ezer-mishpati/legal-ai
1
0
Fork 0
Code Issues Pull Requests 2 Actions Packages Projects Releases Wiki Activity

Replace Haiku with Sonnet in classifier for better accuracy

Browse Source 
classify_document and identify_parties both used Haiku, which produced
parsing failures and 0% confidence on Beit HaKerem documents.
Sonnet handles Hebrew legal documents more reliably.

No more Haiku usage in the entire codebase.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
Chaim
2026-04-04 07:47:12 +00:00
parent 586f1db402
commit 081c7fb17a
1 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
mcp-server/src/legal_mcp/services/classifier.py
View File

@@ -98,8 +98,8 @@ async def classify_document(text: str) -> dict:
client = _get_anthropic() client = _get_anthropic()
message = client.messages.create( message = client.messages.create(
model="claude-haiku-4-5-20251001", model="claude-sonnet-4-20250514",
max_tokens=256, max_tokens=512,
messages=[ messages=[
{ {
"role": "user", "role": "user",
@@ -136,7 +136,7 @@ async def identify_parties(text: str) -> dict:
client = _get_anthropic() client = _get_anthropic()
message = client.messages.create( message = client.messages.create(
model="claude-haiku-4-5-20251001", model="claude-sonnet-4-20250514",
max_tokens=512, max_tokens=512,
messages=[ messages=[
{ {