Ollama's API for handling conversational responses has a bug where it fails to retain conversation history when the `previous_response_id` parameter is used. Instead of continuing the dialogue, the API treats each request with this parameter as a new, isolated interaction, costing the same number of tokens as a fresh start but yielding incorrect or unrelated responses. This issue affects both local and potentially hosted models, leading to plausible but wrong answers or empty outputs when tool calls are involved, making it difficult to detect as the API does not return an error. AI
IMPACT This bug in Ollama's API prevents proper conversation history management, potentially leading to incorrect responses and hindering the development of reliable AI agents that rely on context.
RANK_REASON The cluster describes a bug in an API endpoint for a specific tool, Ollama, which impacts its conversational capabilities.
Read on Mastodon — sigmoid.social →
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →