@elizaos/agent 2.0.0-beta.2 → 2.0.11-beta.6

package/packages/agent/src/cli/benchmark.js

@@ -1,293 +0,0 @@
-/**
- * CLI subcommand for headless benchmark execution.
- *
- * Accepts a task as JSON (from --task <file> or stdin), boots the runtime
- * in headless mode, sends the prompt through handleMessage (the real agent
- * loop including action selection), captures the response, and writes a
- * JSON result to stdout.
- *
- * Task prompts are augmented with type-aware instructions so the agent
- * produces thorough, structured responses through its normal flow.
- *
- * Server mode (--server) keeps the runtime alive and reads line-delimited
- * JSON from stdin, writing one result line per task.
- */
-import crypto from "node:crypto";
-import { readFileSync } from "node:fs";
-import process from "node:process";
-import * as readline from "node:readline";
-import { ChannelType, createMessageMemory, logger, stringToUuid, } from "@elizaos/core";
-// ---------------------------------------------------------------------------
-// Prompt augmentation — steers the agent via the normal message flow
-// ---------------------------------------------------------------------------
-const RESEARCH_AUGMENTATION = "\n\nPlease give a thorough, structured answer with ## headings, bullet points, and a conclusion. Be detailed and comprehensive.";
-const CODING_AUGMENTATION = "\n\nPlease write the complete code implementation directly in your response using ```typescript blocks. Include all imports, types, and error handling. Do not delegate — write the code yourself.";
-const DEFAULT_AUGMENTATION = "\n\nPlease give a detailed, structured answer with headings and bullet points.";
-function detectTaskType(task) {
-    if (task.type)
-        return task.type;
-    if (/\b(implement|build|create|write|code|function|class|module|component|api|endpoint|cli|test suite|refactor|debug|fix.*bug)\b/i.test(task.prompt)) {
-        return "coding";
-    }
-    return "research";
-}
-function augmentPrompt(task) {
-    const taskType = detectTaskType(task);
-    switch (taskType) {
-        case "research":
-            return task.prompt + RESEARCH_AUGMENTATION;
-        case "coding":
-            return task.prompt + CODING_AUGMENTATION;
-        default:
-            return task.prompt + DEFAULT_AUGMENTATION;
-    }
-}
-// ---------------------------------------------------------------------------
-// Task execution — uses handleMessage (the real agent loop)
-// ---------------------------------------------------------------------------
-async function runTask(runtime, task, timeoutMs) {
-    const start = performance.now();
-    const taskType = detectTaskType(task);
-    const userId = crypto.randomUUID();
-    const roomId = stringToUuid(`benchmark-${task.id}`);
-    const worldId = stringToUuid(`benchmark-world-${task.id}`);
-    const messageServerId = stringToUuid(`benchmark-server-${task.id}`);
-    try {
-        await runtime.ensureConnection({
-            entityId: userId,
-            roomId,
-            worldId,
-            userName: "Benchmark",
-            source: "benchmark",
-            channelId: `benchmark-${task.id}`,
-            type: ChannelType.DM,
-            messageServerId,
-            metadata: { ownership: { ownerId: userId } },
-        });
-        const augmentedPrompt = augmentPrompt(task);
-        const message = createMessageMemory({
-            id: crypto.randomUUID(),
-            entityId: userId,
-            roomId,
-            content: {
-                text: augmentedPrompt,
-                source: "benchmark",
-                channelType: ChannelType.DM,
-            },
-        });
-        if (!runtime.messageService) {
-            return {
-                id: task.id,
-                response: "",
-                actions_taken: [taskType],
-                duration_ms: Math.round(performance.now() - start),
-                success: false,
-                error: "runtime.messageService is not available",
-            };
-        }
-        let callbackText = "";
-        let streamText = "";
-        const actionsTaken = [taskType];
-        // Race message handling against the timeout.
-        // Text can arrive via three channels:
-        //   1. callback — action handler results
-        //   2. onStreamChunk — streamed LLM tokens
-        //   3. result.responseContent — final composed response
-        // We capture all three and deduplicate.
-        const result = (await Promise.race([
-            (async () => {
-                const handleResult = await runtime.messageService?.handleMessage(runtime, message, async (content) => {
-                    if (content?.text) {
-                        callbackText += content.text;
-                    }
-                    // Track actions taken
-                    const action = content?.action;
-                    if (typeof action === "string" && !actionsTaken.includes(action)) {
-                        actionsTaken.push(action);
-                    }
-                    return [];
-                }, {
-                    onStreamChunk: async (chunk) => {
-                        if (chunk)
-                            streamText += chunk;
-                    },
-                });
-                return handleResult;
-            })(),
-            new Promise((resolve) => setTimeout(() => resolve("timeout"), timeoutMs)),
-        ]));
-        if (result === "timeout") {
-            const responseText = streamText || callbackText;
-            return {
-                id: task.id,
-                response: responseText,
-                actions_taken: actionsTaken,
-                duration_ms: Math.round(performance.now() - start),
-                success: false,
-                error: `Timeout after ${timeoutMs}ms`,
-            };
-        }
-        // Extract text from all channels, preferring the richest source.
-        // result.responseContent has the final composed text from the LLM.
-        const resultRecord = typeof result === "object" && result !== null
-            ? result
-            : null;
-        const responseContent = resultRecord?.responseContent &&
-            typeof resultRecord.responseContent === "object"
-            ? resultRecord.responseContent
-            : null;
-        const resultText = responseContent && typeof responseContent.text === "string"
-            ? responseContent.text
-            : "";
-        // Also check responseMessages for additional text
-        const responseMessages = Array.isArray(resultRecord?.responseMessages)
-            ? resultRecord.responseMessages
-            : [];
-        const messagesText = responseMessages
-            .map((m) => m.content?.text ?? "")
-            .filter(Boolean)
-            .join("\n");
-        // Pick the best source — longest non-empty wins, since the same
-        // text may appear in multiple channels.
-        const candidates = [
-            resultText,
-            messagesText,
-            streamText,
-            callbackText,
-        ].filter(Boolean);
-        const responseText = candidates.sort((a, b) => b.length - a.length)[0] ?? "";
-        return {
-            id: task.id,
-            response: responseText,
-            actions_taken: actionsTaken,
-            duration_ms: Math.round(performance.now() - start),
-            success: true,
-        };
-    }
-    catch (err) {
-        return {
-            id: task.id,
-            response: "",
-            actions_taken: [taskType],
-            duration_ms: Math.round(performance.now() - start),
-            success: false,
-            error: err instanceof Error ? err.message : String(err),
-        };
-    }
-}
-/** Parse a JSON string into a BenchmarkTask, throwing on invalid input. */
-function parseTask(raw) {
-    const parsed = JSON.parse(raw);
-    if (typeof parsed !== "object" ||
-        parsed === null ||
-        !("id" in parsed) ||
-        !("prompt" in parsed)) {
-        throw new Error('Invalid task JSON: must contain at minimum "id" and "prompt" fields');
-    }
-    return parsed;
-}
-/** Write a result as a single JSON line to stdout. */
-function writeResult(result) {
-    process.stdout.write(`${JSON.stringify(result)}\n`);
-}
-/**
- * Server mode: read line-delimited JSON tasks from stdin, process each
- * against the running runtime, and write results to stdout.
- */
-async function runServerMode(runtime, timeoutMs) {
-    const rl = readline.createInterface({
-        input: process.stdin,
-        terminal: false,
-    });
-    for await (const line of rl) {
-        const trimmed = line.trim();
-        if (!trimmed)
-            continue;
-        try {
-            const task = parseTask(trimmed);
-            const result = await runTask(runtime, task, timeoutMs);
-            writeResult(result);
-        }
-        catch (err) {
-            const errorResult = {
-                id: "unknown",
-                response: "",
-                actions_taken: [],
-                duration_ms: 0,
-                success: false,
-                error: `Failed to parse task: ${err instanceof Error ? err.message : String(err)}`,
-            };
-            writeResult(errorResult);
-        }
-    }
-}
-/**
- * Entry point for the `benchmark` CLI subcommand.
- */
-export async function runBenchmark(opts) {
-    const timeoutMs = Number.parseInt(opts.timeout, 10);
-    if (Number.isNaN(timeoutMs) || timeoutMs <= 0) {
-        console.error("[benchmark] Invalid timeout value");
-        process.exit(2);
-    }
-    // Suppress noisy runtime logs — benchmark output must be clean JSON on stdout
-    if (!process.env.LOG_LEVEL) {
-        process.env.LOG_LEVEL = "error";
-    }
-    process.env.ELIZA_HEADLESS = "1";
-    const { bootElizaRuntime } = await import("../runtime/eliza.js");
-    let runtime;
-    try {
-        runtime = await bootElizaRuntime();
-    }
-    catch (err) {
-        const errorResult = {
-            id: opts.task ? "file" : "stdin",
-            response: "",
-            actions_taken: [],
-            duration_ms: 0,
-            success: false,
-            error: `Runtime boot failed: ${err instanceof Error ? err.message : String(err)}`,
-        };
-        writeResult(errorResult);
-        process.exit(1);
-    }
-    if (opts.server) {
-        await runServerMode(runtime, timeoutMs);
-        logger.info("[benchmark] EOF on stdin, shutting down");
-        process.exit(0);
-    }
-    // Single-task mode
-    let taskJson;
-    if (opts.task) {
-        try {
-            taskJson = readFileSync(opts.task, "utf-8");
-        }
-        catch (err) {
-            console.error(`[benchmark] Failed to read task file: ${err instanceof Error ? err.message : String(err)}`);
-            process.exit(2);
-        }
-    }
-    else {
-        const chunks = [];
-        for await (const chunk of process.stdin) {
-            chunks.push(chunk);
-        }
-        taskJson = Buffer.concat(chunks).toString("utf-8").trim();
-        if (!taskJson) {
-            console.error("[benchmark] No task provided. Use --task <file> or pipe JSON to stdin.");
-            process.exit(2);
-        }
-    }
-    let task;
-    try {
-        task = parseTask(taskJson);
-    }
-    catch (err) {
-        console.error(`[benchmark] Invalid task JSON: ${err instanceof Error ? err.message : String(err)}`);
-        process.exit(2);
-    }
-    const result = await runTask(runtime, task, timeoutMs);
-    writeResult(result);
-    process.exit(result.success ? 0 : 1);
-}

package/packages/agent/src/cli/index.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../../../src/cli/index.ts"],"names":[],"mappings":"AA6BA,wBAAsB,gBAAgB,CACpC,IAAI,GAAE,MAAM,EAAiB,GAC5B,OAAO,CAAC,IAAI,CAAC,CAkFf"}

package/packages/agent/src/cli/index.js

@@ -1,94 +0,0 @@
-import { createRequire } from "node:module";
-import process from "node:process";
-function printHelp() {
-    console.log(`eliza-autonomous
-
-Usage:
-  eliza-autonomous serve
-  eliza-autonomous runtime
-  eliza-autonomous benchmark [options]
-
-Commands:
-  serve      Start the autonomous backend in server-only mode
-  runtime    Boot the runtime without entering the API/CLI wrapper
-  benchmark  Run a benchmark task headlessly against the agent
-
-Benchmark options:
-  --task <path>    Path to task JSON file
-  --server         Keep runtime alive and accept tasks via stdin (line-delimited JSON)
-  --timeout <ms>   Timeout per task in milliseconds (default: 120000)
-`);
-}
-function printVersion() {
-    const require = createRequire(import.meta.url);
-    const pkg = require("../../package.json");
-    console.log(pkg.version);
-}
-export async function runAutonomousCli(argv = process.argv) {
-    const command = argv[2] ?? "serve";
-    if (command === "--version" || command === "-v" || command === "version") {
-        printVersion();
-        return;
-    }
-    if (command === "--help" || command === "-h" || command === "help") {
-        printHelp();
-        return;
-    }
-    if (command === "runtime") {
-        const { bootElizaRuntime } = await import("../runtime/index.js");
-        await bootElizaRuntime();
-        return;
-    }
-    if (command === "serve" || command === "start") {
-        const { startEliza } = await import("../runtime/index.js");
-        const runtime = await startEliza({ serverOnly: true });
-        console.log(`[cli] startEliza returned: runtime=${runtime ? "present" : "null"}, ELIZA_LOCAL_LLAMA=${process.env.ELIZA_LOCAL_LLAMA ?? "(unset)"}`);
-        // AOSP-only post-boot wiring. The upstream `startEliza` does not
-        // register local-inference handlers — that lives in the
-        // `@elizaos/app-core` runtime wrapper, which the mobile agent
-        // bundle cannot import (would create an `agent → app-core →
-        // agent` workspace cycle). Bootstrapping the AOSP llama loader
-        // and ModelType handlers here keeps the registration in the
-        // agent package and out of the bundler's cycle path. No-op when
-        // `ELIZA_LOCAL_LLAMA !== "1"`.
-        if (runtime && process.env.ELIZA_LOCAL_LLAMA?.trim() === "1") {
-            console.log("[cli] importing aosp-local-inference-bootstrap…");
-            const { ensureAospLocalInferenceHandlers } = await import("@elizaos/plugin-aosp-local-inference");
-            console.log("[cli] calling ensureAospLocalInferenceHandlers(runtime)…");
-            const ok = await ensureAospLocalInferenceHandlers(runtime);
-            console.log(`[cli] ensureAospLocalInferenceHandlers returned ${ok}`);
-        }
-        else if (runtime &&
-            process.env.ELIZA_DEVICE_BRIDGE_ENABLED?.trim() === "1") {
-            console.log("[cli] importing mobile-device-bridge-bootstrap…");
-            const { ensureMobileDeviceBridgeInferenceHandlers } = await import("@elizaos/plugin-capacitor-bridge");
-            console.log("[cli] calling ensureMobileDeviceBridgeInferenceHandlers(runtime)…");
-            const ok = await ensureMobileDeviceBridgeInferenceHandlers(runtime);
-            console.log(`[cli] ensureMobileDeviceBridgeInferenceHandlers returned ${ok}`);
-        }
-        return;
-    }
-    if (command === "benchmark") {
-        const { runBenchmark } = await import("./benchmark.js");
-        // Parse benchmark-specific flags from argv
-        const opts = {
-            task: undefined,
-            server: false,
-            timeout: "120000",
-        };
-        for (let i = 3; i < argv.length; i++) {
-            if (argv[i] === "--task" && argv[i + 1]) {
-                opts.task = argv[++i];
-            }
-            else if (argv[i] === "--server") {
-                opts.server = true;
-            }
-            else if (argv[i] === "--timeout" && argv[i + 1]) {
-                opts.timeout = argv[++i];
-            }
-        }
-        await runBenchmark(opts);
-        return;
-    }
-    throw new Error(`Unknown command: ${command}`);
-}

package/packages/agent/src/cli/parse-duration.d.ts

@@ -1,5 +0,0 @@
-export type DurationMsParseOptions = {
-    defaultUnit?: "ms" | "s" | "m" | "h" | "d";
-};
-export declare function parseDurationMs(raw: string, opts?: DurationMsParseOptions): number;
-//# sourceMappingURL=parse-duration.d.ts.map

package/packages/agent/src/cli/parse-duration.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"parse-duration.d.ts","sourceRoot":"","sources":["../../../../../src/cli/parse-duration.ts"],"names":[],"mappings":"AAAA,MAAM,MAAM,sBAAsB,GAAG;IACnC,WAAW,CAAC,EAAE,IAAI,GAAG,GAAG,GAAG,GAAG,GAAG,GAAG,GAAG,GAAG,CAAC;CAC5C,CAAC;AAEF,wBAAgB,eAAe,CAC7B,GAAG,EAAE,MAAM,EACX,IAAI,CAAC,EAAE,sBAAsB,GAC5B,MAAM,CAmCR"}

package/packages/agent/src/cli/parse-duration.js

@@ -1,27 +0,0 @@
-export function parseDurationMs(raw, opts) {
-    const trimmed = String(raw ?? "")
-        .trim()
-        .toLowerCase();
-    if (!trimmed) {
-        throw new Error("invalid duration (empty)");
-    }
-    const m = /^(\d+(?:\.\d+)?)(ms|s|m|h|d)?$/.exec(trimmed);
-    if (!m) {
-        throw new Error(`invalid duration: ${raw}`);
-    }
-    const value = Number(m[1]);
-    if (!Number.isFinite(value) || value < 0) {
-        throw new Error(`invalid duration: ${raw}`);
-    }
-    const unit = (m[2] ?? opts?.defaultUnit ?? "ms");
-    const multiplier = unit === "ms"
-        ? 1
-        : unit === "s"
-            ? 1000
-            : unit === "m"
-                ? 60_000
-                : unit === "h"
-                    ? 3_600_000
-                    : 86_400_000;
-    return Math.round(value * multiplier);
-}

package/packages/agent/src/config/character-schema.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"character-schema.d.ts","sourceRoot":"","sources":["../../../../../src/config/character-schema.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AA+BxB,eAAO,MAAM,eAAe;;;;;;;;;;;;;;;;;;;;;;kBAYjB,CAAC"}

package/packages/agent/src/config/config.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"config.d.ts","sourceRoot":"","sources":["../../../../../src/config/config.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,iBAAiB,CAAC;AAqBnD,YAAY,EAAE,WAAW,EAAE,MAAM,iBAAiB,CAAC;AAyEnD,wBAAgB,eAAe,IAAI,WAAW,CAwH7C;AAwDD,wBAAgB,eAAe,CAAC,MAAM,EAAE,WAAW,GAAG,IAAI,CA8EzD;AAED,wBAAgB,gBAAgB,IAAI,OAAO,CAQ1C"}