Repository navigation
说话人推测名按证据分高/中/低三档(后端) - #286
Merged
Merged
Conversation
added 4 commits
October 30, 2026 22:23
Agent-Executor: glm-pi Agent-Model: glm-5.3-flash Agent-Effort: high Dispatch-Id: dlg-20261030-141106-67ca32 Task-Id: VideoTranscriptAPI-20261030-05
Agent-Executor: glm-pi Agent-Model: glm-5.3-flash Agent-Effort: high Dispatch-Id: dlg-20261030-141106-67ca32 Task-Id: VideoTranscriptAPI-20261030-05
Agent-Executor: glm-pi Agent-Model: glm-5.3-flash Agent-Effort: high Dispatch-Id: dlg-20261030-141106-67ca32 Task-Id: VideoTranscriptAPI-20261030-05
Agent-Executor: glm-pi Agent-Model: glm-5.3-flash Agent-Effort: high Dispatch-Id: dlg-20261030-141106-67ca32 Task-Id: VideoTranscriptAPI-20261030-05
Required Gate v2 — 状态面板当前状态:pass · 可合并 当前裁决: PR 累计主审轮(影子):
Gate 历史(v1;来源为持久化
|
| Run | Attempt | Head | 状态 | PR 累计主审轮(影子) | 收件人动作 |
|---|---|---|---|---|---|
| 41075828585 | 1 | f356b92 |
pass |
1/5 · collecting |
可合并 |
历史行按 run_id + run_attempt 去重并只增不删;删除本评论后可由 gate_terminal 制品重建。
Advisory 未复核 · ocr-minimax这次没扫:评审额度已耗尽。这不是「扫过且干净」——本 PR 少了一层 advisory 扫描。 要不要管:不阻塞合并,无需处理。下个 run 会自动重跑。 查原因:→ job 日志 失败证据
status=budget_exhausted · verdict=unknown · reviewer=ocr-minimax · adapter_rc=20 · sig=OCR 失败链未走完,评审预算耗尽,未启动后续腿 |
Merged
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
背景
证据不够的说话人现在一律显示「说话人N」,参考价值很低。本 PR 改为按程序能核实的证据强弱给每位说话人分档,并修掉卡 C 上线后实测暴露的三个问题(生产只读重跑 10 个两人样本,仅 2 个认出名字)。
改动
新模块
llm/core/speaker_confidence_level.py:每个 meta 条目新增confidence_level(high/medium/low/none)、guess_name、basis,按证据定档:mapping 语义不变:低档推测名不进导出和总结。
提示词:title_credit / author_host 的 name 必须照抄原文(
@sama填sama),删去已过时的「必须另有对话佐证」。两人通路:
规则版本 7 → 8。
验证
uv run --frozen pytest tests/unit/test_speaker_confidence_level.py tests/unit/test_speaker_duo_inference.py tests/unit/i18n_llm/test_prompt_selection.py tests/unit/test_speaker_evidence_gate.py tests/unit/test_speaker_inferencer.py tests/unit/test_speaker_self_intro_verifier.py tests/unit/test_llm_schemas.py tests/unit/test_speaker_artifact.py退出码 0。🤖 Generated with Claude Code