$megatools --few-shot-formatter --client-side
Few-Shot Prompt & Fine-Tuning Dataset Formatter
Construct, validate, and convert few-shot prompt examples into OpenAI JSONL, Anthropic XML, ChatML, and Llama 3 templates.
139 chars
Example #1
Example #2
Example #3
{"messages":[{"role":"system","content":"You are a precise sentiment analysis engine. Classify text into POSITIVE, NEGATIVE, or NEUTRAL with a confidence score and brief rationale."},{"role":"user","content":"The new UI update is incredibly responsive, but the dark mode colors feel a bit washed out."},{"role":"assistant","content":"SENTIMENT: MIXED/POSITIVE | Confidence: 0.85 | Rationale: Praises responsiveness but gives minor visual critique."}]}
{"messages":[{"role":"system","content":"You are a precise sentiment analysis engine. Classify text into POSITIVE, NEGATIVE, or NEUTRAL with a confidence score and brief rationale."},{"role":"user","content":"Customer support resolved my billing issue in under 5 minutes. Amazing service!"},{"role":"assistant","content":"SENTIMENT: POSITIVE | Confidence: 0.98 | Rationale: Strong positive sentiment regarding rapid issue resolution."}]}
{"messages":[{"role":"system","content":"You are a precise sentiment analysis engine. Classify text into POSITIVE, NEGATIVE, or NEUTRAL with a confidence score and brief rationale."},{"role":"user","content":"The application crashed twice during data export, causing complete work loss."},{"role":"assistant","content":"SENTIMENT: NEGATIVE | Confidence: 0.99 | Rationale: Critical software failure resulting in negative user experience."}]}