Different response engines for AI chat
feat~ai.response-engines~1
Description: response engine is an algorithm that supplies the context for LLM
Rationale: different response engines are suitable for different tasks
Needs: impl
Covers: feat~ai.chatting~1
Allow users to select a default response engine
req~ai.response-engines.default~1
Needs: impl
Covers: feat~ai.response-engines~1
“Embedding search” AI response engine
feat~ai.response-engines.embeddings-search~1
Rationale: this response engine is suitable when the user wants to perform a semantic search
Reference: https://arxiv.org/abs/2005.11401
Needs: impl, dsn
Covers: feat~ai.response-engines~1
Allow users to customize injection prompt for “embedding search” AI response engine
req~ai.response-engines.embeddings-search.prompt~1
Rationale: different prompts are suited for different tasks and affect the LLM output
Needs: impl
Covers: feat~ai.response-engines.embeddings-search~1, feat~ai.expert-settings~1
“Full document” AI response engine
feat~ai.response-engines.full-document~1
Rationale: this response engine is suitable when the user wants to get information that depends on the full content of a document
Needs: impl
Reference: https://arxiv.org/abs/2407.16833
Covers: feat~ai.response-engines~1
Allow users to customize injection prompt for “full document” AI response engine
req~ai.response-engines.full-document.prompt~1
Rationale: different prompts are suited for different tasks and affect the LLM output
Needs: impl
Covers: feat~ai.response-engines.full-document~1, feat~ai.expert-settings~1