Bug Description

When a GFM markdown table with a row-label column is rendered for Telegram, the row-label column header (e.g., "維度") is incorrectly included as a duplicate bullet item in the output.

Expected Behavior

Given a markdown table:

維度

• 核心賣點: 核心賣點

• MiniMax M2.7: 自我進化能力、Agent 協作

• Qwen3.5: 多模態原生統一架構

維度

• 核心賣點: 強項場景

• MiniMax M2.7: 編程 Agent、長時任務自主迭代

• Qwen3.5: 視覺理解、工具調用、多模態

The Telegram output should be:

核心賣點
• MiniMax M2.7: 自我進化能力、Agent 協作
• Qwen3.5: 多模態原生統一架構

強項場景
• MiniMax M2.7: 編程 Agent、長時任務自主迭代
• Qwen3.5: 視覺理解、工具調用、多模態

Actual Behavior

The row-label column header ("維度") is included as a spurious bullet:

核心賣點
• 維度: 核心賣點
• MiniMax M2.7: 自我進化能力、Agent 協作
• Qwen3.5: 多模態原生統一架構

Environment

• Hermes Agent version: latest (NousResearch/hermes-agent)
• Platform: Telegram
• File: gateway/platforms/telegram.py, function _render_table_block_for_telegram, lines ~174-205

Root Cause

The table has 4 columns: row-label + 3 data columns.

• headers = ['核心賣點', 'MiniMax M2.7', 'Qwen3.5'] (3 items — the row-label column has no header)
• First data row cells = ['維度', '核心賣點', 'MiniMax M2.7', '多模態原生統一架構'] (4 items)

Line 191 sets heading = '維度' from the first cell. Then the condition at line 196 evaluates len(headers) == len(cells) and headers[0] → 3 == 4 and True → False, so the else branch at line 201 executes zip(headers[1:], cells[1:]). This still pairs headers incorrectly because the first cell of cells[1:] ('核心賣點') gets paired with 'MiniMax M2.7'.

The core issue: the first cell (the row-label column containing "維度") is prepended as a bold heading AND its value is also included in the bullet zip. The row-label column is never correctly excluded from the bullet generation.