@opengeni/contracts 0.44.1 → 1.0.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@opengeni/contracts",
-  "version": "0.44.1",
+  "version": "1.0.0",
   "description": "Shared zod schemas and wire-contract types for the OpenGeni API.",
   "license": "Apache-2.0",
   "repository": {
@@ -18,6 +18,74 @@
   "module": "./dist/index.js",
   "types": "./dist/index.d.ts",
   "exports": {
+    "./connection-account-label": {
+      "types": "./dist/connection-account-label.d.ts",
+      "import": "./dist/connection-account-label.js"
+    },
+    "./modal-native-proof-v2": {
+      "types": "./dist/modal-native-proof-v2.d.ts",
+      "import": "./dist/modal-native-proof-v2.js"
+    },
+    "./allowance-refusal": {
+      "types": "./dist/allowance-refusal.d.ts",
+      "import": "./dist/allowance-refusal.js"
+    },
+    "./usage-allowances": {
+      "types": "./dist/usage-allowances.d.ts",
+      "import": "./dist/usage-allowances.js"
+    },
+    "./browser-storage": {
+      "types": "./dist/browser-storage.d.ts",
+      "import": "./dist/browser-storage.js"
+    },
+    "./managed-sign-in-methods": {
+      "types": "./dist/managed-sign-in-methods.d.ts",
+      "import": "./dist/managed-sign-in-methods.js"
+    },
+    "./client-error-report": {
+      "types": "./dist/client-error-report.d.ts",
+      "import": "./dist/client-error-report.js"
+    },
+    "./analytics-consent-report": {
+      "types": "./dist/analytics-consent-report.d.ts",
+      "import": "./dist/analytics-consent-report.js"
+    },
+    "./mcp-endpoint": {
+      "types": "./dist/mcp-endpoint.d.ts",
+      "import": "./dist/mcp-endpoint.js"
+    },
+    "./plugin-discovery": {
+      "types": "./dist/plugin-discovery.d.ts",
+      "import": "./dist/plugin-discovery.js"
+    },
+    "./host-mcp-bindings": {
+      "types": "./dist/host-mcp-bindings.d.ts",
+      "import": "./dist/host-mcp-bindings.js"
+    },
+    "./external-identities": {
+      "types": "./dist/external-identities.d.ts",
+      "import": "./dist/external-identities.js"
+    },
+    "./organization-access": {
+      "types": "./dist/organization-access.d.ts",
+      "import": "./dist/organization-access.js"
+    },
+    "./connect": {
+      "types": "./dist/connect.d.ts",
+      "import": "./dist/connect.js"
+    },
+    "./site-session-http": {
+      "types": "./dist/site-session-http.d.ts",
+      "import": "./dist/site-session-http.js"
+    },
+    "./organization-model-usage": {
+      "types": "./dist/organization-model-usage.d.ts",
+      "import": "./dist/organization-model-usage.js"
+    },
+    "./insights-usage": {
+      "types": "./dist/insights-usage.d.ts",
+      "import": "./dist/insights-usage.js"
+    },
     ".": {
       "types": "./dist/index.d.ts",
       "import": "./dist/index.js"
@@ -26,22 +94,94 @@
       "types": "./dist/slack-bot-scopes.d.ts",
       "import": "./dist/slack-bot-scopes.js"
     },
+    "./slack-rest-mcp": {
+      "types": "./dist/slack-rest-mcp.d.ts",
+      "import": "./dist/slack-rest-mcp.js"
+    },
     "./google-drive": {
       "types": "./dist/google-drive.d.ts",
       "import": "./dist/google-drive.js"
     },
+    "./personal-github": {
+      "types": "./dist/personal-github.d.ts",
+      "import": "./dist/personal-github.js"
+    },
+    "./github-repository": {
+      "types": "./dist/github-repository.d.ts",
+      "import": "./dist/github-repository.js"
+    },
+    "./github-repository-contracts": {
+      "types": "./dist/github-repository-contracts.d.ts",
+      "import": "./dist/github-repository-contracts.js"
+    },
     "./atlassian": {
       "types": "./dist/atlassian.d.ts",
       "import": "./dist/atlassian.js"
     },
+    "./atlassian-native-retirement": {
+      "types": "./dist/atlassian-native-retirement.d.ts",
+      "import": "./dist/atlassian-native-retirement.js"
+    },
     "./connector-destinations": {
       "types": "./dist/connector-destinations.d.ts",
       "import": "./dist/connector-destinations.js"
     },
+    "./connector-attachments": {
+      "types": "./dist/connector-attachments.d.ts",
+      "import": "./dist/connector-attachments.js"
+    },
+    "./connection-authority": {
+      "types": "./dist/connection-authority.d.ts",
+      "import": "./dist/connection-authority.js"
+    },
     "./video-generation": {
       "types": "./dist/video-generation.d.ts",
       "import": "./dist/video-generation.js"
     },
+    "./canonical-human-identities": {
+      "types": "./dist/canonical-human-identities.d.ts",
+      "import": "./dist/canonical-human-identities.js"
+    },
+    "./session-titles": {
+      "types": "./dist/session-titles.d.ts",
+      "import": "./dist/session-titles.js"
+    },
+    "./session-final-reply": {
+      "types": "./dist/session-final-reply.d.ts",
+      "import": "./dist/session-final-reply.js"
+    },
+    "./managed-auth-session-sets": {
+      "types": "./dist/managed-auth-session-sets.d.ts",
+      "import": "./dist/managed-auth-session-sets.js"
+    },
+    "./organization-recovery": {
+      "types": "./dist/organization-recovery.d.ts",
+      "import": "./dist/organization-recovery.js"
+    },
+    "./codex-provider-account-authority": {
+      "types": "./dist/codex-provider-account-authority.d.ts",
+      "import": "./dist/codex-provider-account-authority.js"
+    },
+    "./xai-provider-account-authority": {
+      "types": "./dist/xai-provider-account-authority.d.ts",
+      "import": "./dist/xai-provider-account-authority.js"
+    },
+    "./model-picker-order": {
+      "types": "./dist/model-picker-order.d.ts",
+      "import": "./dist/model-picker-order.js"
+    },
+    "./model-display": {
+      "types": "./dist/model-display.d.ts",
+      "import": "./dist/model-display.js"
+    },
+    "./workspace-integration-wire": {
+      "types": "./dist/workspace-integration-wire.d.ts",
+      "import": "./dist/workspace-integration-wire.js"
+    },
+    "./code-search": {
+      "types": "./dist/code-search.d.ts",
+      "import": "./dist/code-search.js"
+    },
     "./editable-artifacts": {
       "types": "./dist/editable-artifacts.d.ts",
       "import": "./dist/editable-artifacts.js"
@@ -93,6 +233,10 @@
     "./presentation-artifact-query": {
       "types": "./dist/presentation-artifact-query.d.ts",
       "import": "./dist/presentation-artifact-query.js"
+    },
+    "./session-list-entries": {
+      "types": "./dist/session-list-entries.d.ts",
+      "import": "./dist/session-list-entries.js"
     }
   },
   "publishConfig": {
@@ -106,6 +250,7 @@
   },
   "dependencies": {
     "@noble/hashes": "^1.8.0",
+    "yaml": "2.9.0",
     "zod": "^4.2.1"
   },
   "engines": {

package/src/agent-authored-durable-text.ts

@@ -0,0 +1,116 @@
+import { WORKSPACE_INSTRUCTION_POLICY_CONTENT_MAX_CHARS } from "./workspace-instruction-policies";
+
+/**
+ * Budgets and authoring style for durable text an **agent** writes on a user's
+ * behalf.
+ *
+ * The budget follows the cost, and the cost differs per destination:
+ *
+ * - A workspace instruction policy is composed verbatim into the prompt of
+ *   every session it applies to, for as long as it stays active. A global
+ *   charter or global policy applies to every session in the workspace; a role
+ *   policy applies to every session bound to that role. At most three entries
+ *   compose at once (charter, global policy, matching role policy), so the
+ *   storage limit is shared with the human editor; prompt composition retains
+ *   its separate UTF-8 byte bound.
+ * - A preference is cheaper and in a different way: only its short title and
+ *   description descriptors are prompt-composed, while the full content is
+ *   retrieved on demand behind the exact attempt's retrieval handle. Its length
+ *   is therefore retrieval cost, not standing prompt cost.
+ * - Organization identity and mission are always-on context for root sessions
+ *   across the whole organization. Historical company-profile list fields stay
+ *   in the storage contract for compatibility; nonempty values on an already-
+ *   active historical revision remain composed with an explicit compatibility
+ *   label until an organization owner replaces the profile.
+ * - Knowledge is retrieval evidence and never joins the always-composed prefix.
+ *
+ * Instruction writes share the human limit so localized edits can preserve
+ * existing long policies. The other caps bind only agent-authored writes.
+ * The human-facing limits (`PREFERENCE_REGISTRY_CONTENT_MAX_CHARS`, `COMPANY_PROFILE_SCALAR_MAX_CHARS`,
+ * `COMPANY_PROFILE_ENTRY_MAX_CHARS`, `COMPANY_PROFILE_CONTENT_MAX_UTF8_BYTES`)
+ * are unchanged: a person editing in the UI is making a deliberate, visible
+ * choice, and lowering their limit would reject text they already typed.
+ * Existing stored revisions are never rewritten; only new agent writes are
+ * bounded.
+ */
+
+/** Compatibility export: agents and humans use the same instruction storage limit. */
+export const AGENT_AUTHORED_INSTRUCTION_POLICY_CONTENT_MAX_CHARS =
+  WORKSPACE_INSTRUCTION_POLICY_CONTENT_MAX_CHARS;
+
+/** Retrieved on demand rather than always composed. */
+export const AGENT_AUTHORED_PREFERENCE_CONTENT_MAX_CHARS = 1_200;
+
+/** `identity` and `mission`: a couple of sentences, not a positioning document. */
+export const AGENT_AUTHORED_COMPANY_PROFILE_SCALAR_MAX_CHARS = 400;
+
+/** Legacy company-profile list entry bound; retained for stored-revision compatibility. */
+export const AGENT_AUTHORED_COMPANY_PROFILE_ENTRY_MAX_CHARS = 200;
+
+/** Whole canonical compatibility profile; current agent tools submit only identity and mission. */
+export const AGENT_AUTHORED_COMPANY_PROFILE_CONTENT_MAX_UTF8_BYTES = 4_096;
+
+export const AGENT_AUTHORED_INSTRUCTION_POLICY_CONTENT_TOO_LONG_MESSAGE =
+  "A workspace rule is composed verbatim into the prompt of every session it applies to (every session " +
+  "for a global charter or policy, every session bound to the role for a role policy), for as long as it " +
+  `stays active. The instruction must contain at most ${AGENT_AUTHORED_INSTRUCTION_POLICY_CONTENT_MAX_CHARS} ` +
+  "characters, the same limit as the human editor. Keep new rules concise and preserve unrelated content.";
+
+export const AGENT_AUTHORED_PREFERENCE_CONTENT_TOO_LONG_MESSAGE =
+  "A Skill is durable Agent Knowledge that agents retrieve on demand, so its length is " +
+  "retrieval cost rather than standing prompt cost: only its short title and description are composed into " +
+  `every session prompt. Keep it under ${AGENT_AUTHORED_PREFERENCE_CONTENT_MAX_CHARS} characters: state the ` +
+  "trigger and outcome plainly, include only the necessary procedure, checks, and exceptions, and remove " +
+  "background, repeated guidance, and decorative examples.";
+
+export const AGENT_AUTHORED_COMPANY_PROFILE_TOO_LONG_MESSAGE =
+  "Organization identity and mission are mandatory prompt context in every root session across the " +
+  `organization. Keep each under ${AGENT_AUTHORED_COMPANY_PROFILE_SCALAR_MAX_CHARS} characters: one concise ` +
+  "descriptive statement, no products, customers, goals, constraints, procedure, or marketing copy. Those " +
+  "details belong in organization Documents and are retrieved as evidence when relevant.";
+
+/**
+ * Lane-agnostic guidance for the one `remember` content field, whose schema
+ * bound has to admit the widest lane. Without it an over-long rule would be
+ * refused with a generic "expected length <= 4000" that points at the wrong
+ * number entirely.
+ */
+export const AGENT_AUTHORED_REMEMBER_CONTENT_TOO_LONG_MESSAGE =
+  "Remembered content is bounded by where it lands: a mandatory workspace rule at most " +
+  `${AGENT_AUTHORED_INSTRUCTION_POLICY_CONTENT_MAX_CHARS} characters because it is composed into every ` +
+  `session prompt it applies to, a preference at most ${AGENT_AUTHORED_PREFERENCE_CONTENT_MAX_CHARS}, a ` +
+  "Knowledge fact at most 4000. Write one imperative statement in 1-3 sentences, split unrelated entries, and " +
+  "keep procedure in a Document or Skill that the entry references.";
+
+/**
+ * The shape rule the model reads on every tool that authors a durable *rule or
+ * preference*. Deliberately not reused for the company profile, which is a
+ * descriptive profile rather than an instruction and needs its own wording.
+ */
+export const AGENT_AUTHORED_INSTRUCTION_POLICY_STYLE =
+  "Write the shortest complete imperative rule in 1-3 sentences, with no numbered procedure, examples, " +
+  "rationale, or restated defaults. Split unrelated rules, and move conditional procedure into a Skill.";
+
+export const AGENT_AUTHORED_SKILL_STYLE =
+  "Write one focused Skill with a clear trigger and outcome. Include only the prerequisites, necessary steps, " +
+  "verification, and important failure handling an agent needs to execute it; omit background, repetition, " +
+  "generic advice, and decorative examples. Split unrelated workflows into separate Skills.";
+
+/** @deprecated Use the destination-specific instruction or Skill style. */
+export const AGENT_AUTHORED_DURABLE_TEXT_STYLE = AGENT_AUTHORED_INSTRUCTION_POLICY_STYLE;
+
+/**
+ * Actionable over-budget message including the text the caller actually sent,
+ * so the model can see how far over it is instead of guessing.
+ */
+export function agentAuthoredDurableTextTooLongMessage(input: {
+  kind: "instruction_policy" | "preference";
+  actualChars: number;
+}): string {
+  const subject = input.kind === "instruction_policy" ? "This rule is" : "This preference is";
+  const limit =
+    input.kind === "instruction_policy"
+      ? AGENT_AUTHORED_INSTRUCTION_POLICY_CONTENT_TOO_LONG_MESSAGE
+      : AGENT_AUTHORED_PREFERENCE_CONTENT_TOO_LONG_MESSAGE;
+  return `${subject} ${input.actualChars} characters. ${limit}`;
+}