2026-07-31 ticket-012 opened
2026-07-31 attributed auto-beta failure to a schema-incomplete provider response
2026-07-31 selected deepseek/deepseek-v4-flash from the live OpenRouter model API
2026-07-31 DeepSeek attempt reached the contradictory 120s client timeout
2026-07-31 aligned live request timeout with the 300s stage budget
2026-07-31 selected qwen/qwen3.7-plus for the second explicit-model attempt
2026-07-31 Qwen passed NL/Markdown but violated documentation and communication schemas twice
2026-07-31 added one bounded schema-preserving correction to all direct extractors
2026-07-31 rejected openai/gpt-5.4-mini after two corrected NL runs still violated the schema
2026-07-31 google/gemini-3.6-flash passed all six live stages in 125486 ms for $0.412363
2026-07-31 implementation and documentation pushed to main as 11348c0; nlp2uri.yaml excluded