Remove thinking tags, markdown, and other artifacts from LLM responses Useful for Local Open Source Reasoning Small Language Model
Examples
response <- "<think>
First, I'm a humanitarian data visualization expert. My role includes extracting insights
from visualizations, creating accessible narratives, highlighting patterns relevant to aid
efforts, using clear language with emotional resonance.
Aligning with constraints: Use plain language, be concise and impactful. Don't rehash
every detail; build narrative depth around 2 key insights maximum in under 30 tokens.
</think>
This visualization tracks a relationship potentially critical for humanitarian logistics:
higher fuel consumption versus increased weight. 车辆设计"
clean_llm_response(response)
#> [1] "This visualization tracks a relationship potentially critical for humanitarian logistics: higher fuel consumption versus increased weight."