[metadata PR 2] Add live metadata recommendation agents #222
Conversation
Layer the LLM-backed reagent, condition, literature, and reaction-string recommendation workflow on top of the deterministic metadata foundation. Constraint: keep local tasks artifacts out of the PR split and keep this stacked follow-up below 1000 added lines. Rejected: moving the live Opus test to a special workflow | the requested PR behavior is to run it in the existing package test job. Confidence: high Scope-risk: moderate Directive: open this branch as a stacked PR targeting porting/metadata-foundation, not main, unless the foundation PR has already merged. Tested: uv run ruff check deepretro/metadata.py deepretro/metadata_types.py deepretro/types.py deepretro/utils/langfuse_config.py deepretro/tests/test_metadata.py; uv run --extra dev ty check metadata.py metadata_types.py types.py utils/langfuse_config.py tests/test_metadata.py; set -a; source /Users/shreyasv/Desktop/research/todel/DeepRetro/.env; set +a; uv run pytest tests/test_metadata.py -q; uv run --extra docs sphinx-build -b html ../docs/source ../docs/_build/html -W --keep-going
64d171b to
120ebe3
Compare
riya-singh28
left a comment
There was a problem hiding this comment.
@shreyasvinaya I have left a few comments after the initial review.
| ): | ||
| calls: list[str] = [] | ||
|
|
||
| def reagent_recommender( |
There was a problem hiding this comment.
@shreyasvinaya Can you please put these recommender functions as fixtures in conftest.py ?
|
|
||
|
|
||
| def test_recommend_reaction_metadata_returns_failure_status() -> None: | ||
| def reagent_recommender( |
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
There was a problem hiding this comment.
@shreyasvinaya @rbharath The changes look good to me. But CI is not running for the latest changes of this PR. @shreyasvinaya, can you please check?
|
@shreyasvinaya Also, why is there a Claude commit in this PR? |
|
For the record, it's ok to use Claude code but all code must be carefully checked and verified by hand before merge |
|
We need CI to run though @shreyasvinaya |
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
The ci had not run as the branch head was set to metadata pr 1 I've set the head branch to main and added a small comment to trigger CI |

Layer the LLM-backed reagent, condition, literature, and reaction-string recommendation workflow on top of the deterministic metadata foundation.
Type of change
Please check the option that is related to your PR.
Checklist