clucraft e8edc0c0e3 Fix Ollama Qwen3/DeepSeek thinking mode breaking JSON parsing ... These models output <think>...</think> blocks before their actual JSON response when in "thinking mode". This fix: 1. Adds /nothink prefix to all AI prompts (EXTRACTION_PROMPT, VERIFICATION_PROMPT, STOCK_STATUS_PROMPT, ARBITRATION_PROMPT) to request models disable thinking mode 2. Adds stripThinkingTags() helper function that removes <think> blocks from responses as a fallback if models ignore /nothink 3. Applies stripping in all parse functions before JSON extraction Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>		2026-01-25 11:23:32 -05:00
..
src	Fix Ollama Qwen3/DeepSeek thinking mode breaking JSON parsing	2026-01-25 11:23:32 -05:00
.dockerignore	Initial commit: PriceGhost price tracking application	2026-01-20 13:58:13 -05:00
.env.example	Initial commit: PriceGhost price tracking application	2026-01-20 13:58:13 -05:00
Dockerfile	Add Puppeteer fallback for Cloudflare-protected sites	2026-01-21 20:55:49 -05:00
package-lock.json	Fix stock status false positives for in-stock items	2026-01-23 13:56:26 -05:00
package.json	Fix stock status false positives for in-stock items	2026-01-23 13:56:26 -05:00
tsconfig.json	Initial commit: PriceGhost price tracking application	2026-01-20 13:58:13 -05:00