Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/dev2forge/bridgex/llms.txt

Use this file to discover all available pages before exploring further.

Bridgex integrates with large language model providers to generate natural-language descriptions of JPEG images during conversion. When you open a .jpg or .jpeg file with a configured LLM, the AI-generated description is appended to the Markdown output automatically — no extra steps required beyond setting up your API key once. This page explains exactly what the LLM integration does, which providers and models are supported, how to configure it, and where the settings are stored on disk.

​
What the LLM Integration Does

LLM integration in Bridgex is used exclusively for JPEG image files (.jpg / .jpeg). Opening any other file format — Excel, Word, PDF, HTML, CSV, PowerPoint, ZIP, or RSS/XML — does not trigger any LLM call, regardless of your configuration.
When you open a JPEG file with a valid LLM configuration active, the conversion pipeline in converter.rs works as follows:
  1. Markitdown-rs performs its standard JPEG conversion (typically minimal text output).
  2. If the result does not already contain a # Description: heading, Bridgex synchronously calls the configured LLM provider to generate a description of the image (the call blocks until a response is received or the request fails).
  3. The description is appended to the editor content under the heading # Description:.
The final Markdown in your editor looks like this: 
# Description:
A close-up photograph of a red maple leaf resting on wet pavement, 
with shallow depth of field and soft bokeh in the background.
If the JPEG conversion produces no content and the LLM call also fails (or LLM is not configured), Bridgex writes the following error into the editor instead: 
Conversion failed for JPEG image. Please check the image file and your LLM configuration.

​
Supported Providers and Models

Bridgex supports three LLM providers. Each has a set of suggested models available as quick-select buttons in the LLM Settings popup, but the model field accepts any free-text string — you can enter any model identifier your provider supports.
Provider key: openaiEnvironment variable set: OPENAI_API_KEYSuggested models:
  • gpt-4o
  • gpt-4o-mini
  • gpt-3.5-turbo
Any other OpenAI model string (e.g. gpt-4-turbo) can be entered manually in the model field.

​
Environment Variables

When a JPEG conversion with LLM is triggered, Bridgex sets the appropriate environment variable for the active provider at runtime from the saved configuration. You do not need to set these manually in your shell — Bridgex handles it:
ProviderEnvironment variable
OpenAIOPENAI_API_KEY
GeminiGEMINI_API_KEY
DeepseekDEEPSEEK_API_KEY
Only the variable for the currently selected provider is set. The other two are left unchanged.

​
How to Configure LLM Settings

1

Open LLM Settings

Press Ctrl+K or go to Help → LLM API Key in the menu bar. The LLM Settings popup opens with the API key input auto-focused.
2

Enter your API key

Type or paste your API key into the API Key text field. The placeholder text shows the names of the relevant environment variables (OPENAI_API_KEY, GEMINI_API_KEY, DEEPSEEK_API_KEY) as a reminder.
3

Select a provider

Click one of the three provider buttons: OpenAI, Gemini, or Deepseek. When you switch providers, the model field is automatically updated to the first suggested model for that provider if your current model string is not in that provider’s suggested list.
4

Choose or enter a model

Either click one of the suggested model quick-select buttons to fill the model field, or type a model name directly into the free-text Model input. The model field accepts any valid provider model string.
5

Save and close

Click Done. The configuration is saved to disk immediately and the popup closes. The status line at the bottom of the popup shows the currently active provider and model before you close.

​
Configuration File

Bridgex persists your LLM settings in a JSON file on disk. The file is read at startup and updated each time you click Done in the LLM Settings popup. File location: 
%APPDATA%\Bridgex\llm_settings.json
Typically resolves to C:\Users\<username>\AppData\Roaming\Bridgex\llm_settings.json.
The directory is created automatically if it does not exist. If neither APPDATA (Windows) nor HOME (Linux/macOS) environment variables are set, Bridgex falls back to writing llm_settings.json in the current working directory. File format: 
{
  "llm_api_key": "sk-...",
  "llm_client": "openai",
  "llm_model": "gpt-4o"
}
FieldTypeDescription
llm_api_keystringYour API key for the selected provider
llm_clientstringProvider identifier: "openai", "gemini", or "deepseek"
llm_modelstringModel name string, e.g. "gpt-4o" or "gemini-2.0-flash"
You can edit this file manually in a text editor if needed. Bridgex reads it fresh on every launch.

​
Error Handling

If the LLM call fails — because the API key is incorrect, the model name is invalid, the provider is unreachable, or there is no network connection — Bridgex appends the following message to the editor output instead of a description:
# Description unavailable: check API key, provider, model and network.
No dialog is shown. The error is visible directly in the editor so you can review and retry by opening the same file again after correcting your LLM settings.

​
LLM Scope Reminder

To be explicit: the LLM setting has no effect on any non-JPEG file. The convert_from_path() function checks the file extension and only enables LLM mode when:
  1. The extension is .jpg or .jpeg (.jpeg is normalised to .jpg internally).
  2. The llm_api_key field is non-empty after trimming whitespace.
  3. Both llm_client and llm_model are set and non-empty.
All three conditions must be true simultaneously for an LLM call to occur.

Build docs developers (and LLMs) love

Get started for freeTalk to us