mock("./jev-integration",()=>({resolveJevIntegrationConfig:vi.fn(async()=>({transport:"openrouter",provider:"openrouter",model:"typesafe/jev-1.13"}))}));
+vi.mock("./jev-openrouter",()=>({evaluateOpenRouterJev:mocks.evaluate}));
+vi.mock("./jev-telemetry",()=>({recordJevTelemetry:mocks.record}));
import { compileJevState, estimateJevTokens } from "./jev-state-compiler";
import { getJevDecisionDefinition, listJevDecisionDefinitions } from "./jev-decision-registry";
+import { runJevChoice, runJevNoul, runJevScore } from "./jev-decision-plane";
+beforeEach(()=>{mocks.evaluate.mockReset();mocks.record.mockReset();mocks.record.mockResolvedValue(undefined);});
describe("JEV decision registry and state compiler",()=>{
it("keeps decision policy versioned and decision-specific",()=>{
expect(getJevDecisionDefinition("action.select")).toMatchObject({version:1,type:"noul",threshold:.65,telemetryPolicy:"metadata-only"});
@@ -15,4 +21,13 @@ describe("JEV decision registry and state compiler",()=>{
expect(Buffer.byteLength(json,"utf8")).toBeLessThanOrEqual(24*1024+512);
expect(estimateJevTokens({a:"abcd"})).toBeGreaterThan(0);
});
+ it("rejects malformed provider decisions at the decision-plane boundary",async()=>{
+ const envelope=(decision:unknown)=>({model:"typesafe/jev-1.13",provider:"TypeSafe",answers:{decision},latencyMs:1});
+ mocks.evaluate.mockResolvedValueOnce(envelope({type:"choice",choice:"invented",probabilities:{invented:1},confidence:1}));
+ await expect(runJevChoice("model.route",{})).rejects.toThrow("invalid choice answer");
+ mocks.evaluate.mockResolvedValueOnce(envelope({type:"noul",noul:"yes"}));
+ await expect(runJevNoul("action.verify",{})).rejects.toThrow("invalid noul answer");
+ mocks.evaluate.mockResolvedValueOnce(envelope({type:"score",score:5,probabilities:{0:0,1:0,2:0,3:0,4:1},confidence:1}));
+ await expect(runJevScore("recipe.promote",{})).rejects.toThrow("invalid score answer");
+ });
});
diff --git a/lib/workflow/jev-decision-plane.ts b/lib/workflow/jev-decision-plane.ts
index 87d202cd..b5401295 100644
--- a/lib/workflow/jev-decision-plane.ts
+++ b/lib/workflow/jev-decision-plane.ts
@@ -13,6 +13,11 @@ export type JevScoreResult={kind:"mso.jev-decision.v1";decisionId:JevDecisionId;
const object=(v:unknown):v is Record<string,unknown>=>Boolean(v&&typeof v==="object"&&!Array.isArray(v));
const probability=(v:unknown)=>{const n=Number(v);return Number.isFinite(n)&&n>=0&&n<=1?n:0};
+const validProbability=(v:unknown):v is number=>typeof v==="number"&&Number.isFinite(v)&&v>=0&&v<=1;
+function providerProbabilities(value:Record<string,unknown>,keys:string[],answerType:string):Record<string,number>{
+ const rows=keys.map((key)=>{const p=value[key];if(!validProbability(p))throw new Error(`OpenRouter Jev returned invalid ${answerType} probabilities`);return[key,p] as const;});
+ return Object.fromEntries(rows);
+}
function baseTelemetry(id:JevDecisionId,ctx:JevDecisionContext):Pick<JevTelemetryEvent,"decisionId"|"decisionVersion"|"decisionType"|"correlationId"|"sessionRef"|"workflowRef">{
const d=getJevDecisionDefinition(id);return{decisionId:id,decisionVersion:d.version,decisionType:d.type,...(ctx.correlationId?{correlationId:ctx.correlationId}:{}),...(ctx.sessionRef?{sessionRef:ctx.sessionRef}:{}),...(ctx.workflowRef?{workflowRef:ctx.workflowRef}:{})};
}
@@ -35,10 +40,10 @@ export async function runJevChoice(id:JevDecisionId,state:Record<string,unknown>
const config=await resolveJevIntegrationConfig(raw),threshold=d.threshold;let probabilities:Record<string,number>={},choice="",confidence:number|undefined,model:string|undefined,latencyMs:number|undefined,inputTokens:number|undefined,actualCostUsd:number|undefined,provider:string=config.transport,requestId:string|undefined;
if(config.transport==="openrouter"){
const r=await evaluateOpenRouterJev(compileJevState(state),{decision:{type:"choice",instructions:d.instructions,criteria:d.criteria}},{model:config.model}),a=r.answers.decision;
- if(!object(a)||a.type!=="choice"||typeof a.choice!=="string"||!object(a.probabilities))throw new Error("OpenRouter Jev returned an invalid choice answer");
- choice=a.choice;probabilities=Object.fromEntries(Object.entries(a.probabilities).map(([k,v])=>[k,probability(v)]));confidence=typeof a.confidence==="number"?probability(a.confidence):undefined;model=r.model;latencyMs=r.latencyMs;inputTokens=r.usage?.inputTokens;actualCostUsd=r.usage?.actualCostUsd;provider=r.provider;requestId=r.requestId;
+ if(!object(a)||a.type!=="choice"||typeof a.choice!=="string"||!Object.hasOwn(d.criteria,a.choice)||!object(a.probabilities)||(a.confidence!==undefined&&!validProbability(a.confidence)))throw new Error("OpenRouter Jev returned an invalid choice answer");
+ choice=a.choice;probabilities=providerProbabilities(a.probabilities,Object.keys(d.criteria),"choice");confidence=a.confidence;model=r.model;latencyMs=r.latencyMs;inputTokens=r.usage?.inputTokens;actualCostUsd=r.usage?.actualCostUsd;provider=r.provider;requestId=r.requestId;
}else{
- const rows=fakeCandidates(d.criteria),e=await createResolvedJevWorkflowOptimizerEvaluator(config)(compileJevState(state),rows);probabilities=e.probabilities;
+ const rows=fakeCandidates(d.criteria),e=await createResolvedJevWorkflowOptimizerEvaluator(config)(compileJevState(state),rows);probabilities=Object.fromEntries(Object.keys(d.criteria).map((key)=>[key,probability(e.probabilities[key])]));
choice=Object.entries(probabilities).sort((a,b)=>b[1]-a[1])[0]?.[0]??"";model=e.jev?.model??config.model;latencyMs=e.jev?.latencyMs;inputTokens=e.jev?.inputTokens;actualCostUsd=e.jev?.actualCostUsd;provider=e.jev?.provider??"mcp";requestId=e.jev?.requestId;
}
const accepted=Boolean(choice)&&(probabilities[choice]??0)>=threshold;
@@ -50,7 +55,7 @@ export async function runJevNoul(id:JevDecisionId,state:Record<string,unknown>,i
const d=getJevDecisionDefinition(id);if(d.type!=="noul")throw new Error(`${id} is not a noul decision`);const config=await resolveJevIntegrationConfig(raw),threshold=d.threshold;let p=0,model:string|undefined,latencyMs:number|undefined,inputTokens:number|undefined,actualCostUsd:number|undefined,provider:string=config.transport,requestId:string|undefined;
if(config.transport==="openrouter"){
const r=await evaluateOpenRouterJev(compileJevState(state),{decision:{type:"noul",instructions:instructions??d.instructions}},{model:config.model}),a=r.answers.decision;
- if(!object(a)||a.type!=="noul")throw new Error("OpenRouter Jev returned an invalid noul answer");p=probability(a.noul);model=r.model;latencyMs=r.latencyMs;inputTokens=r.usage?.inputTokens;actualCostUsd=r.usage?.actualCostUsd;provider=r.provider;requestId=r.requestId;
+ if(!object(a)||a.type!=="noul"||!validProbability(a.noul))throw new Error("OpenRouter Jev returned an invalid noul answer");p=a.noul;model=r.model;latencyMs=r.latencyMs;inputTokens=r.usage?.inputTokens;actualCostUsd=r.usage?.actualCostUsd;provider=r.provider;requestId=r.requestId;
}else{
const row:WorkflowOptimizerCandidate={id:"yes",kind:"presentation-group",title:"yes",description:instructions??d.instructions,nodeIds:[],risk:"safe",estimatedNodeDelta:0,hostEligible:true},e=await createResolvedJevWorkflowOptimizerEvaluator(config)(compileJevState(state),[row]);p=e.probabilities.yes??0;model=e.jev?.model??config.model;latencyMs=e.jev?.latencyMs;inputTokens=e.jev?.inputTokens;actualCostUsd=e.jev?.actualCostUsd;provider=e.jev?.provider??"mcp";requestId=e.jev?.requestId;
}
@@ -60,8 +65,8 @@ export async function runJevNoul(id:JevDecisionId,state:Record<string,unknown>,i
export async function runJevScore(id:JevDecisionId,state:Record<string,unknown>,raw?:JevIntegrationInput,ctx:JevDecisionContext={}):Promise<JevScoreResult>{
const d=getJevDecisionDefinition(id);if(d.type!=="score"||!Array.isArray(d.criteria))throw new Error(`${id} is not a score decision`);const config=await resolveJevIntegrationConfig(raw);if(config.transport!=="openrouter")throw new Error("score decisions require the default OpenRouter JEV path");
const r=await evaluateOpenRouterJev(compileJevState(state),{decision:{type:"score",instructions:d.instructions,criteria:d.criteria}},{model:config.model}),a=r.answers.decision;
- if(!object(a)||a.type!=="score"||typeof a.score!=="number"||!object(a.probabilities))throw new Error("OpenRouter Jev returned an invalid score answer");
- const probabilities=Object.fromEntries(Object.entries(a.probabilities).map(([k,v])=>[k,probability(v)])),confidence=typeof a.confidence==="number"?probability(a.confidence):undefined;
+ if(!object(a)||a.type!=="score"||typeof a.score!=="number"||!Number.isFinite(a.score)||a.score<0||a.score>d.criteria.length-1||!object(a.probabilities)||(a.confidence!==undefined&&!validProbability(a.confidence)))throw new Error("OpenRouter Jev returned an invalid score answer");
+ const probabilities=providerProbabilities(a.probabilities,d.criteria.map((_,index)=>String(index)),"score"),confidence=a.confidence;
await recordJevTelemetry({event:"decision",...baseTelemetry(id,ctx),model:r.model,provider:r.provider,requestId:r.requestId,candidateCount:d.criteria.length,selected:[String(Math.round(a.score))],probabilities,threshold:d.threshold,latencyMs:r.latencyMs,actual:{...(r.usage?.inputTokens!=null?{inputTokens:r.usage.inputTokens}:{}),...(r.usage?.outputTokens!=null?{outputTokens:r.usage.outputTokens}:{}),...(r.usage?.actualCostUsd!=null?{costUsd:r.usage.actualCostUsd}:{})}}).catch(()=>undefined);
return{kind:"mso.jev-decision.v1",decisionId:id,decisionVersion:d.version,provider:"jev",score:a.score,probabilities,...(confidence!=null?{confidence}:{}),threshold:d.threshold,model:r.model,latencyMs:r.latencyMs,...(r.usage?.inputTokens!=null?{inputTokens:r.usage.inputTokens}:{}),...(r.usage?.actualCostUsd!=null?{actualCostUsd:r.usage.actualCostUsd}:{})};
}
diff --git a/scripts/e2e/mcp-sessions.mjs b/scripts/e2e/mcp-sessions.mjs
index 66cd193a..0924e613 100644
--- a/scripts/e2e/mcp-sessions.mjs
+++ b/scripts/e2e/mcp-sessions.mjs
@@ -69,8 +69,11 @@ export async function mcpSessionsJourney(page, fixture) {
const useJev = page.getByRole("button", { name: "Use JEV", exact: true });
if (!(await useJev.isVisible().catch(() => false))) {
const sessionDetails = page.getByRole("button", { name: "Session details", exact: true });
- await expect(sessionDetails).toBeVisible();
- await sessionDetails.click();
+ if (await sessionDetails.isVisible()) await sessionDetails.click();
+ else {
+ await page.getByRole("button", { name: "More actions", exact: true }).click();
+ await page.getByRole("menuitem", { name: "Session details", exact: true }).click();
+ }
}
await expect(useJev).toBeVisible();
await useJev.click();
@@ -80,6 +83,11 @@ export async function mcpSessionsJourney(page, fixture) {
await expect(page.locator('[data-slot="jev-capsule-summary"]')).toContainText("Continuation capsule");
await expect(page.getByText("MCP ACTION · not executed", { exact: true })).toBeVisible();
await expect(page.getByText("VERIFIED RESULT · n/a", { exact: true })).toBeVisible();
+ const sessionDialog = page.locator('[data-slot="shell-dialog-surface"]').filter({ has: page.locator('[data-slot="jev-session-output"]') });
+ if (await sessionDialog.isVisible()) {
+ await page.keyboard.press("Escape");
+ await expect(sessionDialog).toHaveCount(0);
+ }
const artifactStep = selectedGraph.steps.find(step => step.actions.some(action => action.artifact));
const artifactAction = artifactStep?.actions.find(action => action.artifact);
expect(artifactStep).toBeTruthy(); expect(artifactAction).toBeTruthy();
diff --git a/scripts/e2e/release.mjs b/scripts/e2e/release.mjs
index ddaa3abc..4c154ac2 100644
--- a/scripts/e2e/release.mjs
+++ b/scripts/e2e/release.mjs
@@ -96,7 +96,7 @@ try {
await expect(page.getByText("Actual JEV input / cost · 30d", { exact: true })).toBeVisible();
await expect(page.getByText("Estimated LLM input / cost avoided · 30d", { exact: true })).toBeVisible();
await page.getByText("JEV Lab · bounded decision test", { exact: true }).click();
- await expect(page.getByText("JEV Lab never executes tools, writes memory, or promotes recipes.", { exact: true })).toBeVisible();
+ await expect(page.getByText("JEV Lab never executes tools, writes memory, or promotes recipes.")).toBeVisible();
await page.getByRole("button", { name: "Run bounded test", exact: true }).click();
await expect(page.getByRole("status")).toContainText("JEV Lab decision complete; no action executed.");
await storeExtensionsJourney(page, fixture);