[Bug]: Special Characters in a models .md file cause the lineage tool to fail

[craigchurch]

What happened?

AI tells me:

Perfect! I found it. At byte position 1901709, there's the byte sequence 0xC3 0x81, which is the UTF-8 encoding for the character "Á" (Latin Capital Letter A with Acute).

The context shows this is part of the text: "Area México, Área Brasil, Área Centroaméric..."

This appears to be in a description or documentation field that includes geographic regions with Portuguese/Spanish names containing accented characters. The word "Área" means "Area" in Portuguese/Spanish.

The problem: The dbt MCP server is trying to read the manifest.json file using Windows' default 'charmap' encoding (CP-1252) instead of UTF-8, so it can't decode these valid UTF-8 accented characters.

Let me search for where this text appears in your dbt project:

Found it! The special characters are in cdol_mdm_org_dim_init.md at line 433:

These are valid UTF-8 characters (Spanish/Portuguese accented letters and Japanese katakana/kanji), but the dbt MCP server is trying to read the compiled manifest.json using Windows' default charmap encoding instead of UTF-8.

Steps to Reproduce

Ask AI to get lineage for a model

Deployment

Remote MCP server

Environment

Windows 11
VSCode

Relevant log output

[VDFaller]

@craigchurch which lineage tool is it running? get_lineage or get_lineage_dev?