Custom LLM connection with Text Generator plugin

[g.raby_1822]

Troubleshooting Custom LLM Connection with Obsidian’s AI Plugin (LM Studio + MythoMax)

Hi everyone,

I’m reaching out after several days of trying to get a custom local LLM integration working with Obsidian’s AI Plugin. My end goal is to use a local model (MythoMax 13B, running in LM Studio) as a conversational assistant within Obsidian, using the plugin’s custom endpoint support.

---

My Setup:

• Model: mythomax-13b (running locally in LM Studio, latest version)
• Endpoint: http://127.0.0.1:1234/v1/chat/completions
• Obsidian Plugin: Obsidian AI Plugin (installed and toggled on)
• Plugin Config:
  • Endpoint set to Custom
  • Model name copied exactly
  • No API key required (left blank)
  • Tried multiple JSON body/header formats

---

Symptoms:

• Occasionally I get TG error.I or JSON5 formatting errors, even with seemingly valid body JSON
• Other times, no error is thrown—but I receive no response from the model (just spinning dots)
• I’ve tried:
  • Resetting plugin config
  • Restarting Obsidian and LM Studio
  • Tweaking body/header values in different combinations

---

My Questions:

1. Has anyone successfully connected LM Studio (or a similar local LLM) to the Obsidian AI Plugin?
2. Is there a known working body format or plugin configuration that works reliably?
3. Is there something I’m missing—perhaps in Advanced Mode, headers, or model registration?

---

Any insights, working examples, screenshots, or help would be greatly appreciated. This plugin is a big part of my creative workflow, so I’d love to get it working with my local model.

Thanks in advance!
– Glenn

[CawlinTeffid]

@g.raby_1822

Obsidian Plugin: Obsidian AI Plugin

There is no plugin named “AI” (the word “Obsidian” is removed from Obsidian plugin names). What is the actual name of the plugin you’re using?

[g.raby_1822]

Text Generator

[CawlinTeffid]

I updated the post title to use the plugin’s name.