@agentproto/eval-reporters 0.2.0

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Jeremy (agentik.net)
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.md

@@ -0,0 +1,45 @@
+# `@agentproto/eval-reporters`
+
+Adapter-kit family for eval reporter backends. It mirrors the tunnel-provider adapter family in `@agentproto/runtime`: each backend is identified by a slug, may require credentials stored in a 0600 creds file, and exposes a `Telemetry<EvalEvent>` sink.
+
+## Backends
+
+| slug | Needs creds | Description |
+|---|---|---|
+| `langfuse` | yes | Langfuse public ingestion API (`@agentproto/telemetry-langfuse`). |
+| `stderr` | no | Human-readable events to `process.stderr`. |
+| `array` | no | In-memory array for tests / inspection. |
+
+## Setup flow
+
+```ts
+import { makeEvalReporterTools } from "@agentproto/eval-reporters"
+
+const tools = makeEvalReporterTools()
+
+// Configure Langfuse
+await tools.setup_eval_reporter.handler({
+  slug: "langfuse",
+  publicKey: process.env.LANGFUSE_PUBLIC_KEY,
+  secretKey: process.env.LANGFUSE_SECRET_KEY,
+  baseUrl: "https://cloud.langfuse.com",
+  environment: "production",
+})
+
+// List backends — status is now `ready` for langfuse
+const list = await tools.list_eval_reporters.handler()
+console.log(JSON.parse(list.content[0].text))
+
+// Resolve and sink
+const { makeEvalReporterResolver, makeEvalReporterCredsStore } = await import("@agentproto/eval-reporters")
+const resolver = makeEvalReporterResolver(makeEvalReporterCredsStore())
+const handle = await resolver("langfuse")
+const sink = handle!.sink()
+
+runEval({ suite, sink })
+await sink.flush?.()
+```
+
+## Security
+
+Credentials live only in `~/.agentproto/eval-reporter-creds/<slug>.json` with mode `0600`. `list_eval_reporters` and `EvalReporterHandle.info()` expose only capability metadata — never a secret value.

package/dist/index.d.ts

@@ -0,0 +1,110 @@
+import { AdapterCatalog, CredsStore, AdapterHandle, AdapterResolver, SetupField } from '@agentproto/provider-kit';
+import { Telemetry } from '@agentproto/telemetry';
+import { EvalEvent } from '@agentproto/eval';
+import { z } from 'zod';
+
+/** Family key used for the eval-reporter creds store path. */
+declare const EVAL_REPORTER_FAMILY: "eval-reporter";
+/** Static catalog of eval reporter backends. */
+declare const EVAL_REPORTER_CATALOG: AdapterCatalog;
+
+/** Credentials required by the Langfuse eval reporter backend. */
+interface LangfuseCreds {
+    readonly publicKey: string;
+    readonly secretKey: string;
+    readonly baseUrl: string;
+    readonly environment?: string;
+}
+/** Union of all eval-reporter credential shapes (per-slug). */
+type EvalReporterCreds = LangfuseCreds;
+/**
+ * Build the eval-reporter creds store (per-slug, 0600) at
+ * `~/.agentproto/eval-reporter-creds/`.
+ */
+declare function makeEvalReporterCredsStore(home?: string): CredsStore<EvalReporterCreds>;
+
+/**
+ * Safe descriptor returned by {@link EvalReporterHandle.info}. NEVER carries
+ * a secret value — only capability metadata.
+ */
+interface EvalReporterInfo {
+    readonly slug: string;
+    readonly capabilities: {
+        readonly needsCreds: boolean;
+    };
+}
+/** Resolved eval reporter backend. */
+interface EvalReporterHandle extends AdapterHandle {
+    /** Safe descriptor — NEVER a secret. */
+    info(): EvalReporterInfo;
+    /** Build the live sink. langfuse requires creds; stderr/array need none. */
+    sink(): Telemetry<EvalEvent> & {
+        flush?(): Promise<unknown>;
+    };
+}
+interface ResolveEvalReporterOptions {
+    /** Stored creds for the slug, or null when none have been configured. */
+    readonly creds: EvalReporterCreds | null;
+}
+/**
+ * Resolve a catalog slug to a concrete eval reporter handle.
+ *
+ * Throws for unknown slugs so the kit's resolver wraps the miss to `null`.
+ * Known slugs always resolve to a descriptor handle; `sink()` throws when
+ * creds are required but missing.
+ */
+declare function resolveEvalReporter(slug: string, options: ResolveEvalReporterOptions): EvalReporterHandle;
+/**
+ * Build the eval-reporter resolver: reads stored creds and resolves a handle,
+ * returning `null` when the slug is unknown (via the kit's wrapper).
+ */
+declare function makeEvalReporterResolver(credsStore: CredsStore<EvalReporterCreds>): AdapterResolver<EvalReporterHandle>;
+
+/**
+ * MCP tool definitions for the eval-reporter adapter family.
+ *
+ * This module returns tool *specs* (name, description, schema, handler) rather
+ * than registering directly on an `McpServer`. That keeps the package
+ * daemon-free and unit-testable; a daemon package can register the returned
+ * specs on its server later.
+ */
+
+/** Fields collected by the `setup_eval_reporter` tool for Langfuse. */
+declare const LANGFUSE_SETUP_FIELDS: readonly SetupField[];
+interface MakeEvalReporterToolsOptions {
+    /** Home dir override (tests). Defaults to `AGENTPROTO_HOME ?? ~/.agentproto`. */
+    home?: string;
+}
+interface EvalReporterToolResult {
+    readonly content: readonly {
+        readonly type: "text";
+        readonly text: string;
+    }[];
+    readonly isError?: boolean;
+}
+interface EvalReporterListToolSpec {
+    readonly name: "list_eval_reporters";
+    readonly description: string;
+    readonly handler: () => Promise<EvalReporterToolResult>;
+}
+interface EvalReporterSetupToolSpec {
+    readonly name: "setup_eval_reporter";
+    readonly description: string;
+    readonly inputSchema: z.ZodObject<Record<string, z.ZodString>>;
+    readonly handler: (args: Record<string, string>) => Promise<EvalReporterToolResult>;
+}
+interface EvalReporterTools {
+    readonly list_eval_reporters: EvalReporterListToolSpec;
+    readonly setup_eval_reporter: EvalReporterSetupToolSpec;
+}
+/**
+ * Build the eval-reporter family's MCP tool specs.
+ *
+ *   - `list_eval_reporters` — parameterless; returns status + capabilities,
+ *     never creds.
+ *   - `setup_eval_reporter` — multi-field form for configurable reporters
+ *     (currently `langfuse`). Sensitive values are never echoed.
+ */
+declare function makeEvalReporterTools(deps?: MakeEvalReporterToolsOptions): EvalReporterTools;
+
+export { EVAL_REPORTER_CATALOG, EVAL_REPORTER_FAMILY, type EvalReporterCreds, type EvalReporterHandle, type EvalReporterInfo, type EvalReporterListToolSpec, type EvalReporterSetupToolSpec, type EvalReporterToolResult, type EvalReporterTools, LANGFUSE_SETUP_FIELDS, type LangfuseCreds, type MakeEvalReporterToolsOptions, type ResolveEvalReporterOptions, makeEvalReporterCredsStore, makeEvalReporterResolver, makeEvalReporterTools, resolveEvalReporter };

package/dist/index.mjs

@@ -0,0 +1,302 @@
+import { makeCredsStore, makeAdapterResolver, computeStatus } from '@agentproto/provider-kit';
+import { arrayTelemetry, stderrTelemetry } from '@agentproto/telemetry';
+import { langfuseTelemetry } from '@agentproto/telemetry-langfuse';
+import { z } from 'zod';
+
+/**
+ * @agentproto/eval-reporters v0.1.0
+ * Adapter-kit family for eval reporter backends.
+ */
+
+// src/catalog.ts
+var EVAL_REPORTER_FAMILY = "eval-reporter";
+var EVAL_REPORTER_CATALOG = [
+  {
+    slug: "langfuse",
+    name: "Langfuse",
+    description: "Langfuse observability platform via the public ingestion API. Requires public key, secret key, and base URL.",
+    packageName: "@agentproto/telemetry-langfuse",
+    hint: "observability \xB7 scores"
+  },
+  {
+    slug: "stderr",
+    name: "Standard Error",
+    description: "Human-readable eval events written to process stderr.",
+    packageName: "@agentproto/eval-reporters",
+    hint: "local \xB7 debug"
+  },
+  {
+    slug: "array",
+    name: "In-Memory Array",
+    description: "Collects eval events in an in-memory array for tests and inspection.",
+    packageName: "@agentproto/eval-reporters",
+    hint: "local \xB7 test"
+  }
+];
+function makeEvalReporterCredsStore(home) {
+  return makeCredsStore({
+    family: EVAL_REPORTER_FAMILY,
+    ...home ? { home } : {}
+  });
+}
+function resolveEvalReporter(slug, options) {
+  const catalogEntry = EVAL_REPORTER_CATALOG.find((entry) => entry.slug === slug);
+  if (catalogEntry === void 0) {
+    throw new Error(`unknown eval reporter slug: ${slug}`);
+  }
+  switch (slug) {
+    case "langfuse": {
+      const creds = options.creds;
+      return {
+        slug,
+        name: catalogEntry.name,
+        description: catalogEntry.description,
+        version: "0.1.0",
+        requiresSetup: true,
+        check: async () => true,
+        info() {
+          return { slug, capabilities: { needsCreds: true } };
+        },
+        sink() {
+          if (creds === null) {
+            throw new Error(
+              `eval reporter '${slug}' is not configured; run setup_eval_reporter first`
+            );
+          }
+          const langfuseCreds = creds;
+          return langfuseTelemetry({
+            publicKey: langfuseCreds.publicKey,
+            secretKey: langfuseCreds.secretKey,
+            baseUrl: langfuseCreds.baseUrl,
+            environment: langfuseCreds.environment
+          });
+        }
+      };
+    }
+    case "stderr": {
+      return {
+        slug,
+        name: catalogEntry.name,
+        description: catalogEntry.description,
+        version: "0.1.0",
+        requiresSetup: false,
+        check: async () => true,
+        info() {
+          return { slug, capabilities: { needsCreds: false } };
+        },
+        sink() {
+          return stderrTelemetry({
+            prefix: "eval: ",
+            format: formatEvalEvent
+          });
+        }
+      };
+    }
+    case "array": {
+      return {
+        slug,
+        name: catalogEntry.name,
+        description: catalogEntry.description,
+        version: "0.1.0",
+        requiresSetup: false,
+        check: async () => true,
+        info() {
+          return { slug, capabilities: { needsCreds: false } };
+        },
+        sink() {
+          return arrayTelemetry();
+        }
+      };
+    }
+  }
+  throw new Error(`unhandled eval reporter slug: ${slug}`);
+}
+function formatEvalEvent(event) {
+  switch (event.kind) {
+    case "eval.started":
+      return `${event.runId} started suite=${event.suiteId} cases=${event.caseCount} scorers=${event.scorerCount}`;
+    case "eval.case.started":
+      return `${event.runId} case.started case=${event.caseId}`;
+    case "eval.case.scored":
+      return `${event.runId} case.scored case=${event.caseId} scorer=${event.scorerId} value=${event.value} passed=${event.passed}`;
+    case "eval.case.finished":
+      return `${event.runId} case.finished case=${event.caseId} passed=${event.passed}`;
+    case "eval.finished":
+      return `${event.runId} finished suite=${event.suiteId} total=${event.total} passed=${event.passedCount} mean=${event.meanValue}ms=${event.durationMs}`;
+    default: {
+      const _exhaustive = event;
+      return JSON.stringify(_exhaustive);
+    }
+  }
+}
+function makeEvalReporterResolver(credsStore) {
+  return makeAdapterResolver({
+    load: async (slug) => {
+      const creds = await credsStore.read(slug);
+      return resolveEvalReporter(slug, { creds });
+    }
+  });
+}
+var LANGFUSE_SETUP_FIELDS = [
+  {
+    name: "publicKey",
+    description: "Langfuse public key",
+    required: true,
+    sensitive: true
+  },
+  {
+    name: "secretKey",
+    description: "Langfuse secret key",
+    required: true,
+    sensitive: true
+  },
+  {
+    name: "baseUrl",
+    description: "Langfuse base URL, e.g. https://cloud.langfuse.com",
+    required: true,
+    sensitive: false
+  },
+  {
+    name: "environment",
+    description: "Optional environment label",
+    required: false,
+    sensitive: false
+  }
+];
+var DEFAULT_LANGFUSE_BASE_URL = "https://cloud.langfuse.com";
+function makeEvalReporterTools(deps = {}) {
+  const credsStore = makeEvalReporterCredsStore(deps.home);
+  const resolver = makeEvalReporterResolver(credsStore);
+  const lister = async () => {
+    const out = [];
+    for (const entry of EVAL_REPORTER_CATALOG) {
+      const handle = await resolver(entry.slug);
+      if (handle === null) {
+        out.push({
+          slug: entry.slug,
+          name: entry.name,
+          description: entry.description,
+          packageName: entry.packageName,
+          ...entry.hint !== void 0 ? { hint: entry.hint } : {},
+          status: "supported",
+          version: "not installed"
+        });
+        continue;
+      }
+      const credsExist = await credsStore.exists(entry.slug);
+      const status = computeStatus({
+        resolved: true,
+        requiresSetup: handle.requiresSetup,
+        ledgerExists: false,
+        credsExist
+      });
+      out.push({
+        slug: handle.slug,
+        name: handle.name,
+        description: handle.description,
+        packageName: entry.packageName,
+        ...entry.hint !== void 0 ? { hint: entry.hint } : {},
+        status,
+        version: handle.version,
+        info: handle.info()
+      });
+    }
+    return out;
+  };
+  const listTool = {
+    name: "list_eval_reporters",
+    description: "List known eval reporter backends with their status (supported/available/ready) and declared capabilities. Credentials are never returned. Use `setup_eval_reporter` to configure a backend that needs credentials.",
+    handler: async () => {
+      const entries = await lister();
+      return {
+        content: [
+          { type: "text", text: JSON.stringify(entries, null, 2) }
+        ]
+      };
+    }
+  };
+  const shape = {
+    slug: z.string().describe("Eval reporter slug to configure")
+  };
+  for (const field of LANGFUSE_SETUP_FIELDS) {
+    const desc = field.sensitive ? `${field.description} SENSITIVE \u2014 never logged and never echoed back in tool results.` : field.description;
+    shape[field.name] = (field.required !== false ? z.string().min(1, `${field.name} is required`) : z.string()).describe(desc);
+  }
+  const inputSchema = z.object(shape);
+  const setupTool = {
+    name: "setup_eval_reporter",
+    description: "Configure an eval reporter backend that requires credentials (currently `langfuse`). Each field is SENSITIVE \u2014 stored with mode 0600 and never echoed back in tool results.",
+    inputSchema,
+    handler: async (args) => {
+      const slug = args.slug ?? "";
+      if (slug !== "langfuse") {
+        return {
+          content: [
+            {
+              type: "text",
+              text: `setup_eval_reporter: unknown slug '${slug}'. Valid: langfuse`
+            }
+          ],
+          isError: true
+        };
+      }
+      const handle = resolveEvalReporter(slug, { creds: null });
+      if (!handle.requiresSetup) {
+        return {
+          content: [
+            {
+              type: "text",
+              text: `setup_eval_reporter: '${slug}' is not configurable`
+            }
+          ],
+          isError: true
+        };
+      }
+      const missing = LANGFUSE_SETUP_FIELDS.filter(
+        (field) => field.required !== false
+      ).map((field) => field.name).filter((name) => {
+        const value = args[name];
+        return value === void 0 || value.length === 0;
+      });
+      if (missing.length > 0) {
+        return {
+          content: [
+            {
+              type: "text",
+              text: `missing required field(s): ${missing.join(", ")}`
+            }
+          ],
+          isError: true
+        };
+      }
+      const publicKey = args.publicKey ?? "";
+      const secretKey = args.secretKey ?? "";
+      const rawBaseUrl = args.baseUrl;
+      const baseUrl = rawBaseUrl !== void 0 && rawBaseUrl.length > 0 ? rawBaseUrl : DEFAULT_LANGFUSE_BASE_URL;
+      const environment = args.environment;
+      const creds = environment !== void 0 && environment.length > 0 ? { publicKey, secretKey, baseUrl, environment } : { publicKey, secretKey, baseUrl };
+      await credsStore.write(slug, creds);
+      return {
+        content: [
+          {
+            type: "text",
+            text: JSON.stringify(
+              {
+                ok: true,
+                slug,
+                hint: `${slug} configured \u2014 status is now ready`
+              },
+              null,
+              2
+            )
+          }
+        ]
+      };
+    }
+  };
+  return { list_eval_reporters: listTool, setup_eval_reporter: setupTool };
+}
+
+export { EVAL_REPORTER_CATALOG, EVAL_REPORTER_FAMILY, LANGFUSE_SETUP_FIELDS, makeEvalReporterCredsStore, makeEvalReporterResolver, makeEvalReporterTools, resolveEvalReporter };
+//# sourceMappingURL=index.mjs.map
+//# sourceMappingURL=index.mjs.map

package/dist/index.mjs.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/catalog.ts","../src/creds.ts","../src/resolve.ts","../src/tools.ts"],"names":[],"mappings":";;;;;;;;;;;AAGO,IAAM,oBAAA,GAAuB;AAG7B,IAAM,qBAAA,GAAwC;AAAA,EACnD;AAAA,IACE,IAAA,EAAM,UAAA;AAAA,IACN,IAAA,EAAM,UAAA;AAAA,IACN,WAAA,EACE,8GAAA;AAAA,IAEF,WAAA,EAAa,gCAAA;AAAA,IACb,IAAA,EAAM;AAAA,GACR;AAAA,EACA;AAAA,IACE,IAAA,EAAM,QAAA;AAAA,IACN,IAAA,EAAM,gBAAA;AAAA,IACN,WAAA,EAAa,uDAAA;AAAA,IACb,WAAA,EAAa,4BAAA;AAAA,IACb,IAAA,EAAM;AAAA,GACR;AAAA,EACA;AAAA,IACE,IAAA,EAAM,OAAA;AAAA,IACN,IAAA,EAAM,iBAAA;AAAA,IACN,WAAA,EAAa,sEAAA;AAAA,IACb,WAAA,EAAa,4BAAA;AAAA,IACb,IAAA,EAAM;AAAA;AAEV;ACZO,SAAS,2BACd,IAAA,EAC+B;AAC/B,EAAA,OAAO,cAAA,CAAkC;AAAA,IACvC,MAAA,EAAQ,oBAAA;AAAA,IACR,GAAI,IAAA,GAAO,EAAE,IAAA,KAAS;AAAC,GACxB,CAAA;AACH;ACkBO,SAAS,mBAAA,CACd,MACA,OAAA,EACoB;AACpB,EAAA,MAAM,eAAe,qBAAA,CAAsB,IAAA,CAAK,CAAC,KAAA,KAAU,KAAA,CAAM,SAAS,IAAI,CAAA;AAC9E,EAAA,IAAI,iBAAiB,MAAA,EAAW;AAC9B,IAAA,MAAM,IAAI,KAAA,CAAM,CAAA,4BAAA,EAA+B,IAAI,CAAA,CAAE,CAAA;AAAA,EACvD;AAEA,EAAA,QAAQ,IAAA;AAAM,IACZ,KAAK,UAAA,EAAY;AACf,MAAA,MAAM,QAAQ,OAAA,CAAQ,KAAA;AACtB,MAAA,OAAO;AAAA,QACL,IAAA;AAAA,QACA,MAAM,YAAA,CAAa,IAAA;AAAA,QACnB,aAAa,YAAA,CAAa,WAAA;AAAA,QAC1B,OAAA,EAAS,OAAA;AAAA,QACT,aAAA,EAAe,IAAA;AAAA,QACf,OAAO,YAAY,IAAA;AAAA,QACnB,IAAA,GAAO;AACL,UAAA,OAAO,EAAE,IAAA,EAAM,YAAA,EAAc,EAAE,UAAA,EAAY,MAAK,EAAE;AAAA,QACpD,CAAA;AAAA,QACA,IAAA,GAAO;AACL,UAAA,IAAI,UAAU,IAAA,EAAM;AAClB,YAAA,MAAM,IAAI,KAAA;AAAA,cACR,kBAAkB,IAAI,CAAA,kDAAA;AAAA,aACxB;AAAA,UACF;AACA,UAAA,MAAM,aAAA,GAA+B,KAAA;AACrC,UAAA,OAAO,iBAAA,CAAkB;AAAA,YACvB,WAAW,aAAA,CAAc,SAAA;AAAA,YACzB,WAAW,aAAA,CAAc,SAAA;AAAA,YACzB,SAAS,aAAA,CAAc,OAAA;AAAA,YACvB,aAAa,aAAA,CAAc;AAAA,WAC5B,CAAA;AAAA,QACH;AAAA,OACF;AAAA,IACF;AAAA,IACA,KAAK,QAAA,EAAU;AACb,MAAA,OAAO;AAAA,QACL,IAAA;AAAA,QACA,MAAM,YAAA,CAAa,IAAA;AAAA,QACnB,aAAa,YAAA,CAAa,WAAA;AAAA,QAC1B,OAAA,EAAS,OAAA;AAAA,QACT,aAAA,EAAe,KAAA;AAAA,QACf,OAAO,YAAY,IAAA;AAAA,QACnB,IAAA,GAAO;AACL,UAAA,OAAO,EAAE,IAAA,EAAM,YAAA,EAAc,EAAE,UAAA,EAAY,OAAM,EAAE;AAAA,QACrD,CAAA;AAAA,QACA,IAAA,GAAO;AACL,UAAA,OAAO,eAAA,CAA2B;AAAA,YAChC,MAAA,EAAQ,QAAA;AAAA,YACR,MAAA,EAAQ;AAAA,WACT,CAAA;AAAA,QACH;AAAA,OACF;AAAA,IACF;AAAA,IACA,KAAK,OAAA,EAAS;AACZ,MAAA,OAAO;AAAA,QACL,IAAA;AAAA,QACA,MAAM,YAAA,CAAa,IAAA;AAAA,QACnB,aAAa,YAAA,CAAa,WAAA;AAAA,QAC1B,OAAA,EAAS,OAAA;AAAA,QACT,aAAA,EAAe,KAAA;AAAA,QACf,OAAO,YAAY,IAAA;AAAA,QACnB,IAAA,GAAO;AACL,UAAA,OAAO,EAAE,IAAA,EAAM,YAAA,EAAc,EAAE,UAAA,EAAY,OAAM,EAAE;AAAA,QACrD,CAAA;AAAA,QACA,IAAA,GAAO;AACL,UAAA,OAAO,cAAA,EAA0B;AAAA,QACnC;AAAA,OACF;AAAA,IACF;AAAA;AAGF,EAAA,MAAM,IAAI,KAAA,CAAM,CAAA,8BAAA,EAAiC,IAAI,CAAA,CAAE,CAAA;AACzD;AAEA,SAAS,gBAAgB,KAAA,EAA0B;AACjD,EAAA,QAAQ,MAAM,IAAA;AAAM,IAClB,KAAK,cAAA;AACH,MAAA,OAAO,CAAA,EAAG,KAAA,CAAM,KAAK,CAAA,eAAA,EAAkB,KAAA,CAAM,OAAO,CAAA,OAAA,EAAU,KAAA,CAAM,SAAS,CAAA,SAAA,EAAY,KAAA,CAAM,WAAW,CAAA,CAAA;AAAA,IAC5G,KAAK,mBAAA;AACH,MAAA,OAAO,CAAA,EAAG,KAAA,CAAM,KAAK,CAAA,mBAAA,EAAsB,MAAM,MAAM,CAAA,CAAA;AAAA,IACzD,KAAK,kBAAA;AACH,MAAA,OAAO,CAAA,EAAG,KAAA,CAAM,KAAK,CAAA,kBAAA,EAAqB,MAAM,MAAM,CAAA,QAAA,EAAW,KAAA,CAAM,QAAQ,CAAA,OAAA,EAAU,KAAA,CAAM,KAAK,CAAA,QAAA,EAAW,MAAM,MAAM,CAAA,CAAA;AAAA,IAC7H,KAAK,oBAAA;AACH,MAAA,OAAO,CAAA,EAAG,MAAM,KAAK,CAAA,oBAAA,EAAuB,MAAM,MAAM,CAAA,QAAA,EAAW,MAAM,MAAM,CAAA,CAAA;AAAA,IACjF,KAAK,eAAA;AACH,MAAA,OAAO,GAAG,KAAA,CAAM,KAAK,CAAA,gBAAA,EAAmB,KAAA,CAAM,OAAO,CAAA,OAAA,EAAU,KAAA,CAAM,KAAK,CAAA,QAAA,EAAW,MAAM,WAAW,CAAA,MAAA,EAAS,MAAM,SAAS,CAAA,GAAA,EAAM,MAAM,UAAU,CAAA,CAAA;AAAA,IACtJ,SAAS;AACP,MAAA,MAAM,WAAA,GAAqB,KAAA;AAC3B,MAAA,OAAO,IAAA,CAAK,UAAU,WAAW,CAAA;AAAA,IACnC;AAAA;AAEJ;AAMO,SAAS,yBACd,UAAA,EACqC;AACrC,EAAA,OAAO,mBAAA,CAAwC;AAAA,IAC7C,IAAA,EAAM,OAAO,IAAA,KAA8C;AACzD,MAAA,MAAM,KAAA,GAAQ,MAAM,UAAA,CAAW,IAAA,CAAK,IAAI,CAAA;AACxC,MAAA,OAAO,mBAAA,CAAoB,IAAA,EAAM,EAAE,KAAA,EAAO,CAAA;AAAA,IAC5C;AAAA,GACD,CAAA;AACH;AClIO,IAAM,qBAAA,GAA+C;AAAA,EAC1D;AAAA,IACE,IAAA,EAAM,WAAA;AAAA,IACN,WAAA,EAAa,qBAAA;AAAA,IACb,QAAA,EAAU,IAAA;AAAA,IACV,SAAA,EAAW;AAAA,GACb;AAAA,EACA;AAAA,IACE,IAAA,EAAM,WAAA;AAAA,IACN,WAAA,EAAa,qBAAA;AAAA,IACb,QAAA,EAAU,IAAA;AAAA,IACV,SAAA,EAAW;AAAA,GACb;AAAA,EACA;AAAA,IACE,IAAA,EAAM,SAAA;AAAA,IACN,WAAA,EAAa,oDAAA;AAAA,IACb,QAAA,EAAU,IAAA;AAAA,IACV,SAAA,EAAW;AAAA,GACb;AAAA,EACA;AAAA,IACE,IAAA,EAAM,aAAA;AAAA,IACN,WAAA,EAAa,4BAAA;AAAA,IACb,QAAA,EAAU,KAAA;AAAA,IACV,SAAA,EAAW;AAAA;AAEf;AAEA,IAAM,yBAAA,GAA4B,4BAAA;AAsC3B,SAAS,qBAAA,CACd,IAAA,GAAqC,EAAC,EACnB;AACnB,EAAA,MAAM,UAAA,GAAa,0BAAA,CAA2B,IAAA,CAAK,IAAI,CAAA;AACvD,EAAA,MAAM,QAAA,GAAW,yBAAyB,UAAU,CAAA;AAEpD,EAAA,MAAM,SAA0C,YAAY;AAC1D,IAAA,MAAM,MAAwC,EAAC;AAC/C,IAAA,KAAA,MAAW,SAAS,qBAAA,EAAuB;AACzC,MAAA,MAAM,MAAA,GAAS,MAAM,QAAA,CAAS,KAAA,CAAM,IAAI,CAAA;AACxC,MAAA,IAAI,WAAW,IAAA,EAAM;AACnB,QAAA,GAAA,CAAI,IAAA,CAAK;AAAA,UACP,MAAM,KAAA,CAAM,IAAA;AAAA,UACZ,MAAM,KAAA,CAAM,IAAA;AAAA,UACZ,aAAa,KAAA,CAAM,WAAA;AAAA,UACnB,aAAa,KAAA,CAAM,WAAA;AAAA,UACnB,GAAI,MAAM,IAAA,KAAS,MAAA,GAAY,EAAE,IAAA,EAAM,KAAA,CAAM,IAAA,EAAK,GAAI,EAAC;AAAA,UACvD,MAAA,EAAQ,WAAA;AAAA,UACR,OAAA,EAAS;AAAA,SACV,CAAA;AACD,QAAA;AAAA,MACF;AAEA,MAAA,MAAM,UAAA,GAAa,MAAM,UAAA,CAAW,MAAA,CAAO,MAAM,IAAI,CAAA;AACrD,MAAA,MAAM,SAAS,aAAA,CAAc;AAAA,QAC3B,QAAA,EAAU,IAAA;AAAA,QACV,eAAe,MAAA,CAAO,aAAA;AAAA,QACtB,YAAA,EAAc,KAAA;AAAA,QACd;AAAA,OACD,CAAA;AAED,MAAA,GAAA,CAAI,IAAA,CAAK;AAAA,QACP,MAAM,MAAA,CAAO,IAAA;AAAA,QACb,MAAM,MAAA,CAAO,IAAA;AAAA,QACb,aAAa,MAAA,CAAO,WAAA;AAAA,QACpB,aAAa,KAAA,CAAM,WAAA;AAAA,QACnB,GAAI,MAAM,IAAA,KAAS,MAAA,GAAY,EAAE,IAAA,EAAM,KAAA,CAAM,IAAA,EAAK,GAAI,EAAC;AAAA,QACvD,MAAA;AAAA,QACA,SAAS,MAAA,CAAO,OAAA;AAAA,QAChB,IAAA,EAAM,OAAO,IAAA;AAAK,OACnB,CAAA;AAAA,IACH;AACA,IAAA,OAAO,GAAA;AAAA,EACT,CAAA;AAEA,EAAA,MAAM,QAAA,GAAqC;AAAA,IACzC,IAAA,EAAM,qBAAA;AAAA,IACN,WAAA,EACE,qNAAA;AAAA,IAIF,SAAS,YAAY;AACnB,MAAA,MAAM,OAAA,GAAU,MAAM,MAAA,EAAO;AAC7B,MAAA,OAAO;AAAA,QACL,OAAA,EAAS;AAAA,UACP,EAAE,MAAM,MAAA,EAAiB,IAAA,EAAM,KAAK,SAAA,CAAU,OAAA,EAAS,IAAA,EAAM,CAAC,CAAA;AAAE;AAClE,OACF;AAAA,IACF;AAAA,GACF;AAEA,EAAA,MAAM,KAAA,GAAqC;AAAA,IACzC,IAAA,EAAM,CAAA,CAAE,MAAA,EAAO,CAAE,SAAS,iCAAiC;AAAA,GAC7D;AACA,EAAA,KAAA,MAAW,SAAS,qBAAA,EAAuB;AACzC,IAAA,MAAM,OAAO,KAAA,CAAM,SAAA,GACf,GAAG,KAAA,CAAM,WAAW,0EACpB,KAAA,CAAM,WAAA;AACV,IAAA,KAAA,CAAM,KAAA,CAAM,IAAI,CAAA,GAAA,CACd,KAAA,CAAM,aAAa,KAAA,GACf,CAAA,CAAE,QAAO,CAAE,GAAA,CAAI,GAAG,CAAA,EAAG,KAAA,CAAM,IAAI,CAAA,YAAA,CAAc,CAAA,GAC7C,EAAE,MAAA,EAAO,EACb,SAAS,IAAI,CAAA;AAAA,EACjB;AACA,EAAA,MAAM,WAAA,GAAc,CAAA,CAAE,MAAA,CAAO,KAAK,CAAA;AAElC,EAAA,MAAM,SAAA,GAAuC;AAAA,IAC3C,IAAA,EAAM,qBAAA;AAAA,IACN,WAAA,EACE,kLAAA;AAAA,IAGF,WAAA;AAAA,IACA,OAAA,EAAS,OAAO,IAAA,KAAiC;AAC/C,MAAA,MAAM,IAAA,GAAO,KAAK,IAAA,IAAQ,EAAA;AAC1B,MAAA,IAAI,SAAS,UAAA,EAAY;AACvB,QAAA,OAAO;AAAA,UACL,OAAA,EAAS;AAAA,YACP;AAAA,cACE,IAAA,EAAM,MAAA;AAAA,cACN,IAAA,EAAM,sCAAsC,IAAI,CAAA,kBAAA;AAAA;AAClD,WACF;AAAA,UACA,OAAA,EAAS;AAAA,SACX;AAAA,MACF;AAEA,MAAA,MAAM,SAAS,mBAAA,CAAoB,IAAA,EAAM,EAAE,KAAA,EAAO,MAAM,CAAA;AACxD,MAAA,IAAI,CAAC,OAAO,aAAA,EAAe;AACzB,QAAA,OAAO;AAAA,UACL,OAAA,EAAS;AAAA,YACP;AAAA,cACE,IAAA,EAAM,MAAA;AAAA,cACN,IAAA,EAAM,yBAAyB,IAAI,CAAA,qBAAA;AAAA;AACrC,WACF;AAAA,UACA,OAAA,EAAS;AAAA,SACX;AAAA,MACF;AAEA,MAAA,MAAM,UAAU,qBAAA,CAAsB,MAAA;AAAA,QACpC,CAAC,KAAA,KAAU,KAAA,CAAM,QAAA,KAAa;AAAA,OAChC,CACG,IAAI,CAAC,KAAA,KAAU,MAAM,IAAI,CAAA,CACzB,MAAA,CAAO,CAAC,IAAA,KAAS;AAChB,QAAA,MAAM,KAAA,GAAQ,KAAK,IAAI,CAAA;AACvB,QAAA,OAAO,KAAA,KAAU,MAAA,IAAa,KAAA,CAAM,MAAA,KAAW,CAAA;AAAA,MACjD,CAAC,CAAA;AACH,MAAA,IAAI,OAAA,CAAQ,SAAS,CAAA,EAAG;AACtB,QAAA,OAAO;AAAA,UACL,OAAA,EAAS;AAAA,YACP;AAAA,cACE,IAAA,EAAM,MAAA;AAAA,cACN,IAAA,EAAM,CAAA,2BAAA,EAA8B,OAAA,CAAQ,IAAA,CAAK,IAAI,CAAC,CAAA;AAAA;AACxD,WACF;AAAA,UACA,OAAA,EAAS;AAAA,SACX;AAAA,MACF;AAEA,MAAA,MAAM,SAAA,GAAY,KAAK,SAAA,IAAa,EAAA;AACpC,MAAA,MAAM,SAAA,GAAY,KAAK,SAAA,IAAa,EAAA;AACpC,MAAA,MAAM,aAAa,IAAA,CAAK,OAAA;AACxB,MAAA,MAAM,UACJ,UAAA,KAAe,MAAA,IAAa,UAAA,CAAW,MAAA,GAAS,IAC5C,UAAA,GACA,yBAAA;AACN,MAAA,MAAM,cAAc,IAAA,CAAK,WAAA;AACzB,MAAA,MAAM,KAAA,GACJ,WAAA,KAAgB,MAAA,IAAa,WAAA,CAAY,SAAS,CAAA,GAC9C,EAAE,SAAA,EAAW,SAAA,EAAW,SAAS,WAAA,EAAY,GAC7C,EAAE,SAAA,EAAW,WAAW,OAAA,EAAQ;AACtC,MAAA,MAAM,UAAA,CAAW,KAAA,CAAM,IAAA,EAAM,KAAK,CAAA;AAElC,MAAA,OAAO;AAAA,QACL,OAAA,EAAS;AAAA,UACP;AAAA,YACE,IAAA,EAAM,MAAA;AAAA,YACN,MAAM,IAAA,CAAK,SAAA;AAAA,cACT;AAAA,gBACE,EAAA,EAAI,IAAA;AAAA,gBACJ,IAAA;AAAA,gBACA,IAAA,EAAM,GAAG,IAAI,CAAA,sCAAA;AAAA,eACf;AAAA,cACA,IAAA;AAAA,cACA;AAAA;AACF;AACF;AACF,OACF;AAAA,IACF;AAAA,GACF;AAEA,EAAA,OAAO,EAAE,mBAAA,EAAqB,QAAA,EAAU,mBAAA,EAAqB,SAAA,EAAU;AACzE","file":"index.mjs","sourcesContent":["import type { AdapterCatalog } from \"@agentproto/provider-kit\"\n\n/** Family key used for the eval-reporter creds store path. */\nexport const EVAL_REPORTER_FAMILY = \"eval-reporter\" as const\n\n/** Static catalog of eval reporter backends. */\nexport const EVAL_REPORTER_CATALOG: AdapterCatalog = [\n  {\n    slug: \"langfuse\",\n    name: \"Langfuse\",\n    description:\n      \"Langfuse observability platform via the public ingestion API. \" +\n      \"Requires public key, secret key, and base URL.\",\n    packageName: \"@agentproto/telemetry-langfuse\",\n    hint: \"observability · scores\",\n  },\n  {\n    slug: \"stderr\",\n    name: \"Standard Error\",\n    description: \"Human-readable eval events written to process stderr.\",\n    packageName: \"@agentproto/eval-reporters\",\n    hint: \"local · debug\",\n  },\n  {\n    slug: \"array\",\n    name: \"In-Memory Array\",\n    description: \"Collects eval events in an in-memory array for tests and inspection.\",\n    packageName: \"@agentproto/eval-reporters\",\n    hint: \"local · test\",\n  },\n]\n","import { makeCredsStore, type CredsStore } from \"@agentproto/provider-kit\"\nimport { EVAL_REPORTER_FAMILY } from \"./catalog.js\"\n\n/** Credentials required by the Langfuse eval reporter backend. */\nexport interface LangfuseCreds {\n  readonly publicKey: string\n  readonly secretKey: string\n  readonly baseUrl: string\n  readonly environment?: string\n}\n\n/** Union of all eval-reporter credential shapes (per-slug). */\nexport type EvalReporterCreds = LangfuseCreds\n\n/**\n * Build the eval-reporter creds store (per-slug, 0600) at\n * `~/.agentproto/eval-reporter-creds/`.\n */\nexport function makeEvalReporterCredsStore(\n  home?: string,\n): CredsStore<EvalReporterCreds> {\n  return makeCredsStore<EvalReporterCreds>({\n    family: EVAL_REPORTER_FAMILY,\n    ...(home ? { home } : {}),\n  })\n}\n","import { stderrTelemetry, arrayTelemetry, type Telemetry } from \"@agentproto/telemetry\"\nimport { langfuseTelemetry } from \"@agentproto/telemetry-langfuse\"\nimport type { EvalEvent } from \"@agentproto/eval\"\nimport {\n  makeAdapterResolver,\n  type AdapterHandle,\n  type AdapterResolver,\n  type CredsStore,\n} from \"@agentproto/provider-kit\"\nimport { EVAL_REPORTER_CATALOG } from \"./catalog.js\"\nimport type { EvalReporterCreds, LangfuseCreds } from \"./creds.js\"\n\n/**\n * Safe descriptor returned by {@link EvalReporterHandle.info}. NEVER carries\n * a secret value — only capability metadata.\n */\nexport interface EvalReporterInfo {\n  readonly slug: string\n  readonly capabilities: {\n    readonly needsCreds: boolean\n  }\n}\n\n/** Resolved eval reporter backend. */\nexport interface EvalReporterHandle extends AdapterHandle {\n  /** Safe descriptor — NEVER a secret. */\n  info(): EvalReporterInfo\n  /** Build the live sink. langfuse requires creds; stderr/array need none. */\n  sink(): Telemetry<EvalEvent> & { flush?(): Promise<unknown> }\n}\n\nexport interface ResolveEvalReporterOptions {\n  /** Stored creds for the slug, or null when none have been configured. */\n  readonly creds: EvalReporterCreds | null\n}\n\n/**\n * Resolve a catalog slug to a concrete eval reporter handle.\n *\n * Throws for unknown slugs so the kit's resolver wraps the miss to `null`.\n * Known slugs always resolve to a descriptor handle; `sink()` throws when\n * creds are required but missing.\n */\nexport function resolveEvalReporter(\n  slug: string,\n  options: ResolveEvalReporterOptions,\n): EvalReporterHandle {\n  const catalogEntry = EVAL_REPORTER_CATALOG.find((entry) => entry.slug === slug)\n  if (catalogEntry === undefined) {\n    throw new Error(`unknown eval reporter slug: ${slug}`)\n  }\n\n  switch (slug) {\n    case \"langfuse\": {\n      const creds = options.creds\n      return {\n        slug,\n        name: catalogEntry.name,\n        description: catalogEntry.description,\n        version: \"0.1.0\",\n        requiresSetup: true,\n        check: async () => true,\n        info() {\n          return { slug, capabilities: { needsCreds: true } }\n        },\n        sink() {\n          if (creds === null) {\n            throw new Error(\n              `eval reporter '${slug}' is not configured; run setup_eval_reporter first`,\n            )\n          }\n          const langfuseCreds: LangfuseCreds = creds\n          return langfuseTelemetry({\n            publicKey: langfuseCreds.publicKey,\n            secretKey: langfuseCreds.secretKey,\n            baseUrl: langfuseCreds.baseUrl,\n            environment: langfuseCreds.environment,\n          })\n        },\n      }\n    }\n    case \"stderr\": {\n      return {\n        slug,\n        name: catalogEntry.name,\n        description: catalogEntry.description,\n        version: \"0.1.0\",\n        requiresSetup: false,\n        check: async () => true,\n        info() {\n          return { slug, capabilities: { needsCreds: false } }\n        },\n        sink() {\n          return stderrTelemetry<EvalEvent>({\n            prefix: \"eval: \",\n            format: formatEvalEvent,\n          })\n        },\n      }\n    }\n    case \"array\": {\n      return {\n        slug,\n        name: catalogEntry.name,\n        description: catalogEntry.description,\n        version: \"0.1.0\",\n        requiresSetup: false,\n        check: async () => true,\n        info() {\n          return { slug, capabilities: { needsCreds: false } }\n        },\n        sink() {\n          return arrayTelemetry<EvalEvent>()\n        },\n      }\n    }\n  }\n\n  throw new Error(`unhandled eval reporter slug: ${slug}`)\n}\n\nfunction formatEvalEvent(event: EvalEvent): string {\n  switch (event.kind) {\n    case \"eval.started\":\n      return `${event.runId} started suite=${event.suiteId} cases=${event.caseCount} scorers=${event.scorerCount}`\n    case \"eval.case.started\":\n      return `${event.runId} case.started case=${event.caseId}`\n    case \"eval.case.scored\":\n      return `${event.runId} case.scored case=${event.caseId} scorer=${event.scorerId} value=${event.value} passed=${event.passed}`\n    case \"eval.case.finished\":\n      return `${event.runId} case.finished case=${event.caseId} passed=${event.passed}`\n    case \"eval.finished\":\n      return `${event.runId} finished suite=${event.suiteId} total=${event.total} passed=${event.passedCount} mean=${event.meanValue}ms=${event.durationMs}`\n    default: {\n      const _exhaustive: never = event\n      return JSON.stringify(_exhaustive)\n    }\n  }\n}\n\n/**\n * Build the eval-reporter resolver: reads stored creds and resolves a handle,\n * returning `null` when the slug is unknown (via the kit's wrapper).\n */\nexport function makeEvalReporterResolver(\n  credsStore: CredsStore<EvalReporterCreds>,\n): AdapterResolver<EvalReporterHandle> {\n  return makeAdapterResolver<EvalReporterHandle>({\n    load: async (slug: string): Promise<EvalReporterHandle> => {\n      const creds = await credsStore.read(slug)\n      return resolveEvalReporter(slug, { creds })\n    },\n  })\n}\n","/**\n * MCP tool definitions for the eval-reporter adapter family.\n *\n * This module returns tool *specs* (name, description, schema, handler) rather\n * than registering directly on an `McpServer`. That keeps the package\n * daemon-free and unit-testable; a daemon package can register the returned\n * specs on its server later.\n */\n\nimport { z } from \"zod\"\nimport type { AdapterEntry, AdapterLister, SetupField } from \"@agentproto/provider-kit\"\nimport { computeStatus } from \"@agentproto/provider-kit\"\nimport type { EvalEvent } from \"@agentproto/eval\"\nimport { EVAL_REPORTER_CATALOG } from \"./catalog.js\"\nimport {\n  makeEvalReporterCredsStore,\n  type EvalReporterCreds,\n  type LangfuseCreds,\n} from \"./creds.js\"\nimport type { EvalReporterHandle, EvalReporterInfo } from \"./resolve.js\"\nimport { makeEvalReporterResolver, resolveEvalReporter } from \"./resolve.js\"\n\n/** Fields collected by the `setup_eval_reporter` tool for Langfuse. */\nexport const LANGFUSE_SETUP_FIELDS: readonly SetupField[] = [\n  {\n    name: \"publicKey\",\n    description: \"Langfuse public key\",\n    required: true,\n    sensitive: true,\n  },\n  {\n    name: \"secretKey\",\n    description: \"Langfuse secret key\",\n    required: true,\n    sensitive: true,\n  },\n  {\n    name: \"baseUrl\",\n    description: \"Langfuse base URL, e.g. https://cloud.langfuse.com\",\n    required: true,\n    sensitive: false,\n  },\n  {\n    name: \"environment\",\n    description: \"Optional environment label\",\n    required: false,\n    sensitive: false,\n  },\n]\n\nconst DEFAULT_LANGFUSE_BASE_URL = \"https://cloud.langfuse.com\"\n\nexport interface MakeEvalReporterToolsOptions {\n  /** Home dir override (tests). Defaults to `AGENTPROTO_HOME ?? ~/.agentproto`. */\n  home?: string\n}\n\nexport interface EvalReporterToolResult {\n  readonly content: readonly { readonly type: \"text\"; readonly text: string }[]\n  readonly isError?: boolean\n}\n\nexport interface EvalReporterListToolSpec {\n  readonly name: \"list_eval_reporters\"\n  readonly description: string\n  readonly handler: () => Promise<EvalReporterToolResult>\n}\n\nexport interface EvalReporterSetupToolSpec {\n  readonly name: \"setup_eval_reporter\"\n  readonly description: string\n  readonly inputSchema: z.ZodObject<Record<string, z.ZodString>>\n  readonly handler: (args: Record<string, string>) => Promise<EvalReporterToolResult>\n}\n\nexport interface EvalReporterTools {\n  readonly list_eval_reporters: EvalReporterListToolSpec\n  readonly setup_eval_reporter: EvalReporterSetupToolSpec\n}\n\n/**\n * Build the eval-reporter family's MCP tool specs.\n *\n *   - `list_eval_reporters` — parameterless; returns status + capabilities,\n *     never creds.\n *   - `setup_eval_reporter` — multi-field form for configurable reporters\n *     (currently `langfuse`). Sensitive values are never echoed.\n */\nexport function makeEvalReporterTools(\n  deps: MakeEvalReporterToolsOptions = {},\n): EvalReporterTools {\n  const credsStore = makeEvalReporterCredsStore(deps.home)\n  const resolver = makeEvalReporterResolver(credsStore)\n\n  const lister: AdapterLister<EvalReporterInfo> = async () => {\n    const out: AdapterEntry<EvalReporterInfo>[] = []\n    for (const entry of EVAL_REPORTER_CATALOG) {\n      const handle = await resolver(entry.slug)\n      if (handle === null) {\n        out.push({\n          slug: entry.slug,\n          name: entry.name,\n          description: entry.description,\n          packageName: entry.packageName,\n          ...(entry.hint !== undefined ? { hint: entry.hint } : {}),\n          status: \"supported\",\n          version: \"not installed\",\n        })\n        continue\n      }\n\n      const credsExist = await credsStore.exists(entry.slug)\n      const status = computeStatus({\n        resolved: true,\n        requiresSetup: handle.requiresSetup,\n        ledgerExists: false,\n        credsExist,\n      })\n\n      out.push({\n        slug: handle.slug,\n        name: handle.name,\n        description: handle.description,\n        packageName: entry.packageName,\n        ...(entry.hint !== undefined ? { hint: entry.hint } : {}),\n        status,\n        version: handle.version,\n        info: handle.info(),\n      })\n    }\n    return out\n  }\n\n  const listTool: EvalReporterListToolSpec = {\n    name: \"list_eval_reporters\",\n    description:\n      \"List known eval reporter backends with their status (supported/\" +\n      \"available/ready) and declared capabilities. Credentials are never \" +\n      \"returned. Use `setup_eval_reporter` to configure a backend that \" +\n      \"needs credentials.\",\n    handler: async () => {\n      const entries = await lister()\n      return {\n        content: [\n          { type: \"text\" as const, text: JSON.stringify(entries, null, 2) },\n        ],\n      }\n    },\n  }\n\n  const shape: Record<string, z.ZodString> = {\n    slug: z.string().describe(\"Eval reporter slug to configure\"),\n  }\n  for (const field of LANGFUSE_SETUP_FIELDS) {\n    const desc = field.sensitive\n      ? `${field.description} SENSITIVE — never logged and never echoed back in tool results.`\n      : field.description\n    shape[field.name] = (\n      field.required !== false\n        ? z.string().min(1, `${field.name} is required`)\n        : z.string()\n    ).describe(desc)\n  }\n  const inputSchema = z.object(shape)\n\n  const setupTool: EvalReporterSetupToolSpec = {\n    name: \"setup_eval_reporter\",\n    description:\n      \"Configure an eval reporter backend that requires credentials \" +\n      \"(currently `langfuse`). Each field is SENSITIVE — stored with mode \" +\n      \"0600 and never echoed back in tool results.\",\n    inputSchema,\n    handler: async (args: Record<string, string>) => {\n      const slug = args.slug ?? \"\"\n      if (slug !== \"langfuse\") {\n        return {\n          content: [\n            {\n              type: \"text\" as const,\n              text: `setup_eval_reporter: unknown slug '${slug}'. Valid: langfuse`,\n            },\n          ],\n          isError: true,\n        }\n      }\n\n      const handle = resolveEvalReporter(slug, { creds: null })\n      if (!handle.requiresSetup) {\n        return {\n          content: [\n            {\n              type: \"text\" as const,\n              text: `setup_eval_reporter: '${slug}' is not configurable`,\n            },\n          ],\n          isError: true,\n        }\n      }\n\n      const missing = LANGFUSE_SETUP_FIELDS.filter(\n        (field) => field.required !== false,\n      )\n        .map((field) => field.name)\n        .filter((name) => {\n          const value = args[name]\n          return value === undefined || value.length === 0\n        })\n      if (missing.length > 0) {\n        return {\n          content: [\n            {\n              type: \"text\" as const,\n              text: `missing required field(s): ${missing.join(\", \")}`,\n            },\n          ],\n          isError: true,\n        }\n      }\n\n      const publicKey = args.publicKey ?? \"\"\n      const secretKey = args.secretKey ?? \"\"\n      const rawBaseUrl = args.baseUrl\n      const baseUrl =\n        rawBaseUrl !== undefined && rawBaseUrl.length > 0\n          ? rawBaseUrl\n          : DEFAULT_LANGFUSE_BASE_URL\n      const environment = args.environment\n      const creds: LangfuseCreds =\n        environment !== undefined && environment.length > 0\n          ? { publicKey, secretKey, baseUrl, environment }\n          : { publicKey, secretKey, baseUrl }\n      await credsStore.write(slug, creds)\n\n      return {\n        content: [\n          {\n            type: \"text\" as const,\n            text: JSON.stringify(\n              {\n                ok: true,\n                slug,\n                hint: `${slug} configured — status is now ready`,\n              },\n              null,\n              2,\n            ),\n          },\n        ],\n      }\n    },\n  }\n\n  return { list_eval_reporters: listTool, setup_eval_reporter: setupTool }\n}\n"]}

package/package.json

@@ -0,0 +1,70 @@
+{
+  "name": "@agentproto/eval-reporters",
+  "version": "0.2.0",
+  "description": "Adapter-kit family for eval reporters. Pluggable backends (Langfuse, stderr, in-memory array) configured by slug + credentials, mirroring the tunnel-provider adapter family.",
+  "keywords": [
+    "agentproto",
+    "eval",
+    "reporter",
+    "adapter",
+    "langfuse",
+    "telemetry",
+    "open-standard",
+    "agentic"
+  ],
+  "homepage": "https://agentproto.sh",
+  "repository": {
+    "type": "git",
+    "url": "https://github.com/agentproto/ts",
+    "directory": "packages/eval-reporters"
+  },
+  "bugs": {
+    "url": "https://github.com/agentproto/ts/issues"
+  },
+  "license": "MIT",
+  "type": "module",
+  "main": "dist/index.mjs",
+  "module": "dist/index.mjs",
+  "types": "dist/index.d.ts",
+  "exports": {
+    ".": {
+      "types": "./dist/index.d.ts",
+      "import": "./dist/index.mjs",
+      "default": "./dist/index.mjs"
+    },
+    "./package.json": "./package.json"
+  },
+  "files": [
+    "dist",
+    "README.md",
+    "LICENSE"
+  ],
+  "publishConfig": {
+    "access": "public"
+  },
+  "dependencies": {
+    "zod": "^4.4.3",
+    "@agentproto/eval": "0.2.0",
+    "@agentproto/provider-kit": "0.2.0",
+    "@agentproto/telemetry": "0.2.0",
+    "@agentproto/telemetry-langfuse": "0.2.0"
+  },
+  "devDependencies": {
+    "@types/node": "^25.6.2",
+    "tsup": "^8.5.1",
+    "typescript": "^5.9.3",
+    "vitest": "^3.2.4",
+    "@agentproto/tooling": "0.1.0-alpha.0"
+  },
+  "engines": {
+    "node": ">=20.9.0"
+  },
+  "scripts": {
+    "dev": "tsup --watch",
+    "build": "tsup",
+    "clean": "rm -rf dist",
+    "check-types": "tsc --noEmit",
+    "test": "vitest run",
+    "test:watch": "vitest"
+  }
+}