Skip to content

fix(evaluation): Force tool choice for metrics #1203

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
AkhileshNegi wants to merge 1 commit into
base: main
Choose a base branch
from
Open

fix(evaluation): Force tool choice for metrics #1203

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 5 additions & 2 deletions backend/app/crud/evaluations/fast.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -505,10 +505,13 @@ def run_response_chunk(
)

# Ask OpenAI to return the file_search hits so knowledge_base can judge them.
# tool_choice stays at the model default (auto) — consistent with normal calls;
# a row where the model doesn't query the KB is scored N/A, not forced to search.
# tool_choice is forced rather than left at auto: a model that declines to search
# (gpt-4.1 usually does) returns no chunks, which drops the knowledge_base metric
# for every row and leaves the score blank run-wide. Forcing the call makes the
# metric measure groundedness given retrieval, not eagerness to retrieve.
if any(t.get("type") == "file_search" for t in base_params.get("tools", [])):
base_params["include"] = ["file_search_call.results"]
base_params["tool_choice"] = "required"

results: list[dict[str, Any]] = []
max_workers = max(
Expand Down
Loading