rdf fix

app.py

@@ -306,42 +306,83 @@ def get_openai_client():
         timeout=120.0  # Increase timeout for cold starts
     )
 
-# Sample RDF data for examples
+# Sample RDF data for examples (based on real Library of Congress BibFrame)
 SAMPLE_VALID_RDF = '''<?xml version="1.0" encoding="UTF-8"?>
-<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
-         xmlns:bf="http://id.loc.gov/ontologies/bibframe/"
-         xmlns:rdfs="http://www.w3.org/2000/01/rdf-schema#">
+<rdf:RDF xmlns:bf="http://id.loc.gov/ontologies/bibframe/"
+         xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
+         xmlns:rdfs="http://www.w3.org/2000/01/rdf-schema#"
+         xmlns:bflc="http://id.loc.gov/ontologies/bflc/"
+         xmlns:madsrdf="http://www.loc.gov/mads/rdf/v1#">
     
     <bf:Work rdf:about="http://example.org/work/1">
         <rdf:type rdf:resource="http://id.loc.gov/ontologies/bibframe/Text"/>
+        <rdf:type rdf:resource="http://id.loc.gov/ontologies/bibframe/Monograph"/>
+        
         <bf:title>
             <bf:Title>
-                <bf:mainTitle>Sample Monograph Title</bf:mainTitle>
+                <bf:mainTitle>The knitter's handy book of patterns</bf:mainTitle>
+                <bf:subtitle>basic designs in multiple sizes & gauges</bf:subtitle>
             </bf:Title>
         </bf:title>
+        
         <bf:contribution>
-            <bf:Contribution>
+            <bf:PrimaryContribution>
                 <bf:agent>
-                    <bf:Agent>
-                        <rdfs:label>Sample Author</rdfs:label>
+                    <bf:Agent rdf:about="http://id.loc.gov/rwo/agents/n2001017606">
+                        <rdf:type rdf:resource="http://id.loc.gov/ontologies/bibframe/Person"/>
+                        <rdfs:label>Budd, Ann, 1956-</rdfs:label>
                     </bf:Agent>
                 </bf:agent>
                 <bf:role>
-                    <bf:Role>
-                        <rdfs:label>Author</rdfs:label>
+                    <bf:Role rdf:about="http://id.loc.gov/vocabulary/relators/aut">
+                        <rdfs:label>author</rdfs:label>
+                        <bf:code>aut</bf:code>
                     </bf:Role>
                 </bf:role>
-            </bf:Contribution>
+            </bf:PrimaryContribution>
         </bf:contribution>
-        <bf:language rdf:resource="http://id.loc.gov/vocabulary/languages/eng"/>
-        <bf:content rdf:resource="http://id.loc.gov/vocabulary/contentTypes/txt"/>
-        <bf:adminMetadata>
-            <bf:AdminMetadata>
+        
+        <bf:language>
+            <bf:Language rdf:about="http://id.loc.gov/vocabulary/languages/eng">
+                <rdfs:label xml:lang="en">English</rdfs:label>
+                <bf:code rdf:datatype="http://www.w3.org/2001/XMLSchema#string">eng</bf:code>
+            </bf:Language>
+        </bf:language>
+        
+        <bf:content>
+            <bf:Content rdf:about="http://id.loc.gov/vocabulary/contentTypes/txt">
+                <rdfs:label>text</rdfs:label>
+                <bf:code>txt</bf:code>
+            </bf:Content>
+        </bf:content>
+        
+        <bf:classification>
+            <bf:ClassificationLcc>
+                <bf:classificationPortion>TT820</bf:classificationPortion>
+                <bf:itemPortion>.B877 2002</bf:itemPortion>
                 <bf:assigner>
-                    <bf:Organization>
-                        <rdfs:label>Example Library</rdfs:label>
+                    <bf:Organization rdf:about="http://id.loc.gov/vocabulary/organizations/dlc">
+                        <rdfs:label>United States, Library of Congress</rdfs:label>
                     </bf:Organization>
                 </bf:assigner>
+            </bf:ClassificationLcc>
+        </bf:classification>
+        
+        <bf:adminMetadata>
+            <bf:AdminMetadata>
+                <bf:status>
+                    <bf:Status rdf:about="http://id.loc.gov/vocabulary/mstatus/n">
+                        <rdfs:label>new</rdfs:label>
+                        <bf:code>n</bf:code>
+                    </bf:Status>
+                </bf:status>
+                <bf:date rdf:datatype="http://www.w3.org/2001/XMLSchema#date">2001-12-12</bf:date>
+                <bf:agent>
+                    <bf:Agent rdf:about="http://id.loc.gov/vocabulary/organizations/dlc">
+                        <rdf:type rdf:resource="http://id.loc.gov/ontologies/bibframe/Organization"/>
+                        <rdfs:label>United States, Library of Congress</rdfs:label>
+                    </bf:Agent>
+                </bf:agent>
             </bf:AdminMetadata>
         </bf:adminMetadata>
     </bf:Work>
@@ -360,6 +401,127 @@ SAMPLE_INVALID_RDF = '''<?xml version="1.0" encoding="UTF-8"?>
     </bf:Work>
 </rdf:RDF>'''
 
+# BibFrame Few-Shot Examples (based on real Library of Congress records)
+BIBFRAME_CORRECTION_EXAMPLES = {
+    "title_structure": {
+        "pattern": r"bf:title",
+        "wrong": """<bf:title>Simple Title String</bf:title>""",
+        "correct": """<bf:title>
+    <bf:Title>
+        <bf:mainTitle>The knitter's handy book of patterns</bf:mainTitle>
+        <bf:subtitle>basic designs in multiple sizes & gauges</bf:subtitle>
+    </bf:Title>
+</bf:title>"""
+    },
+    "adminmetadata": {
+        "pattern": r"bf:adminMetadata|->bf:assigner",
+        "wrong": """<bf:adminMetadata>
+    <bf:AdminMetadata>
+        <bf:agent rdf:resource="http://example.org/org"/>
+        <bf:status>new</bf:status>
+    </bf:AdminMetadata>
+</bf:adminMetadata>""",
+        "correct": """<bf:adminMetadata>
+    <bf:AdminMetadata>
+        <bf:status>
+            <bf:Status rdf:about="http://id.loc.gov/vocabulary/mstatus/n">
+                <rdfs:label>new</rdfs:label>
+                <bf:code>n</bf:code>
+            </bf:Status>
+        </bf:status>
+        <bf:date rdf:datatype="http://www.w3.org/2001/XMLSchema#date">2001-12-12</bf:date>
+        <bf:agent>
+            <bf:Agent rdf:about="http://id.loc.gov/vocabulary/organizations/dlc">
+                <rdf:type rdf:resource="http://id.loc.gov/ontologies/bibframe/Organization"/>
+                <rdfs:label>United States, Library of Congress</rdfs:label>
+            </bf:Agent>
+        </bf:agent>
+    </bf:AdminMetadata>
+</bf:adminMetadata>"""
+    },
+    "contribution": {
+        "pattern": r"bf:contribution",
+        "wrong": """<bf:contribution>Author Name</bf:contribution>""",
+        "correct": """<bf:contribution>
+    <bf:PrimaryContribution>
+        <bf:agent>
+            <bf:Agent rdf:about="http://id.loc.gov/rwo/agents/n2001017606">
+                <rdf:type rdf:resource="http://id.loc.gov/ontologies/bibframe/Person"/>
+                <rdfs:label>Budd, Ann, 1956-</rdfs:label>
+            </bf:Agent>
+        </bf:agent>
+        <bf:role>
+            <bf:Role rdf:about="http://id.loc.gov/vocabulary/relators/ctb">
+                <rdfs:label>contributor</rdfs:label>
+                <bf:code>ctb</bf:code>
+            </bf:Role>
+        </bf:role>
+    </bf:PrimaryContribution>
+</bf:contribution>"""
+    },
+    "language": {
+        "pattern": r"bf:language",
+        "wrong": """<bf:language>English</bf:language>""",
+        "correct": """<bf:language>
+    <bf:Language rdf:about="http://id.loc.gov/vocabulary/languages/eng">
+        <rdfs:label xml:lang="en">English</rdfs:label>
+        <bf:code rdf:datatype="http://www.w3.org/2001/XMLSchema#string">eng</bf:code>
+    </bf:Language>
+</bf:language>"""
+    },
+    "content": {
+        "pattern": r"bf:content",
+        "wrong": """<bf:content>Text</bf:content>""",
+        "correct": """<bf:content>
+    <bf:Content rdf:about="http://id.loc.gov/vocabulary/contentTypes/txt">
+        <rdfs:label>text</rdfs:label>
+        <bf:code>txt</bf:code>
+    </bf:Content>
+</bf:content>"""
+    },
+    "classification": {
+        "pattern": r"bf:classification",
+        "wrong": """<bf:classification>TT820 .B877 2002</bf:classification>""",
+        "correct": """<bf:classification>
+    <bf:ClassificationLcc>
+        <bf:classificationPortion>TT820</bf:classificationPortion>
+        <bf:itemPortion>.B877 2002</bf:itemPortion>
+        <bf:assigner>
+            <bf:Organization rdf:about="http://id.loc.gov/vocabulary/organizations/dlc">
+                <rdfs:label>United States, Library of Congress</rdfs:label>
+                <bf:code rdf:datatype="http://id.loc.gov/datatypes/orgs/code">DLC</bf:code>
+            </bf:Organization>
+        </bf:assigner>
+        <bf:status>
+            <bf:Status rdf:about="http://id.loc.gov/vocabulary/mstatus/uba">
+                <rdfs:label>used by assigner</rdfs:label>
+                <bf:code>uba</bf:code>
+            </bf:Status>
+        </bf:status>
+    </bf:ClassificationLcc>
+</bf:classification>"""
+    },
+    "subject": {
+        "pattern": r"bf:subject",
+        "wrong": """<bf:subject>Knitting--Patterns</bf:subject>""",
+        "correct": """<bf:subject>
+    <bf:Topic rdf:about="http://id.loc.gov/authorities/subjects/sh85072708">
+        <rdfs:label xml:lang="en">Knitting--Patterns</rdfs:label>
+        <madsrdf:componentList rdf:parseType="Collection">
+            <madsrdf:Authority>
+                <madsrdf:authoritativeLabel xml:lang="en">Knitting</madsrdf:authoritativeLabel>
+                <madsrdf:elementList>
+                    <madsrdf:TopicElement>
+                        <madsrdf:elementValue xml:lang="en">Knitting</madsrdf:elementValue>
+                    </madsrdf:TopicElement>
+                </madsrdf:elementList>
+            </madsrdf:Authority>
+        </madsrdf:componentList>
+    </bf:Topic>
+</bf:subject>"""
+    }
+}
+
 # MCP Server Tools (can be used independently)
 def validate_rdf_tool(rdf_content: str, template: str = "monograph") -> dict:
     """
@@ -838,6 +1000,58 @@ def extract_rdf_from_response(response: str) -> str:
     # If no code blocks found, return the response as-is
     return response
 
+def fix_common_rdf_errors(rdf_xml: str) -> str:
+    """
+    Fix common RDF/XML errors that AI models generate.
+    
+    Args:
+        rdf_xml (str): RDF/XML that may contain common errors
+        
+    Returns:
+        str: Fixed RDF/XML
+    """
+    import re
+    
+    # Remove any rdf:parseType attributes (common AI mistake)
+    rdf_xml = re.sub(r'\s+rdf:parseType="[^"]*"', '', rdf_xml)
+    
+    # Fix bf:title if it's just a string (should be nested structure)
+    rdf_xml = re.sub(
+        r'<bf:title>([^<]+)</bf:title>',
+        r'<bf:title><bf:Title><bf:mainTitle>\1</bf:mainTitle></bf:Title></bf:title>',
+        rdf_xml
+    )
+    
+    # Fix bf:language if it's a string instead of URI
+    language_map = {
+        'English': 'http://id.loc.gov/vocabulary/languages/eng',
+        'eng': 'http://id.loc.gov/vocabulary/languages/eng',
+        'Spanish': 'http://id.loc.gov/vocabulary/languages/spa',
+        'French': 'http://id.loc.gov/vocabulary/languages/fre',
+    }
+    for lang_text, lang_uri in language_map.items():
+        rdf_xml = re.sub(
+            f'<bf:language>{lang_text}</bf:language>',
+            f'<bf:language rdf:resource="{lang_uri}"/>',
+            rdf_xml,
+            flags=re.IGNORECASE
+        )
+    
+    # Fix bf:content if it's a string
+    content_map = {
+        'Text': 'http://id.loc.gov/vocabulary/contentTypes/txt',
+        'text': 'http://id.loc.gov/vocabulary/contentTypes/txt',
+    }
+    for content_text, content_uri in content_map.items():
+        rdf_xml = re.sub(
+            f'<bf:content>{content_text}</bf:content>',
+            f'<bf:content rdf:resource="{content_uri}"/>',
+            rdf_xml,
+            flags=re.IGNORECASE
+        )
+    
+    return rdf_xml
+
 def get_ai_correction(validation_results: str, rdf_content: str, template: str = 'monograph', max_attempts: int = None, include_warnings: bool = False, enable_validation_loop: bool | None = None, steps_log: Optional[List[str]] = None) -> str:
     """
     Generate AI-powered corrected RDF/XML based on validation errors.
@@ -950,6 +1164,39 @@ Rules:
 - Else if <bf:descriptionModifier rdf:resource=\"...\"/> exists, add <bf:assigner rdf:resource=\"...\"/> with the SAME URI.
 - Else if a <bf:identifiedBy> block contains <bf:assigner rdf:resource=\"...\"/>, copy that URI to a TOP-LEVEL <bf:assigner>.
 Keep all existing content; only add missing <bf:assigner> where required.
+"""
+
+            # Build few-shot examples based on the errors found
+            examples_to_include = []
+            validation_lower = validation_results.lower()
+            
+            # Check each example pattern against validation results
+            for name, example in BIBFRAME_CORRECTION_EXAMPLES.items():
+                pattern = example.get("pattern", name)
+                if re.search(pattern, validation_results, re.IGNORECASE):
+                    examples_to_include.append((name, example))
+                    if steps_log is not None:
+                        steps_log.append(f"Including {name} example based on pattern match")
+            
+            few_shot_section = ""
+            if examples_to_include:
+                few_shot_section = "\n\nCORRECT BIBFRAME PATTERNS (from Library of Congress records):\n"
+                few_shot_section += "NEVER use simple strings - always use nested structures as shown below:\n\n"
+                for name, example in examples_to_include:
+                    few_shot_section += f"{name.upper()}:\n"
+                    few_shot_section += f"❌ WRONG:\n```xml\n{example['wrong']}\n```\n"
+                    few_shot_section += f"✅ CORRECT:\n```xml\n{example['correct']}\n```\n\n"
+
+            # Add critical rules based on real patterns
+            critical_rules = """
+CRITICAL RDF/XML RULES (from real BibFrame):
+1. NEVER use rdf:parseType except for "Collection" on madsrdf:componentList
+2. Properties like bf:title, bf:language, bf:content MUST have nested typed resources
+3. Use rdf:about for resource URIs, not rdf:resource on the property element
+4. bf:adminMetadata can appear multiple times in one record
+5. Status, Role, Language etc. are OBJECTS with rdf:about URIs, not literals
+6. Date values use rdf:datatype for typing (e.g., xsd:date, xsd:dateTime)
+7. Every bf:AdminMetadata needs BOTH bf:agent AND bf:assigner if validation requires it
 """
 
             prompt = f"""You are an expert in RDF/XML and BibFrame cataloging. Fix the following RDF/XML based on the validation errors and official BibFrame documentation.
@@ -957,6 +1204,8 @@ Keep all existing content; only add missing <bf:assigner> where required.
 {severity_instruction}
 {admin_guidance}
 {guidance_section}
+{critical_rules}
+{few_shot_section}
 
 Validation Errors:
 {validation_results}
@@ -966,21 +1215,30 @@ Original RDF/XML:
 
 {f"Previous attempt {attempt} still had validation errors. Please fix ALL issues this time." if attempt > 0 else ""}
 
-Please provide the corrected RDF/XML that addresses all validation issues.
-- Follow the BibFrame specifications above
-- Return only the corrected XML without additional explanation
-- Maintain the original structure while fixing errors
-- Ensure all namespace declarations are present
-- Add any missing required properties per BibFrame specs
-- Fix any syntax or structural issues"""
+INSTRUCTIONS:
+1. Return ONLY valid RDF/XML - no explanations
+2. Follow the EXACT patterns shown in the examples above
+3. Use proper nested structures - NO simple string values for complex properties
+4. Keep ALL namespace declarations
+5. Fix ALL validation errors"""
             
             try:
+                # Update system prompt to be even more explicit
+                system_prompt = """You are an RDF/XML expert following Library of Congress BibFrame patterns.
+Output ONLY valid RDF/XML following these rules:
+- Start with <?xml version="1.0" encoding="UTF-8"?>
+- NO markdown, NO explanations
+- Use EXACT structure patterns from the examples
+- Complex properties need nested typed resources
+- rdf:parseType ONLY for Collection on madsrdf:componentList
+- Status/Role/Language are OBJECTS with URIs, not strings"""
+
                 chat_completion = client.chat.completions.create(
                     model=HF_MODEL,
                     messages=[
                         {
                             "role": "system",
-                            "content": "Return only valid RDF/XML content. No prose, no markdown, no code fences, no explanations."
+                            "content": system_prompt
                         },
                         {
                             "role": "user", 
@@ -994,11 +1252,14 @@ Please provide the corrected RDF/XML that addresses all validation issues.
                 
                 corrected_rdf = chat_completion.choices[0].message.content.strip()
                 if steps_log is not None:
-                    steps_log.append(f"Attempt {attempt_no}: model responded; extracting XML block if present")
+                    steps_log.append(f"Attempt {attempt_no}: model responded; extracting and fixing common errors")
                 
                 # Extract RDF content if it's wrapped in code blocks
                 corrected_rdf = extract_rdf_from_response(corrected_rdf)
                 
+                # Fix common AI mistakes
+                corrected_rdf = fix_common_rdf_errors(corrected_rdf)
+                
                 # Only validate if we have the validator and haven't hit timeout
                 if VALIDATOR_AVAILABLE and (time.time() - start_time < timeout - 10):
                     try: