Problem Statement
The Open WebUI API provider in GhidrAssist currently supports only Ollama API endpoints. However, Open WebUI can connect to multiple LLM backends (such as vLLM, llama.cpp, and other OpenAI-compatible servers), but GhidrAssist cannot utilize these alternative backends through the Open WebUI provider.
Impact
Users running Open WebUI with non-Ollama backends (or without Ollama installed) find the Open WebUI API provider unusable and must resort to workarounds:
• Currently, users must switch to the OpenAI API provider as an alternative
• The OpenAI provider has a hardcoded embedding model that probably won’t match the user’s local setup
• Users are forced to create a dummy embedding model in Open WebUI as a workaround, which duplicates configuration and adds unnecessary overhead
Proposed Solution
Add support for Open WebUI’s OpenAI-compatible endpoints (/api/v1/* or equivalent) as an additional configuration option in the Open WebUI API provider. This would allow direct communication with any backend that Open WebUI supports without requiring Ollama.
Additional Enhancement
It would be beneficial to allow users to specify a custom embedding model in the configuration (either in the Open WebUI provider or globally), rather than relying on hardcoded defaults.
Problem Statement
The Open WebUI API provider in GhidrAssist currently supports only Ollama API endpoints. However, Open WebUI can connect to multiple LLM backends (such as vLLM, llama.cpp, and other OpenAI-compatible servers), but GhidrAssist cannot utilize these alternative backends through the Open WebUI provider.
Impact
Users running Open WebUI with non-Ollama backends (or without Ollama installed) find the Open WebUI API provider unusable and must resort to workarounds:
• Currently, users must switch to the OpenAI API provider as an alternative
• The OpenAI provider has a hardcoded embedding model that probably won’t match the user’s local setup
• Users are forced to create a dummy embedding model in Open WebUI as a workaround, which duplicates configuration and adds unnecessary overhead
Proposed Solution
Add support for Open WebUI’s OpenAI-compatible endpoints (
/api/v1/*or equivalent) as an additional configuration option in the Open WebUI API provider. This would allow direct communication with any backend that Open WebUI supports without requiring Ollama.Additional Enhancement
It would be beneficial to allow users to specify a custom embedding model in the configuration (either in the Open WebUI provider or globally), rather than relying on hardcoded defaults.