fix(evaluator): unwrap dimension_scores.item (llm_client.py)

fix(evaluator): unwrap dimension_scores.item (evaluator.py)

feat(runs): batch concurrency + question + delete (runs.py)

feat(runs): batch concurrency + question + delete (evaluator.py)

feat(runs): batch concurrency + question + delete (config.py)

feat(runs): batch concurrency + question + delete (schemas.py)

fix(evaluator): bump max_tokens 8k→16k + raise on truncated (evaluator.py)

fix(evaluator): bump max_tokens 8k→16k + raise on truncated (llm_client.py)

fix(evaluator): coerce list-vs-dict (retry 1, llm_client.py)

fix(evaluator): coerce list-vs-dict (evaluator.py)

docs: add agent.md (architecture overview + operational notes)

fix(evaluator): transpose parallel-array root_causes (evaluator.py)

fix(evaluator): transpose parallel-array root_causes (llm_client.py)

fix(docker): drop COPY 数据测试集/ line — data uploaded via UI

scan-disk: graceful no-op when 数据测试集/ not bundled

remove 数据测试集/ (per user — only ship platform code)

sync from GitHub bb1f476: fix(testsets): preserve `tags` field across all import paths

fix(skills): route order — /file and /tree before greedy catch-all

feat(skills): add /tree endpoint for Skill Detail 5-tab layout

refactor: rename expected_answer → agent_answer (evaluator.py)

refactor: rename expected_answer → agent_answer (testsets.py)

refactor: rename expected_answer → agent_answer (schemas.py)

refactor: rename expected_answer → agent_answer (models.py)

refactor: rename expected_answer → agent_answer (db.py)

fix(evaluator): sanitize LLM judge output

update persistence.py (v2 fresh deploy)

update config.py (v2 fresh deploy)

init: import all code from Fin-EvalOps (with cache-fix middleware)

initial commit