A gateway/proxy server for AI chat with built-in tools, RAG, skills, MCP client support, and sandbox execution. Provides a web UI and JSON API for chat sessions with tool calling, file attachments, and external service integration.
Significant gaps across naming, descriptions, and schemas. Tool names lack consistent verb prefixes (e.g., 'fetch_attachment' is good, but 'tls_cert', 'rag_grep', 'get_user_location' are inconsistent). Many tools have adequate descriptions (50-150 chars), but parameter descriptions are sparse or missing entirely. Input schemas are visible for most tools but lack depth: enums are present for some (e.g., browser_control.action, fetch_attachment.mode), but many parameters lack type constraints or validation guidance. Output schemas are not documented. Error handling is minimal, no recovery guidance or actionable error messages visible in source.
Control a headless browser for web automation, screenshot capture, and interaction
Execute ComfyUI workflows for image generation and processing. Each workflow is a separate tool.
Fetch and read file attachments (PDF, text, images, audio, etc.) from chat uploads or URLs
Generate QR codes with optional logo overlay and custom colors
Get the current user's stored location (city, country, coordinates) from GeoIP data
Tools exposed by connected MCP servers. Each MCP server's tools are prefixed with mcp_<server>_
Search RAG collections using regex patterns with linear-time matching
Search RAG collections using vector similarity (semantic search)
Dynamically generated tools (typst_<template>, comfyui_<workflow>, mcp_<server>_<tool>) lack explicit schema registration. Tool definitions are inferred from code patterns rather than directly visible in tool registration.
Parameter descriptions are sparse or missing. 'tls_cert' port parameter lacks description of default (443). 'get_user_location' has empty input schema, no parameters documented. 'browser_control' selector and text params lack format/constraint guidance.
Output schemas are not documented. No tool explicitly declares what fields it returns or their types. LLMs cannot plan downstream tool calls or extract required data.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | 2026-07-28+ | v2 |
Render Typst documents in a sandbox with full access to system fonts and custom packages
Inspect TLS certificates from a remote server without verification (including expired/self-signed)
Render documents using Typst templates (e.g., typst_letter, typst_invoice). Each template is a separate tool.
Naming inconsistency: 'tls_cert' should be 'get_tls_cert' or 'inspect_tls_cert' to start with action verb. 'rag_grep' and 'rag_search' lack clear verb prefixes.
Error handling is minimal. No visible recovery guidance, actionable error messages, or error classification (retryable vs. user-fixable vs. fatal).