@opengeni/runtime 4.0.0 → 4.0.1-canary.35916963870001

package/dist/{chunk-5HJEVOMJ.js → chunk-ROCWOHGM.js}

@@ -26,7 +26,7 @@ import {
   withRunCredentialsClient,
   withRunCredentialsSession,
   wrapCapabilityToolsForTurnCancellation
-} from "./chunk-WIHICNWS.js";
+} from "./chunk-JFMOX5OE.js";
 import {
   MemorySession,
   connectMcpServers
@@ -10316,6 +10316,8 @@ Display images with ![descriptive alt text](artifact:<artifactId>). Use the exac
 
 Publish files you deliberately deliver so they are retained and discoverable in Artifacts; do not publish every temporary file. Reuse retained references for unchanged outputs. Source-code navigation may still use workspace file links. Inline HTML stays in chat unless explicitly saved as a Site.
 
+For published files, [Open file](artifact:<artifactId>) opens the retained file in Artifacts. ![Preview](artifact:<artifactId>) displays images, video, audio, or PDFs inline in the OpenGeni console, with an Artifact link for other formats. Replace <artifactId> with the exact artifact.artifactId from the publication receipt and use a descriptive label. Sites and native documents keep their tool-returned canonical links. Never substitute a storage URL or a sandbox path for a published artifact reference.
+
 # Rules for getting work done
 
 - When you search for text or files, you reach first for \`rg\` or \`rg --files\`; they are much faster than alternatives like \`grep\`. If \`rg\` is unavailable, you use the next best tool without fuss.
@@ -10353,6 +10355,8 @@ When presented with clarifying questions or objections from the user, lead with
 
 If completion requires new authority, external coordination, or a meaningful expansion beyond the user\u2019s implied intent and task scope (e.g. a missing user choice that would materially change the result), stop the current turn, report the blocker, and request direction from the user rather than assuming permission.
 
+Decide the design before building it. First check whether OpenGeni already provides the capability natively (for example, a Site reaches the model and workspace tools through the host bridge and needs no server of its own). If a native path fits, use it. Follow the project's established architecture and choices the user has already authorized or delegated; the user need not name each host, provider, or credential. Ask before making a new external commitment or materially departing from the established architecture beyond the authorized scope: name the candidate designs, and do not start parallel work that assumes the unresolved choice. The absence of a native path alone does not require a question; continue with routine implementation choices within scope.
+
 # Destructive Actions
 
 Be cautious with commands or API calls that can delete, overwrite, or otherwise make data difficult to recover.
@@ -10385,7 +10389,7 @@ Skills are reusable instructions supplied dynamically for the current session. W
 
 # Integration setup
 
-Use available integration tools directly. If the task needs an integration you cannot access, discover it with \`capability_catalog_search\` before declaring a setup blocker. For a suitable match with \`setup.nextAction\`, call \`capability_authorization_request\` with the returned capability ID and a brief task-specific rationale to show its Connect card in chat. Requesting the card does not need integration-management permission; the authenticated human must authorize setup. After setup, rediscover the tools and continue the task, verifying the access it needs. If access remains blocked, explain the specific blocker from the returned facts. If either setup tool is unavailable, report the missing setup path.
+Use available integration tools directly. If the task needs an integration you cannot access, discover it with \`capability_catalog_search\` before declaring a setup blocker. For a suitable match with \`setup.nextAction\`, call \`capability_authorization_request\` with the returned capability ID and a brief task-specific rationale to show its Connect card in chat. Requesting the card does not need integration-management permission; the authenticated human must authorize setup. A card is for an integration required by the authorized design, including established or delegated choices. Resolve out-of-scope architecture choices before requesting setup. After setup, rediscover the tools and continue the task, verifying the access it needs. If access remains blocked, explain the specific blocker from the returned facts. If either setup tool is unavailable, report the missing setup path.
 
 # Session coordination
 
@@ -12344,7 +12348,7 @@ import { existsSync, readdirSync } from "fs";
 // ../sdk/package.json
 var package_default = {
   name: "@opengeni/sdk",
-  version: "7.0.0",
+  version: "7.1.0-canary.35916963870001",
   description: "Framework-agnostic TypeScript SDK for the OpenGeni API: typed client, session lifecycle, SSE event streaming with reconnect + replay-by-sequence, and proxy re-streaming helpers.",
   license: "Apache-2.0",
   repository: {
@@ -12496,7 +12500,7 @@ var package_default = {
 // ../react/package.json
 var package_default2 = {
   name: "@opengeni/react",
-  version: "7.0.0",
+  version: "7.1.0-canary.35916963870001",
   description: "React hooks and styled components for OpenGeni: live session streaming, chat composer, message timeline, session status, and fleet views \u2014 token-themed (CSS variables), dark-first, built on Tailwind v4 + Radix + Motion.",
   license: "Apache-2.0",
   repository: {
@@ -12749,7 +12753,7 @@ var package_default2 = {
 // ../codemode/package.json
 var package_default3 = {
   name: "@opengeni/codemode",
-  version: "0.6.0",
+  version: "0.6.1-canary.35916963870001",
   description: "Attempt-frozen programmatic tool catalog and execution authority for OpenGeni.",
   license: "Apache-2.0",
   repository: {
@@ -12799,7 +12803,7 @@ var package_default3 = {
 // ../ogtool/package.json
 var package_default4 = {
   name: "@opengeni/ogtool",
-  version: "0.3.42",
+  version: "0.3.43-canary.35916963870001",
   description: "Codemode CLI and typed client entrypoint for an OpenGeni execution attempt.",
   license: "Apache-2.0",
   repository: {
@@ -15149,7 +15153,8 @@ var BrowserTabsInput = external_exports.discriminatedUnion("operation", [
 ]);
 var BrowserObserveInput = external_exports.object({
   browserSessionId: external_exports.string().uuid(),
-  targetId: external_exports.string().min(1).max(512)
+  targetId: external_exports.string().min(1).max(512),
+  includeScreenshot: external_exports.boolean().optional()
 }).strict();
 var BrowserActInput = external_exports.object({
   browserSessionId: external_exports.string().uuid(),
@@ -15289,7 +15294,7 @@ var ComputerActInput = external_exports.object({
   computerSessionId: external_exports.string().uuid(),
   targetId: external_exports.string().min(1).max(512),
   expectedTargetGeneration: external_exports.string().min(1).max(256).optional(),
-  expectedObservationId: external_exports.string().uuid().nullable().optional(),
+  expectedObservationId: ComputerObservation.shape.observationId.nullable().optional(),
   expectedFrameId: external_exports.string().min(1).max(256).nullable().optional(),
   action: ComputerAction
 }).strict();
@@ -15380,7 +15385,7 @@ function createInteractionAttemptToolDefinitions(input) {
     name: "browser_open",
     codemodePath: ["interaction", "browser", "open"],
     title: "Open or reuse browser",
-    description: "Open a managed BrowserSession on the current agent placement, reuse a relevant compatible live session by default, attach to an explicit workspace BrowserSession, or open an attached Chrome profile by passing placement={kind:'attached_device',deviceId}. This does not infer or attach the user's existing Chrome: for requests about 'my browser', 'my tabs', or current Chrome, call interaction_discover first and select an actual attachedBrowsers device; if none exists, explain that the Chrome extension must be connected instead of silently creating a blank managed browser. Managed Chromium defaults to headed so OAuth and later human interaction use a supported browser; request headless=true only for agent-only work that will not require sign-in or human control. Returns exact session and tab state.",
+    description: "Open a managed BrowserSession on the current agent placement, reuse a relevant compatible live session by default, attach to an explicit workspace BrowserSession, or open an attached Chrome profile by passing placement={kind:'attached_device',deviceId}. This does not infer or attach the user's existing Chrome: for requests about 'my browser', 'my tabs', or current Chrome, call interaction_discover first and select an actual attachedBrowsers device; if none exists, explain that the Chrome extension must be connected instead of silently creating a blank managed browser. BrowserSessions persist across tool calls; shell-launched browser daemons do not survive remote-command cleanup. Never switch to the user's attached Chrome as a fallback for a failed managed browser unless the user requested that profile. A new attached session opens a dedicated tab. Managed Chromium defaults to headed so OAuth and later human interaction use a supported browser; request headless=true only for agent-only work that will not require sign-in or human control. Returns exact session and tab state.",
     input: BrowserOpenInput,
     output: BrowserOpenOutput,
     readOnly: false,
@@ -15391,7 +15396,7 @@ function createInteractionAttemptToolDefinitions(input) {
     name: "browser_tabs",
     codemodePath: ["interaction", "browser", "tabs"],
     title: "Manage browser tabs",
-    description: "List, open, select, or close tabs in one exact BrowserSession. Returns the authoritative complete tab list after the operation.",
+    description: "List, open, logically select, or close tabs in one exact BrowserSession. Selection changes the BrowserSession's default target, not the visible desktop tab. New attached-Chrome tabs open in the background. Use browser_act activate only when foregrounding the owned tab is explicitly intended. Returns the authoritative complete tab list after the operation.",
     input: BrowserTabsInput,
     output: BrowserTargetListResponse,
     readOnly: false,
@@ -15421,22 +15426,47 @@ function createInteractionAttemptToolDefinitions(input) {
     name: "browser_observe",
     codemodePath: ["interaction", "browser", "observe"],
     title: "Observe browser tab",
-    description: "Read one tab's current URL/title, causal generations, compact semantic accessibility tree, dialog, and diagnostic counts without taking control.",
+    description: "Read one tab's current URL/title, measured page viewport, causal generations, compact semantic accessibility tree, dialog, and diagnostic counts without taking control. Set includeScreenshot=true to receive a current still image as tool image content; the structured screenshot field is only a retained-artifact reference and can remain null.",
     input: BrowserObserveInput,
     output: BrowserObservation,
     readOnly: true,
     idempotent: true,
-    execute: async (value) => await input.transport.observeBrowserTarget(
-      input.workspaceId,
-      value.browserSessionId,
-      value.targetId
-    )
+    execute: async (value) => {
+      let observation = await input.transport.observeBrowserTarget(
+        input.workspaceId,
+        value.browserSessionId,
+        value.targetId
+      );
+      if (!value.includeScreenshot) return observation;
+      const frame = await input.transport.captureBrowserTarget(
+        input.workspaceId,
+        value.browserSessionId,
+        value.targetId
+      );
+      if (observation.target.targetGeneration !== frame.targetGeneration || observation.target.documentGeneration !== frame.documentGeneration) {
+        observation = await input.transport.observeBrowserTarget(
+          input.workspaceId,
+          value.browserSessionId,
+          value.targetId
+        );
+      }
+      if (observation.target.targetGeneration !== frame.targetGeneration || observation.target.documentGeneration !== frame.documentGeneration) {
+        throw new Error("browser target changed while its visual observation was captured");
+      }
+      return new InteractionExecutionResult(observation, [
+        {
+          type: "image",
+          data: Buffer.from(frame.data).toString("base64"),
+          mimeType: frame.mediaType
+        }
+      ]);
+    }
   });
   add({
     name: "browser_act",
     codemodePath: ["interaction", "browser", "act"],
     title: "Act in browser tab",
-    description: "Perform one semantic-first browser action or bounded batch, including setting a managed browser's web permission for this tab's exact current top-level origin. Omit generation fences to use a fresh observation automatically; provide them to require exact previously observed state. Returns the durable receipt and changed observation.",
+    description: "Perform one semantic-first browser action or bounded batch. Use viewport to set page width, height, desktop/mobile layout and touch emulation, then check the measured viewport in the returned observation; emulation does not prove physical mobile-browser behavior. Use history back/forward for tab navigation; keypress shortcuts are page input and may not navigate browser history. The explicit activate action foregrounds the target in the user's desktop browser; use only when that is intended. Permission actions set a managed browser's web permission for this tab's exact current top-level origin. Omit generation fences to use a fresh observation automatically; provide them to require exact previously observed state. Returns the durable receipt and changed observation.",
     input: BrowserActInput,
     output: BrowserActionReceipt,
     readOnly: false,
@@ -15804,7 +15834,7 @@ function createInteractionAttemptToolDefinitions(input) {
     name: "computer_act",
     codemodePath: ["interaction", "computer", "act"],
     title: "Act in app or window",
-    description: "Perform one action in an exact ComputerSession target. Prefer semantic actions from computer_observe: on macOS they can invoke controls and set values without foregrounding the app. Pointer, keyboard, target focus, screen actions, and clipboard paste use the physical graphical seat and may change the user's foreground app; use them only when foreground control is explicitly intended. Omit fences to use a fresh observation automatically. Returns the durable causal receipt.",
+    description: "Perform one action in an exact ComputerSession target. Prefer semantic actions from computer_observe: on macOS they can invoke controls and set values without foregrounding the app. Pointer, keyboard, target focus, screen actions, and clipboard paste use the physical graphical seat and may change the user's foreground app; use them only when foreground control is explicitly intended. For pointer actions, set action.frameId from the observed frame and use coordinates in the returned screenshot's pixels; do not scale them to target.bounds. Omit fences to use a fresh observation automatically. Returns the durable causal receipt.",
     input: ComputerActInput,
     output: ComputerActionReceipt,
     readOnly: false,
@@ -15897,10 +15927,12 @@ async function openBrowser(transport, workspaceId, sourceSessionId, value, conte
   if (value.browserSessionId) {
     session = await transport.getBrowserSession(workspaceId, value.browserSessionId);
   } else {
-    const listed = await transport.listBrowserSessions(workspaceId);
+    const listed = value.mode === "new" ? { sessions: [] } : await transport.listBrowserSessions(workspaceId);
     const requestedHeadless = value.headless ?? false;
     const reusable = value.mode === "new" ? null : newestRelevant(
-      listed.sessions.filter((candidate) => candidate.headless === requestedHeadless),
+      listed.sessions.filter(
+        (candidate) => candidate.headless === requestedHeadless && compatibleInteractionPlacement(candidate.placement, value.placement) && (value.identityId === void 0 || candidate.identityId === value.identityId) && (value.baseRevisionId === void 0 || candidate.baseRevisionId === value.baseRevisionId) && (value.networkRouteId === void 0 || candidate.networkRouteId === value.networkRouteId) && (value.linkedComputerSessionId === void 0 || candidate.linkedComputerSessionId === value.linkedComputerSessionId)
+      ),
       sourceSessionId
     );
     if (reusable) {
@@ -15946,8 +15978,13 @@ async function openComputer(transport, workspaceId, sourceSessionId, value, cont
   if (value.computerSessionId) {
     session = await transport.getComputerSession(workspaceId, value.computerSessionId);
   } else {
-    const listed = await transport.listComputerSessions(workspaceId);
-    const reusable = value.mode === "new" ? null : newestRelevant(listed.sessions, sourceSessionId);
+    const listed = value.mode === "new" ? { sessions: [] } : await transport.listComputerSessions(workspaceId);
+    const reusable = newestRelevant(
+      listed.sessions.filter(
+        (candidate) => compatibleInteractionPlacement(candidate.placement, value.placement)
+      ),
+      sourceSessionId
+    );
     session = reusable ? reusable : (await transport.createComputerSession(workspaceId, {
       operationId: context.operationId,
       sessionId: sourceSessionId,
@@ -15960,6 +15997,12 @@ async function openComputer(transport, workspaceId, sourceSessionId, value, cont
     targets: session.lifecycle === "active" ? (await transport.listComputerTargets(workspaceId, session.id)).targets : []
   };
 }
+function compatibleInteractionPlacement(candidate, requested) {
+  if (!requested) return candidate.kind !== "attached_device";
+  return Object.entries(requested).every(
+    ([key, value]) => candidate[key] === value
+  );
+}
 function newestRelevant(sessions, sourceSessionId) {
   return sessions.filter(
     (session) => !["ending", "ended", "failed", "lost", "repair_required"].includes(session.lifecycle) && session.associations.some((association) => association.sessionId === sourceSessionId)
@@ -18430,6 +18473,7 @@ var DeferredPreparedMcpServer = class {
   deferredPreparation = true;
   name;
   listedModelNames = /* @__PURE__ */ new Set();
+  preparedTarget = null;
   toolMetaResolver = async (context) => {
     const target = await this.resolveTarget();
     return await target?.toolMetaResolver?.(context);
@@ -18446,6 +18490,7 @@ var DeferredPreparedMcpServer = class {
     if (!this.isPrepared()) return [];
     const target = await this.resolveTarget();
     const tools = target ? await target.listTools() : [];
+    this.preparedTarget = target instanceof PrefixedMcpServer ? target : null;
     this.listedModelNames.clear();
     for (const tool2 of tools) this.listedModelNames.add(tool2.name);
     return tools;
@@ -18453,6 +18498,9 @@ var DeferredPreparedMcpServer = class {
   modelToolNames() {
     return this.listedModelNames.values();
   }
+  toolDisplayMetadata(name) {
+    return this.preparedTarget?.toolDisplayMetadata(name);
+  }
   async unprefixedToolName(name) {
     const target = await this.requiredTarget();
     if (!(target instanceof PrefixedMcpServer)) throw new Error("Unknown prepared MCP identity");
@@ -20201,7 +20249,31 @@ function normalizeUrl(raw) {
 }
 function prefixedMcpToolName(registryId, toolName) {
   const name = sharedPrefixedMcpToolName(registryId, toolName);
-  return name.length <= 64 && /^[A-Za-z0-9_]+$/.test(name) ? name : createHash5("sha256").update(JSON.stringify([registryId, toolName])).digest("hex");
+  if (name.length <= 64 && /^[A-Za-z0-9_]+$/.test(name)) return name;
+  const digest = legacyMcpToolDigest(registryId, toolName).slice(0, 24);
+  const leaf = toolName.replace(/[^A-Za-z0-9_]/g, "_").slice(0, 34) || "tool";
+  return `mcp_${digest}__${leaf}`;
+}
+function legacyMcpToolDigest(registryId, toolName) {
+  return createHash5("sha256").update(JSON.stringify([registryId, toolName])).digest("hex");
+}
+function mcpToolDisplayMetadata(servers, name) {
+  for (const server of servers) {
+    if (server instanceof PrefixedMcpServer || server instanceof DeferredPreparedMcpServer) {
+      const display = server.toolDisplayMetadata(name);
+      if (display) return display;
+    }
+  }
+  return void 0;
+}
+function withMcpToolDisplayMetadata(servers, payload) {
+  if (!payload || typeof payload !== "object" || Array.isArray(payload)) return payload;
+  const item = payload;
+  const raw = item.rawItem && typeof item.rawItem === "object" ? item.rawItem : {};
+  const name = item.name ?? item.toolName ?? raw.name;
+  if (typeof name !== "string") return payload;
+  const display = mcpToolDisplayMetadata(servers, name);
+  return display ? { ...item, display } : payload;
 }
 var MCP_SDK_LIFECYCLE_NAME = "opengeni-mcp-lifecycle";
 function publicMcpLifecycleError(error, phase, serverId, options = {}) {
@@ -20364,6 +20436,10 @@ var PrefixedMcpServer = class _PrefixedMcpServer {
   listedToolSchemaTokens = 0;
   frozenTools = null;
   originalToolNames = /* @__PURE__ */ new Map();
+  displayMetadata = /* @__PURE__ */ new Map();
+  toolDisplayMetadata(name) {
+    return this.displayMetadata.get(name);
+  }
   /** Exact attempt-local classification, including names too long for a prefix. */
   modelToolNames() {
     return this.originalToolNames.keys();
@@ -20509,6 +20585,13 @@ var PrefixedMcpServer = class _PrefixedMcpServer {
         if (prior !== void 0 && prior !== tool2.name)
           throw new Error("MCP model tool identity collision");
         this.originalToolNames.set(name, tool2.name);
+        const display = {
+          toolName: tool2.name,
+          ...tool2.title ? { title: tool2.title } : {},
+          ...this.accountLabel ? { accountLabel: this.accountLabel } : {}
+        };
+        this.displayMetadata.set(name, display);
+        this.displayMetadata.set(legacyMcpToolDigest(this.registryId, tool2.name), display);
         return {
           ...tool2,
           name,
@@ -20748,9 +20831,7 @@ async function restoreInterruptedRunState(agent, serializedRunState) {
   }
   return await RunState.fromString(agent, serializedRunState, {
     clientToolSearchRehydration: "preserve_history",
-    ...lazyRuntime ? {
-      resolveMissingFunctionTool: createResolveMissingFunctionTool(lazyRuntime)
-    } : {}
+    resolveMissingFunctionTool: mcpToolResolverForAgent(agent)
   });
 }
 var TURN_OPERATIONAL_NOTICE_PREFIX = "[OpenGeni turn-scoped operational notice]\n";
@@ -21276,10 +21357,35 @@ function appendSandboxFileDownloadFailureNote(input, failures) {
 }
 function lazyToolRunBindings(agent) {
   const runtime = lazyToolRuntimeForAgent(agent);
-  if (!runtime) return {};
   return {
-    toolNotFoundBehavior: "return_error_to_model",
-    resolveMissingFunctionTool: createResolveMissingFunctionTool(runtime)
+    ...runtime ? { toolNotFoundBehavior: "return_error_to_model" } : {},
+    resolveMissingFunctionTool: mcpToolResolverForAgent(agent)
+  };
+}
+function mcpToolResolverForAgent(agent) {
+  const lazyRuntime = lazyToolRuntimeForAgent(agent);
+  const resolveCurrent = lazyRuntime ? createResolveMissingFunctionTool(lazyRuntime) : null;
+  return async (input) => {
+    const current = await resolveCurrent?.(input);
+    if (current) return current;
+    if (!/^[a-f0-9]{64}$/.test(input.name)) return null;
+    const tools = await agent.getMcpTools(new RunContext());
+    let resolvedName;
+    for (const server of agent.mcpServers) {
+      if (!(server instanceof PrefixedMcpServer || server instanceof DeferredPreparedMcpServer))
+        continue;
+      for (const descriptor of await server.listTools()) {
+        const rawName = await server.unprefixedToolName(descriptor.name);
+        if (legacyMcpToolDigest(server.registryId, rawName) !== input.name) continue;
+        if (resolvedName !== void 0 && resolvedName !== descriptor.name)
+          throw new Error("MCP legacy model tool identity collision");
+        resolvedName = descriptor.name;
+      }
+    }
+    const tool2 = tools.find(
+      (candidate) => candidate.type === "function" && candidate.name === resolvedName
+    );
+    return tool2?.type === "function" ? { ...tool2, name: input.name } : null;
   };
 }
 function runScopedRunner(settings, agent, inputWaitYield) {
@@ -23828,6 +23934,8 @@ export {
   mcpTransportErrorWithRetryMetadata,
   safeMcpTransportError,
   prefixedMcpToolName,
+  mcpToolDisplayMetadata,
+  withMcpToolDisplayMetadata,
   PrefixedMcpServer,
   restoreInterruptedRunState,
   TURN_OPERATIONAL_NOTICE_PREFIX,
@@ -23872,4 +23980,4 @@ export {
   runAzureCliLoginHook,
   renderSessionGoalContext2 as renderSessionGoalContext
 };
-//# sourceMappingURL=chunk-5HJEVOMJ.js.map
+//# sourceMappingURL=chunk-ROCWOHGM.js.map