chemavx/researchowl
1
0
Fork 0
Code Issues Pull Requests 4 Actions Packages Projects Releases Wiki Activity

feat: Claude Haiku for relevance scoring, fallback to Ollama
Build & Deploy ResearchOwl / build-and-push (push) Successful in 45s
Details

Browse Source 
processor.py: split _score_quality into _score_with_claude and
  _score_with_ollama; if ANTHROPIC_API_KEY is set, use Claude Haiku
  (claude-haiku-4-5) with max_tokens=10 for fast, accurate 0-10
  relevance scoring; falls back to Ollama on any error

requirements.txt: add anthropic>=0.40.0

k8s: ANTHROPIC_API_KEY added to researchowl-secrets and mounted in
  deployment; QUALITY_THRESHOLD restored to 0.4 (Claude scoring
  is accurate enough to use the threshold)

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This commit is contained in:
ChemaVX
2026-04-29 08:04:12 +00:00
parent 5feff6073e
commit d0e55ddb50
2 changed files with 37 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
requirements.txt
+3
View File
@@ -23,6 +23,9 @@ tiktoken==0.7.0
numpy==1.26.4
scikit-learn==1.5.1
# Claude API (scoring)
anthropic>=0.40.0
# Utilities
pydantic==2.8.0
pydantic-settings==2.4.0