Have you looked into semantic prompt compression? We built TokenShrink Gateway specifically to handle this—it acts as an infra proxy that compresses your prompts while preserving meaning before they hit the API. It’s been helping devs cut their OpenAI/Anthropic spend by about 50-60% without needing to swap models.
TACiT
Posted on
Discussion: Dev.to Discussion: 'Struggling with escalating OpenAI API costs in production' | 0707-0615
For further actions, you may consider blocking this person and/or reporting abuse
Top comments (0)