You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
patent_filings_by_ticker tool — USPTO granted patents mapped to a single ticker by corporate assignee (patent ID, grant date, title, type, claims, citations, CPC classification, inventors, filing-to-grant time)
screener_patent_filings tool — cross-ticker patent filings screening with summary stats (total patents, total tickers, top CPC sections)
Integration tests for patent filings (ticker, screener, and tool-level round-trip)
Fixed
Screener summary block is now populated for screener_government_contracts (regression since 0.2.3) and screener_reddit_mentions (since 0.2.2) — both previously returned all-null summary fields because the SDK unwraps screener responses down to the rows and discards the summary. The adapter now fetches the full envelope so the backend-computed summary is preserved.
Changed
Bumped finbrain-python dependency from >=0.2.3 to >=0.2.5 (required for patent filings endpoints)
FBClient now caches the underlying finbrain-python client per API key, so repeated tool calls reuse a single HTTP session/connection pool instead of opening a new one each call
Strengthened screener integration tests to assert summary values (not just key presence), which is what previously masked the empty-summary bug
Internal cleanup: removed the unused _make_client helper from the server entrypoint, documented latest_slice's chronological-ordering assumption, and corrected README inaccuracies (a non-existent settings.py in the project tree and the API-key configuration note)