LLMs in AI features
feat~ai.llms~1
Provides the core connectivity and abstraction layer for interacting with various Large Language Model backends.
Needs: impl
User can select from different LLM providers
feat~ai.llms.providers~1
Different providers offer varying trade-offs between cost, performance, privacy, and reasoning capabilities.
Needs: req, impl
JabRef must support OpenAI LLM provider
req~ai.llms.providers.openai~1
OpenAI is a popular and widely used LLM provider.
Needs: impl
Covers:
- feat~ai.llms.providers~1
JabRef must support HuggingFace LLM provider
req~ai.llms.providers.huggingface~1
HuggingFace provides access to a wide variety of open-weight models and community contributions.
Needs: impl
Covers:
- feat~ai.llms.providers~1
JabRef must support Google Gemini LLM provider
req~ai.llms.providers.gemini~1
Google Gemini is a popular and widely used LLM provider.
Needs: impl
Covers:
- feat~ai.llms.providers~1
JabRef must support Mistral LLM provider
req~ai.llms.providers.mistral~1
Mistral is a popular LLM provider.
Needs: impl
Covers:
- feat~ai.llms.providers~1
User can connect to local and custom LLMs
feat~ai.llms.custom~1
Allows users to connect to self-hosted models or proxy services, ensuring data privacy and cost control.
Needs: req, impl
JabRef must support OpenAI-compatible LLM provider
req~ai.llms.custom.openai-compatible~1
Many local inference servers (e.g., vLLM, Ollama) use the OpenAI API schema, making this a universal connector for local AI.
Needs: impl
Covers:
- feat~ai.llms.custom~1
OpenAI-compatible LLM provider must allow configuring API base URL
req~ai.llms.custom.base-url~1
User needs to point the client to their specific local server address (e.g., localhost:8000) or a private enterprise proxy.
Needs: impl
Covers:
- feat~ai.llms.custom~1
AI preferences must allow testing connection to configured provider
req~ai.llms.test-connection~1
User can check provider, chat model, API key, and API base URL entered in the preferences with a minimal chat request before saving them.
Needs: impl
Covers:
- feat~ai.llms.custom~1
Base URL of LLM provider must be redacted in logs and dialogs
req~ai.llms.base-url-redacted~1
Base URL might contain secrets (username and password for connection), so they need to be removed in logs and dialogs.
Needs: impl, utest
Covers:
- feat~ai.llms.custom~1