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

Add local rule-based classifier with Claude Code headless fallback

Browse Source 
Replaces API-based classifier with:
1. Filename pattern matching (covers 95%+ of legal docs)
2. Content keyword matching for ambiguous filenames
3. Claude Code headless (claude -p) fallback for edge cases

No Anthropic API calls needed for classification.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
Chaim
2026-04-04 13:14:13 +00:00
parent 9e7492e761
commit 52ee3419d3
2 changed files with 118 additions and 30 deletions
Download Patch File Download Diff File Expand all files Collapse all files

43
mcp-server/src/legal_mcp/services/processor.py
View File

@@ -37,39 +37,22 @@ async def process_document(document_id: UUID, case_id: UUID) -> dict:
page_count=page_count,
)
# Step 1.5: Classify document and identify parties (non-fatal)
# Step 1.5: Classify document — local rules first, Claude Code headless fallback
classification_result = {}
try:
logger.info("Classifying document")
case_number = ""
if case_id:
case = await db.get_case(case_id)
if case:
case_number = case.get("case_number", "")
classification_result = await classifier.classify_and_identify(text, case_number)
await db.update_document(
document_id,
metadata=classification_result,
)
logger.info(
"Classification: %s (confidence: %.2f), parties found: %d appellants, %d respondents",
classification_result["classification"].get("doc_type", "?"),
classification_result["classification"].get("confidence", 0),
len(classification_result["parties"].get("appellants", [])),
len(classification_result["parties"].get("respondents", [])),
)
from legal_mcp.services import local_classifier
filename = Path(doc["file_path"]).name
doc_type, confidence = local_classifier.classify(filename, text)
if confidence < 0.8:
doc_type, confidence = local_classifier.classify_with_claude_code(filename, text)
# Update case parties if empty
if case_id and case:
parties = classification_result.get("parties", {})
updates = {}
if not case.get("appellants") and parties.get("appellants"):
updates["appellants"] = parties["appellants"]
if not case.get("respondents") and parties.get("respondents"):
updates["respondents"] = parties["respondents"]
if updates:
await db.update_case(case_id, **updates)
logger.info("Updated case parties: %s", updates)
# Update doc_type if we got a good classification and current type is generic
if confidence >= 0.5 and doc.get("doc_type") in ("reference", "auto"):
await db.update_document(document_id, doc_type=doc_type)
logger.info("Auto-classified: %s → %s (confidence %.2f)", filename, doc_type, confidence)
classification_result = {"classification": {"doc_type": doc_type, "confidence": confidence}}
await db.update_document(document_id, metadata=classification_result)
except Exception as e:
logger.warning("Classification failed (non-fatal): %s", e)