hub-gif f06fd12f1f fix(llm): 策略稿润色预留 completion 预算，缓解输出截断 ...

generate_strategy_draft_markdown_llm 按摘要/第九章节选/规则底稿分级收紧，满足最小 completion token 后再调用；底稿与 JSON 超限仍截断并提示。系统提示补充收束全文要求。

Made-with: Cursor

2026-04-17 16:48:05 +08:00

..

__init__.py

feat(pipeline): JD 流水线 CSV 纯中文表头与模块目录重构

2026-04-15 16:02:29 +08:00

generate_competitor_full.py

refactor(pipeline): 拆分 llm/generate 为网关与各场景子模块

2026-04-17 11:24:43 +08:00

generate_group_summaries.py

refactor(pipeline): 拆分 llm/generate 为网关与各场景子模块

2026-04-17 11:24:43 +08:00

generate_sections.py

refactor(pipeline): 拆分 llm/generate 为网关与各场景子模块

2026-04-17 11:24:43 +08:00

generate_strategy.py

fix(llm): 策略稿润色预留 completion 预算，缓解输出截断

2026-04-17 16:48:05 +08:00

generate.py

refactor(pipeline): 拆分 llm/generate 为网关与各场景子模块

2026-04-17 11:24:43 +08:00

keyword_suggest.py

feat(pipeline): JD 流水线 CSV 纯中文表头与模块目录重构

2026-04-15 16:02:29 +08:00

llm_client.py

refactor(pipeline): 拆分 llm/generate 为网关与各场景子模块

2026-04-17 11:24:43 +08:00