vbagent 0.2.0__tar.gz → 0.2.1__tar.gz

{vbagent-0.2.0 → vbagent-0.2.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: vbagent
-Version: 0.2.0
+Version: 0.2.1
 Summary: Physics question processing library and CLI - classification, LaTeX extraction, variant generation
 Author: vaibhavblayer
 Author-email: vaibhavblayer@gmail.com

{vbagent-0.2.0 → vbagent-0.2.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "vbagent"
-version = "0.2.0"
+version = "0.2.1"
 description = "Physics question processing library and CLI - classification, LaTeX extraction, variant generation"
 authors = [
     {name = "vaibhavblayer", email = "vaibhavblayer@gmail.com"}

{vbagent-0.2.0 → vbagent-0.2.1}/vbagent/agents/base.py

@@ -8,7 +8,7 @@ from typing import TYPE_CHECKING, Any, Optional
 if TYPE_CHECKING:
     from agents import Agent, ModelSettings
 
-from vbagent.config import get_model, get_model_settings
+from vbagent.config import get_model, get_model_settings, apply_provider_config
 
 
 def _get_agent_class():
@@ -115,6 +115,9 @@ def create_agent(
     """
     Agent = _get_agent_class()
     
+    # Apply provider config (base_url, api_key) before creating agent
+    apply_provider_config()
+    
     # Get model and settings from config if not explicitly provided
     if model is None:
         model = get_model(agent_type or "default")

vbagent-0.2.1/vbagent/agents/classifier.py

@@ -0,0 +1,68 @@
+"""Classifier agent for question image classification.
+
+Uses openai-agents SDK to analyze question images and extract
+structured metadata including question type, difficulty, topic, etc.
+"""
+
+from typing import TYPE_CHECKING, Optional
+
+if TYPE_CHECKING:
+    from agents import Agent
+
+from vbagent.agents.base import (
+    create_agent,
+    create_image_message,
+    run_agent_sync,
+)
+from vbagent.config import get_config
+from vbagent.models.classification import ClassificationResult
+from vbagent.prompts.classifier import get_classifier_prompt, get_user_template
+
+
+def create_classifier_agent(subject: Optional[str] = None) -> "Agent":
+    """Create a classifier agent for a subject.
+    
+    Args:
+        subject: Subject override (uses config if not provided)
+        
+    Returns:
+        Configured Agent instance for classification
+    """
+    if subject is None:
+        subject = get_config().subject
+    
+    prompt = get_classifier_prompt(subject)
+    
+    return create_agent(
+        name=f"Classifier-{subject}",
+        instructions=prompt,
+        output_type=ClassificationResult,
+        agent_type="classifier",
+    )
+
+
+# Legacy: Create default classifier agent for backward compatibility
+classifier_agent = create_classifier_agent("physics")
+
+
+def classify(image_path: str, subject: Optional[str] = None) -> ClassificationResult:
+    """Analyze a question image and return structured metadata.
+    
+    Args:
+        image_path: Path to the image file to classify
+        subject: Subject override (uses config if not provided)
+        
+    Returns:
+        ClassificationResult with extracted metadata
+        
+    Raises:
+        FileNotFoundError: If the image file doesn't exist
+    """
+    if subject is None:
+        subject = get_config().subject
+    
+    agent = create_classifier_agent(subject)
+    user_template = get_user_template(subject)
+    message = create_image_message(image_path, user_template)
+    result = run_agent_sync(agent, message)
+    return result

vbagent-0.2.1/vbagent/agents/compile_fixer.py

@@ -0,0 +1,57 @@
+"""Agent for fixing LaTeX compilation errors.
+
+Takes a LaTeX snippet and pdflatex error output, returns corrected LaTeX.
+"""
+
+from vbagent.agents.base import create_agent, run_agent_sync
+
+
+SYSTEM_PROMPT = r"""You are a LaTeX error fixer. You receive LaTeX code that failed to compile and the pdflatex error output.
+
+Your job:
+1. Read the error messages carefully
+2. Fix ONLY the errors — do not change the content or structure
+3. Common fixes: missing braces, undefined commands, wrong environment names, missing $ delimiters
+4. Output ONLY the corrected LaTeX code — no explanations, no markdown, no code blocks
+
+CRITICAL: Output the EXACT same content with ONLY the compilation errors fixed. Do not add \documentclass, preamble, or any wrapping."""
+
+USER_TEMPLATE = """Fix the compilation errors in this LaTeX code.
+
+**Errors from pdflatex:**
+```
+{errors}
+```
+
+**LaTeX code to fix:**
+```latex
+{latex}
+```
+
+Output ONLY the corrected LaTeX code:"""
+
+
+def fix_latex(error_summary: str, latex: str) -> str:
+    """Send LaTeX + errors to agent and get fixed version.
+
+    Args:
+        error_summary: Parsed pdflatex error output
+        latex: The LaTeX code that failed to compile
+
+    Returns:
+        Corrected LaTeX code
+    """
+    agent = create_agent(
+        name="LaTeX-Fixer",
+        instructions=SYSTEM_PROMPT,
+        agent_type="converter",  # Use converter model (lighter/cheaper)
+    )
+
+    prompt = USER_TEMPLATE.format(errors=error_summary, latex=latex)
+    result = run_agent_sync(agent, prompt)
+
+    # Clean markdown artifacts
+    import re
+    result = re.sub(r'^```(?:latex|tex)?\s*\n?', '', result, flags=re.IGNORECASE)
+    result = re.sub(r'\n?```\s*$', '', result)
+    return result.strip()

{vbagent-0.2.0 → vbagent-0.2.1}/vbagent/agents/scanner.py

@@ -1,11 +1,11 @@
-"""Scanner agent for extracting LaTeX from physics question images.
+"""Scanner agent for extracting LaTeX from question images.
 
-Uses openai-agents SDK to analyze physics question images and extract
-LaTeX code using type-specific prompts.
+Uses openai-agents SDK to analyze question images and extract
+LaTeX code using type-specific and subject-specific prompts.
 """
 
 import re
-from typing import TYPE_CHECKING
+from typing import TYPE_CHECKING, Optional
 
 if TYPE_CHECKING:
     from agents import Agent
@@ -15,9 +15,10 @@ from vbagent.agents.base import (
     create_image_message,
     run_agent_sync,
 )
+from vbagent.config import get_config
 from vbagent.models.classification import ClassificationResult
 from vbagent.models.scan import ScanResult
-from vbagent.prompts.scanner import get_scanner_prompt, USER_TEMPLATE
+from vbagent.prompts.scanner import get_scanner_prompt, get_user_template
 from vbagent.references.context import get_context_prompt_section
 
 
@@ -51,17 +52,26 @@ def clean_latex_output(latex: str) -> str:
     return latex.strip()
 
 
-def create_scanner_agent(question_type: str, use_context: bool = True) -> "Agent":
+def create_scanner_agent(
+    question_type: str,
+    use_context: bool = True,
+    subject: Optional[str] = None,
+) -> "Agent":
     """Create a scanner agent with type-specific prompt.
     
     Args:
         question_type: The type of question (mcq_sc, mcq_mc, etc.)
         use_context: Whether to include reference context in prompt
+        subject: Subject override (uses config if not provided)
         
     Returns:
         Configured Agent instance for scanning that question type
     """
-    prompt = get_scanner_prompt(question_type)
+    # Get subject from config if not provided
+    if subject is None:
+        subject = get_config().subject
+    
+    prompt = get_scanner_prompt(question_type, subject)
     
     # Add reference context if enabled
     context = get_context_prompt_section("latex", use_context)
@@ -69,7 +79,7 @@ def create_scanner_agent(question_type: str, use_context: bool = True) -> "Agent
         prompt = prompt + "\n" + context
     
     return create_agent(
-        name=f"Scanner-{question_type}",
+        name=f"Scanner-{question_type}-{subject}",
         instructions=prompt,
         agent_type="scanner",
     )
@@ -79,8 +89,9 @@ def scan(
     image_path: str,
     classification: ClassificationResult,
     use_context: bool = True,
+    subject: Optional[str] = None,
 ) -> ScanResult:
-    """Extract LaTeX from a physics question image.
+    """Extract LaTeX from a question image.
     
     Uses the classification result to select the appropriate prompt
     for the question type.
@@ -89,6 +100,7 @@ def scan(
         image_path: Path to the image file to scan
         classification: Classification result with question type info
         use_context: Whether to include reference context in prompt
+        subject: Subject override (uses config if not provided)
         
     Returns:
         ScanResult with extracted LaTeX and diagram info
@@ -96,8 +108,13 @@ def scan(
     Raises:
         FileNotFoundError: If the image file doesn't exist
     """
-    agent = create_scanner_agent(classification.question_type, use_context)
-    message = create_image_message(image_path, USER_TEMPLATE)
+    # Get subject from config if not provided
+    if subject is None:
+        subject = get_config().subject
+    
+    agent = create_scanner_agent(classification.question_type, use_context, subject)
+    user_template = get_user_template(subject)
+    message = create_image_message(image_path, user_template)
     raw_latex = run_agent_sync(agent, message)
     
     # Clean up markdown artifacts from LLM output
@@ -114,8 +131,9 @@ def scan_with_type(
     image_path: str,
     question_type: str,
     use_context: bool = True,
+    subject: Optional[str] = None,
 ) -> ScanResult:
-    """Extract LaTeX from a physics question image with explicit type.
+    """Extract LaTeX from a question image with explicit type.
     
     Bypasses classification and uses the provided question type directly.
     
@@ -123,6 +141,7 @@ def scan_with_type(
         image_path: Path to the image file to scan
         question_type: The type of question (mcq_sc, mcq_mc, etc.)
         use_context: Whether to include reference context in prompt
+        subject: Subject override (uses config if not provided)
         
     Returns:
         ScanResult with extracted LaTeX
@@ -130,8 +149,13 @@ def scan_with_type(
     Raises:
         FileNotFoundError: If the image file doesn't exist
     """
-    agent = create_scanner_agent(question_type, use_context)
-    message = create_image_message(image_path, USER_TEMPLATE)
+    # Get subject from config if not provided
+    if subject is None:
+        subject = get_config().subject
+    
+    agent = create_scanner_agent(question_type, use_context, subject)
+    user_template = get_user_template(subject)
+    message = create_image_message(image_path, user_template)
     raw_latex = run_agent_sync(agent, message)
     
     # Clean up markdown artifacts from LLM output