@lightsage/mcp 0.2.0 → 0.2.2

package/src/mcp-server/server.ts

@@ -91,7 +91,7 @@ export function createMCPServer(deps: {
       sizes: ["any"],
     }],
     name: "LightsageMcp",
-    version: "0.2.0",
+    version: "0.2.2",
   });
 
   const getClient = deps.getSDK || (() =>

package/src/mcp-server/tools/evalsMonitor.ts

@@ -15,7 +15,7 @@ export const tool$evalsMonitor: ToolDefinition<typeof args> = {
   name: "evals-monitor",
   description: `Monitor Runs
 
-Streams an immediate snapshot followed by live status and sanitized activity for one public v2 eval run or the active runs captured when the connection opens. Exactly one of \`run_id\` or \`all=true\` is required.`,
+Streams an immediate snapshot followed by live status and sanitized activity. Provide \`run_id\` to monitor one public v2 eval run, or omit it to capture all active runs when the connection opens.`,
   annotations: {
     "title": "",
     "destructiveHint": false,

package/src/mcp-server/tools/evalsRun.ts

@@ -13,9 +13,16 @@ const args = {
 
 export const tool$evalsRun: ToolDefinition<typeof args> = {
   name: "evals-run",
-  description: `Create Run
+  description: `Start an eval run
 
-Queues a durable eval job and returns immediately with a polling ID. Poll \`GET /v2/eval-runs/{run_id}\` until its status reaches a terminal value. Choose exactly one request mode. For saved mode, provide \`eval_id\`, \`configuration_id\`, and \`agent_id\`. For direct mode, omit \`eval_id\` and provide \`prompt\`, \`judges\`, and \`agent_id\`. Direct runs may also provide a repository, write-only environment variables, and public skill packages.`,
+Queues a durable eval job and returns immediately with a polling ID. Poll \`GET /v2/eval-runs/{run_id}\` until its status reaches a terminal value.
+
+Send one of these flat request forms:
+
+- **Saved eval:** provide \`eval_id\`, \`configuration_id\`, and \`agent_id\`.
+- **Direct create-and-run:** provide \`prompt\`, \`judges\`, and \`agent_id\`. You may also provide a repository, write-only environment variables, and public skill packages.
+
+Do not combine saved-run and direct-run fields. No separate \`mode\` field is required.`,
   annotations: {
     "title": "",
     "destructiveHint": false,

package/src/mcp-server/tools/repositoriesV2RepositoriesList.ts

@@ -10,7 +10,7 @@ export const tool$repositoriesV2RepositoriesList: ToolDefinition = {
   name: "repositories-v2-repositories-list",
   description: `List Repositories
 
-Returns Git repositories available as reusable workspaces for eval configurations.`,
+Returns built-in templates and enabled saved Git repositories. Pass an item's \`id\` directly as \`repository\` in a direct eval run. Public GitHub URLs are also accepted by eval runs but are not included in this list.`,
   annotations: {
     "title": "",
     "destructiveHint": false,

package/src/models/createv2evalrunrequest.ts

@@ -5,6 +5,9 @@
 
 import * as z from "zod";
 
+/**
+ * Start an eval run using one of two flat request forms. Saved runs require `eval_id`, `configuration_id`, and `agent_id`. Direct create-and-run requests require `prompt`, at least one item in `judges`, and `agent_id`. Do not combine saved-run and direct-run fields; no separate mode field is required.
+ */
 export type CreateV2EvalRunRequest = {
   eval_id?: string | undefined;
   configuration_id?: string | undefined;
@@ -19,26 +22,30 @@ export type CreateV2EvalRunRequest = {
 export const CreateV2EvalRunRequest$zodSchema: z.ZodType<
   CreateV2EvalRunRequest
 > = z.object({
-  agent_id: z.string().describe("Runnable agent ID from `GET /v2/agents`."),
+  agent_id: z.string().describe(
+    "Required in both request forms. Runnable agent ID from `GET /v2/agents`.",
+  ),
   configuration_id: z.string().optional().describe(
-    "Saved runtime configuration. Required when `eval_id` is provided.",
+    "Saved-run form only. Saved runtime configuration; required when `eval_id` is provided.",
   ),
   env: z.record(z.string(), z.string()).optional().describe(
-    "Write-only environment variables for a direct run. Values are stored only in the queued job's private configuration.",
+    "Direct create-and-run form only. Write-only environment variables; values are stored only in the queued job's private configuration.",
   ),
   eval_id: z.string().optional().describe(
-    "Existing eval definition to run. Omit to create and run an eval directly.",
+    "Saved-run form only. Existing eval definition to run. Omit this field to use the direct create-and-run form.",
   ),
   judges: z.array(z.string()).optional().describe(
-    "Success criteria for a direct create-and-run request. CLI clients may expose this string array as a repeatable `--judge` flag.",
+    "Direct create-and-run form only. Provide at least one success criterion. CLI clients may expose this string array as a repeatable `--judge` flag.",
   ),
   prompt: z.string().optional().describe(
-    "Task prompt for a direct create-and-run request.",
+    "Direct create-and-run form only. Task prompt; required when `eval_id` is omitted.",
   ),
   repository: z.string().optional().describe(
-    "Repository for a direct run: a saved repository ID, built-in alias such as `nextjs-starter`, or a public GitHub HTTPS URL.",
+    "Direct create-and-run form only. A saved repository ID, built-in alias such as `nextjs-starter`, or a public GitHub HTTPS URL.",
   ),
   skills: z.array(z.string()).optional().describe(
-    "Public skill packages to install before the agent starts, using GitHub `owner/repository` references such as `resend/resend-skills`.",
+    "Additional public skill packages to install before the agent starts in either request form, using GitHub `owner/repository` references such as `resend/resend-skills`. CLI clients may expose this string array as a repeatable `--skill` flag.",
   ),
-});
+}).describe(
+  "Start an eval run using one of two flat request forms. Saved runs require `eval_id`, `configuration_id`, and `agent_id`. Direct create-and-run requests require `prompt`, at least one item in `judges`, and `agent_id`. Do not combine saved-run and direct-run fields; no separate mode field is required.",
+);

package/src/models/v2evalrunsmonitorop.ts

@@ -29,7 +29,6 @@ export const Detail$zodSchema = z.enum([
 
 export type V2EvalRunsMonitorRequest = {
   run_id?: string | undefined;
-  all?: boolean | undefined;
   detail?: Detail | undefined;
   after_id?: number | undefined;
 };
@@ -40,14 +39,12 @@ export const V2EvalRunsMonitorRequest$zodSchema: z.ZodType<
   after_id: z.int().default(0).describe(
     "Durable child-event cursor used when reconnecting.",
   ),
-  all: z.boolean().default(false).describe(
-    "Capture all currently active public v2 eval runs.",
-  ),
   detail: Detail$zodSchema.default("summary").describe(
     "Reduced activity or the full sanitized durable trace.",
   ),
-  run_id: z.string().describe("Parent run ID returned by `POST /v2/eval-runs`.")
-    .optional(),
+  run_id: z.string().describe(
+    "Parent run ID returned by `POST /v2/eval-runs`. Omit to capture all currently active public v2 eval runs.",
+  ).optional(),
 });
 
 export type V2EvalRunsMonitorResponse = string | HTTPValidationError;

package/src/models/v2repository.ts

@@ -6,6 +6,25 @@
 import * as z from "zod";
 import { ClosedEnum } from "../types/enums.js";
 
+/**
+ * Whether the repository is a read-only built-in template or an organization-owned saved repository.
+ */
+export const Kind = {
+  Template: "template",
+  Saved: "saved",
+} as const;
+/**
+ * Whether the repository is a read-only built-in template or an organization-owned saved repository.
+ */
+export type Kind = ClosedEnum<typeof Kind>;
+
+export const Kind$zodSchema = z.enum([
+  "template",
+  "saved",
+]).describe(
+  "Whether the repository is a read-only built-in template or an organization-owned saved repository.",
+);
+
 export const V2RepositoryFramework = {
   Fastify: "fastify",
   Nuxt: "nuxt",
@@ -35,6 +54,7 @@ export const V2RepositoryFramework$zodSchema = z.enum([
 
 export type V2Repository = {
   id: string;
+  kind: Kind;
   name: string;
   repo_url: string;
   ref: string;
@@ -50,8 +70,13 @@ export const V2Repository$zodSchema: z.ZodType<V2Repository> = z.object({
   created_at: z.string().nullable().optional(),
   enabled: z.boolean(),
   framework: V2RepositoryFramework$zodSchema.default("custom"),
-  id: z.string(),
+  id: z.string().describe(
+    "Value to pass as `repository` in a direct eval run. Built-in templates use aliases such as `nextjs-starter`; saved repositories use their persisted IDs.",
+  ),
   install_steps: z.array(z.string()).optional(),
+  kind: Kind$zodSchema.describe(
+    "Whether the repository is a read-only built-in template or an organization-owned saved repository.",
+  ),
   name: z.string(),
   ref: z.string(),
   repo_url: z.string(),

package/src/tool-names.ts

@@ -150,11 +150,11 @@ export const toolNames: Array<{ name: string; description: string }>= [
   },
   {
     "name": "evals-run",
-    "description": "Create Run\n\nQueues a durable eval job and returns immediately with a polling ID. Poll `GET /v2/eval-runs/{run_id}` until its status reaches a terminal value. Choose exactly one request mode. For saved mode, provide `eval_id`, `configuration_id`, and `agent_id`. For direct mode, omit `eval_id` and provide `prompt`, `judges`, and `agent_id`. Direct runs may also provide a repository, write-only environment variables, and public skill packages."
+    "description": "Start an eval run\n\nQueues a durable eval job and returns immediately with a polling ID. Poll `GET /v2/eval-runs/{run_id}` until its status reaches a terminal value.\n\nSend one of these flat request forms:\n\n- **Saved eval:** provide `eval_id`, `configuration_id`, and `agent_id`.\n- **Direct create-and-run:** provide `prompt`, `judges`, and `agent_id`. You may also provide a repository, write-only environment variables, and public skill packages.\n\nDo not combine saved-run and direct-run fields. No separate `mode` field is required."
   },
   {
     "name": "evals-monitor",
-    "description": "Monitor Runs\n\nStreams an immediate snapshot followed by live status and sanitized activity for one public v2 eval run or the active runs captured when the connection opens. Exactly one of `run_id` or `all=true` is required."
+    "description": "Monitor Runs\n\nStreams an immediate snapshot followed by live status and sanitized activity. Provide `run_id` to monitor one public v2 eval run, or omit it to capture all active runs when the connection opens."
   },
   {
     "name": "eval-runs-v2-eval-runs-list",
@@ -178,7 +178,7 @@ export const toolNames: Array<{ name: string; description: string }>= [
   },
   {
     "name": "repositories-v2-repositories-list",
-    "description": "List Repositories\n\nReturns Git repositories available as reusable workspaces for eval configurations."
+    "description": "List Repositories\n\nReturns built-in templates and enabled saved Git repositories. Pass an item's `id` directly as `repository` in a direct eval run. Public GitHub URLs are also accepted by eval runs but are not included in this list."
   },
   {
     "name": "repositories-v2-repositories-create",