Skip to content

samihalawa/grokbot

v1.3.3MIT

Universal Grok & Grok Bot orchestrator: multi-account pooling, automatic failover, and token-efficient export.

Changelog

All notable changes to grokbot will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[1.1.0] - 2026-10-29

Added

  • Visual Branding & Architecture Diagrams: High-resolution SVG banner and topology diagram in docs/ for enhanced GitHub presentation and developer clarity.
  • Parallel Batch Dispatch (grokbot batch): Concurrently dispatches prompt queues across healthy accounts in the pool with automatic rate-limit rebalancing.
  • Context & Specialized Knowledge Classifier: Auto-detects prompt intent (coding, legal/filing, scrapers, deepsearch) and routes to optimal account profiles automatically.
  • Real-Time Web Dashboard (grokbot serve): Dark-mode web interface at http://localhost:8000/ with live account cards, cooldown timers, request metrics, and interactive test playground.
  • Token Density Calculator (grokbot tokens): Calculates token counts and estimates savings for files or prompts against raw JSON baselines.

[1.0.0] - 2026-10-29

Added

  • Multi-Account Parallel Pool: Dynamic pooling across multiple Grok accounts with round-robin, least-busy, and specialized knowledge/role routing.
  • Automatic Quota Failover: Detects HTTP 429 and rate limits, places account in cooldown (15 min), and transparently routes to the next candidate with zero dropped requests.
  • 3-Layer Conversation Exporter: Unified exporter recovering transcripts from Desktop Persistent Slices (sand-client-persistence), Web REST conversations (grok.com/rest/app-chat), and gateway bots.
  • Token-Efficient Plaintext Export: Formats exported transcripts into clean, readable Markdown or text, reducing token overhead by 70-80% compared to verbose JSON.
  • Local OpenAI-Compatible Server: Built-in node:http proxy (/v1/chat/completions) enabling instant drop-in integration with Cursor, Claude Code, Codex, Aider, and OpenAI SDKs.
  • Skill Management & Pool Sync: Prevents prompt token bloat (re-injecting 50k+ characters per turn) by installing skills into account-level libraries and syncing across pools.
  • System Doctor: Built-in diagnostics for desktop storage, 1Password CLI session, token validity, and account connectivity.
  • Native Model Context Protocol (MCP): Built-in stdio MCP server for seamless agentic workflows.
  • Zero Runtime Dependencies: Pure native Node.js implementation with no external database or npm dependencies.