Skip to main content
Glama

Convex MCP server

Official
by get-convex
logs.ts7.24 kB
import { Context } from "../../bundler/context.js"; import { logMessage, logOutput, logWarning } from "../../bundler/log.js"; import { nextBackoff } from "./dev.js"; import chalk from "chalk"; import { deploymentFetch } from "./utils/utils.js"; export type LogMode = "always" | "pause-on-deploy" | "disable"; export class LogManager { private paused: boolean = false; constructor(private mode: LogMode) {} async waitForUnpaused() { while (this.paused) { await new Promise((resolve) => setTimeout(resolve, 100)); } } beginDeploy() { if (this.mode === "pause-on-deploy") { this.paused = true; } } endDeploy() { if (this.mode === "pause-on-deploy") { this.paused = false; } } } const MAX_UDF_STREAM_FAILURE_COUNT = 5; type LogDestination = "stdout" | "stderr"; export async function logsForDeployment( ctx: Context, credentials: { url: string; adminKey: string; }, options: { success: boolean; history: number; deploymentNotice: string; }, ) { logMessage(chalk.yellow(`Watching logs${options.deploymentNotice}...`)); await watchLogs(ctx, credentials.url, credentials.adminKey, "stdout", { history: options.history, success: options.success, }); } export async function watchLogs( ctx: Context, url: string, adminKey: string, dest: LogDestination, options?: { success: boolean; history?: number | boolean; logManager?: LogManager; }, ) { let numFailures = 0; let isFirst = true; let cursorMs = 0; for (;;) { try { const { entries, newCursor } = await pollUdfLog( ctx, cursorMs, url, adminKey, ); cursorMs = newCursor; numFailures = 0; // Delay printing logs until the log manager is unpaused. await options?.logManager?.waitForUnpaused(); // The first execution, we just want to fetch the current head cursor so we don't send stale // logs to the client. if (isFirst) { isFirst = false; if ( options?.history === true || (typeof options?.history === "number" && options?.history > 0) ) { const entriesSlice = options?.history === true ? entries : entries.slice(entries.length - options?.history); processLogs(ctx, entriesSlice, dest, options?.success); } } else { processLogs(ctx, entries, dest, options?.success === true); } } catch { numFailures += 1; } // Handle backoff if (numFailures > 0) { const backoff = nextBackoff(numFailures); // If we exceed a threshold number of failures, warn the user and display backoff. if (numFailures > MAX_UDF_STREAM_FAILURE_COUNT) { logWarning( `Convex [WARN] Failed to fetch logs. Waiting ${backoff}ms before next retry.`, ); } await new Promise((resolve) => { setTimeout(() => resolve(null), backoff); }); } } } type UdfType = "Query" | "Mutation" | "Action" | "HttpAction"; type StructuredLogLine = { messages: string[]; level: "LOG" | "DEBUG" | "INFO" | "WARN" | "ERROR"; timestamp: number; isTruncated: boolean; }; type LogLine = string | StructuredLogLine; type UdfExecutionResponse = { identifier: string; udfType: UdfType; logLines: LogLine[]; // Unix timestamp (in seconds) timestamp: number; // UDF execution duration (in seconds) executionTime: number; error: string | null; kind: "Completion" | "Progress"; }; async function pollUdfLog( ctx: Context, cursor: number, url: string, adminKey: string, ): Promise<{ entries: UdfExecutionResponse[]; newCursor: number }> { const fetch = deploymentFetch(ctx, { deploymentUrl: url, adminKey, }); const response = await fetch(`/api/stream_function_logs?cursor=${cursor}`, { method: "GET", }); return await response.json(); } const prefixForSource = (udfType: UdfType): string => { return udfType.charAt(0); }; function processLogs( ctx: Context, rawLogs: UdfExecutionResponse[], dest: LogDestination, shouldShowSuccessLogs: boolean, ) { for (let i = 0; i < rawLogs.length; i++) { const log = rawLogs[i]; if (log.logLines) { const id = log.identifier; const udfType = log.udfType; const timestampMs = log.timestamp * 1000; const executionTimeMs = log.executionTime * 1000; for (let j = 0; j < log.logLines.length; j++) { logToTerminal( ctx, "info", timestampMs, udfType, id, log.logLines[j], dest, ); } if (log.error) { logToTerminal(ctx, "error", timestampMs, udfType, id, log.error!, dest); } else if (log.kind === "Completion" && shouldShowSuccessLogs) { logFunctionExecution( ctx, timestampMs, log.udfType, id, executionTimeMs, dest, ); } } } } function logFunctionExecution( ctx: Context, timestampMs: number, udfType: UdfType, udfPath: string, executionTimeMs: number, dest: LogDestination, ) { logToDestination( ctx, dest, chalk.green( `${prefixLog( timestampMs, udfType, udfPath, )} Function executed in ${Math.ceil(executionTimeMs)} ms`, ), ); } function logToTerminal( ctx: Context, type: "info" | "error", timestampMs: number, udfType: UdfType, udfPath: string, message: LogLine, dest: LogDestination, ) { const prefix = prefixForSource(udfType); if (typeof message === "string") { if (type === "info") { const match = message.match(/^\[.*?\] /); if (match === null) { logToDestination( ctx, dest, chalk.red( `[CONVEX ${prefix}(${udfPath})] Could not parse console.log`, ), ); return; } const level = message.slice(1, match[0].length - 2); const args = message.slice(match[0].length); logToDestination( ctx, dest, chalk.cyan(`${prefixLog(timestampMs, udfType, udfPath)} [${level}]`), args, ); } else { logToDestination( ctx, dest, chalk.red(`${prefixLog(timestampMs, udfType, udfPath)} ${message}`), ); } } else { const level = message.level; const formattedMessage = `${message.messages.join(" ")}${message.isTruncated ? " (truncated due to length)" : ""}`; logToDestination( ctx, dest, chalk.cyan( // timestamp is in ms since epoch `${prefixLog(message.timestamp, udfType, udfPath)} [${level}]`, ), formattedMessage, ); } } function logToDestination(ctx: Context, dest: LogDestination, ...logged: any) { switch (dest) { case "stdout": logOutput(...logged); break; case "stderr": logMessage(...logged); break; } } function prefixLog(timestampMs: number, udfType: UdfType, udfPath: string) { const prefix = prefixForSource(udfType); const localizedTimestamp = new Date(timestampMs).toLocaleString(); return `${localizedTimestamp} [CONVEX ${prefix}(${udfPath})]`; }

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/get-convex/convex-backend'

If you have feedback or need assistance with the MCP directory API, please join our Discord server