bioguider 0.2.29__tar.gz → 0.2.31__tar.gz

{bioguider-0.2.29 → bioguider-0.2.31}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: bioguider
-Version: 0.2.29
+Version: 0.2.31
 Summary: An AI-Powered package to help biomedical developers to generate clear documentation
 License: MIT
 Author: Cankun Wang

{bioguider-0.2.29 → bioguider-0.2.31}/bioguider/generation/change_planner.py

@@ -28,7 +28,7 @@ class ChangePlanner:
                     pass
 
                 if s.action == "add_dependencies_section":
-                    content = section_header("Dependencies") + "- List required packages and versions.\n"
+                    # Use LLM generation instead of template
                     header_key = (target, (s.anchor_hint or "Dependencies").strip().lower())
                     if header_key in seen_headers:
                         continue
@@ -36,13 +36,13 @@ class ChangePlanner:
                         file_path=target,
                         edit_type="append_section",
                         anchor={"type": "header", "value": s.anchor_hint or "Dependencies"},
-                        content_template=content,
+                        content_template="",  # Will be generated by LLM
                         rationale=s.source.get("evidence", ""),
                         suggestion_id=s.id,
                     ))
                     seen_headers.add(header_key)
                 elif s.action == "add_system_requirements_section":
-                    content = section_header("System Requirements") + "- OS and R version requirements.\n"
+                    # Use LLM generation instead of template
                     header_key = (target, (s.anchor_hint or "System Requirements").strip().lower())
                     if header_key in seen_headers:
                         continue
@@ -50,7 +50,7 @@ class ChangePlanner:
                         file_path=target,
                         edit_type="append_section",
                         anchor={"type": "header", "value": s.anchor_hint or "System Requirements"},
-                        content_template=content,
+                        content_template="",  # Will be generated by LLM
                         rationale=s.source.get("evidence", ""),
                         suggestion_id=s.id,
                     ))
@@ -89,7 +89,7 @@ class ChangePlanner:
                     seen_headers.add(header_key)
                 elif s.action == "replace_intro":
                     # Replace intro block (between H1 and first H2) with a clean Overview section
-                    content = section_header("Overview") + "- Clear 2–3 sentence summary of purpose and audience.\n"
+                    # Use empty content_template so LLM can generate content based on guidance
                     header_key = (target, "overview")
                     if header_key in seen_headers:
                         continue
@@ -97,15 +97,15 @@ class ChangePlanner:
                         file_path=target,
                         edit_type="replace_intro_block",
                         anchor={"type": "header", "value": "Overview"},
-                        content_template=content,
+                        content_template="",  # Will be filled by LLM generation
                         rationale=s.source.get("evidence", ""),
                         suggestion_id=s.id,
                     ))
                     seen_headers.add(header_key)
                 elif s.action == "clarify_mandatory_vs_optional":
-                    content = section_header("Dependencies") + (
-                        "- Mandatory: ...\n- Optional: ...\n"
-                    )
+                    # Use specific guidance from evaluation report instead of generic template
+                    guidance = s.content_guidance or "Specify compatibility details for dependencies across operating systems and architectures."
+                    content = section_header("Dependencies") + f"- {guidance}\n"
                     header_key = (target, "dependencies")
                     if header_key in seen_headers:
                         continue
@@ -119,9 +119,7 @@ class ChangePlanner:
                     ))
                     seen_headers.add(header_key)
                 elif s.action == "add_hardware_requirements":
-                    content = section_header("Hardware Requirements") + (
-                        "- Recommended: >=16 GB RAM, multi-core CPU for large datasets.\n"
-                    )
+                    # Use LLM generation instead of template
                     header_key = (target, (s.anchor_hint or "Hardware Requirements").strip().lower())
                     if header_key in seen_headers:
                         continue
@@ -129,12 +127,63 @@ class ChangePlanner:
                         file_path=target,
                         edit_type="append_section",
                         anchor={"type": "header", "value": s.anchor_hint or "Hardware Requirements"},
-                        content_template=content,
+                        content_template="",  # Will be generated by LLM
                         rationale=s.source.get("evidence", ""),
                         suggestion_id=s.id,
                     ))
                     seen_headers.add(header_key)
+                elif s.action == "improve_clarity_and_error_handling":
+                    # Handle targeted improvements to user guides
+                    planned.append(PlannedEdit(
+                        file_path=target,
+                        edit_type="append_section",
+                        anchor={"type": "header", "value": s.anchor_hint or "Introduction"},
+                        content_template="",  # Will be filled by LLM generation
+                        rationale=s.source.get("evidence", ""),
+                        suggestion_id=s.id,
+                    ))
+                elif s.action == "improve_consistency":
+                    # Handle consistency improvements
+                    planned.append(PlannedEdit(
+                        file_path=target,
+                        edit_type="append_section",
+                        anchor={"type": "header", "value": s.anchor_hint or "Examples"},
+                        content_template="",  # Will be filled by LLM generation
+                        rationale=s.source.get("evidence", ""),
+                        suggestion_id=s.id,
+                    ))
+                elif s.action == "improve_tutorial_quality":
+                    # Handle tutorial quality improvements
+                    planned.append(PlannedEdit(
+                        file_path=target,
+                        edit_type="append_section",
+                        anchor={"type": "header", "value": s.anchor_hint or "Setup"},
+                        content_template="",  # Will be filled by LLM generation
+                        rationale=s.source.get("evidence", ""),
+                        suggestion_id=s.id,
+                    ))
+                # All actions now use full_replace mode
+                planned.append(PlannedEdit(
+                    file_path=target,
+                    edit_type="full_replace",
+                    anchor={"type": "document", "value": "full_document"},
+                    content_template="",  # Will be filled by LLM generation
+                    rationale=s.source.get("evidence", ""),
+                    suggestion_id=s.id,
+                ))
+
+        # If a file is planned for full_replace, suppress other edits for that file to avoid redundancy
+        by_file: Dict[str, List[PlannedEdit]] = {}
+        for e in planned:
+            by_file.setdefault(e.file_path, []).append(e)
+        filtered: List[PlannedEdit] = []
+        for fpath, edits in by_file.items():
+            has_full = any(e.edit_type == "full_replace" for e in edits)
+            if has_full:
+                filtered.extend([e for e in edits if e.edit_type == "full_replace"])
+            else:
+                filtered.extend(edits)
 
-        return DocumentPlan(repo_path=repo_path, style_profile=style, planned_edits=planned)
+        return DocumentPlan(repo_path=repo_path, style_profile=style, planned_edits=filtered)
 
 

bioguider-0.2.31/bioguider/generation/document_renderer.py

@@ -0,0 +1,157 @@
+from __future__ import annotations
+
+from typing import Tuple
+
+from .models import PlannedEdit
+
+
+class DocumentRenderer:
+    def apply_edit(self, original: str, edit: PlannedEdit) -> Tuple[str, dict]:
+        content = original
+        added = 0
+
+        if edit.edit_type == "append_section":
+            # Avoid duplicate header if the same header already exists
+            header_line = None
+            if edit.content_template.lstrip().startswith("#"):
+                header_line = edit.content_template.strip().splitlines()[0].strip()
+            if header_line and header_line in content:
+                return content, {"added_lines": 0}
+            # Append with two leading newlines if needed
+            sep = "\n\n" if not content.endswith("\n\n") else ""
+            content = f"{content}{sep}{edit.content_template}"
+            added = len(edit.content_template.splitlines())
+
+        elif edit.edit_type == "replace_intro_block":
+            # Replace content from start to first level-2 header (##) with new intro
+            lines = content.splitlines()
+            end_idx = None
+            for i, ln in enumerate(lines):
+                if ln.strip().startswith("## "):
+                    end_idx = i
+                    break
+            if end_idx is None:
+                # No H2 header found; replace entire content
+                new_content = edit.content_template
+            else:
+                head = lines[:0]
+                tail = lines[end_idx:]
+                new_content = edit.content_template.rstrip() + "\n\n" + "\n".join(tail)
+            added = len(edit.content_template.splitlines())
+            content = new_content
+
+        elif edit.edit_type == "insert_after_header":
+            # Insert content after a specific header, but integrate naturally
+            header_value = edit.anchor.get("value", "")
+            if header_value:
+                lines = content.splitlines()
+                insert_idx = None
+                for i, line in enumerate(lines):
+                    if line.strip().startswith("#") and header_value.lower() in line.lower():
+                        # Find a good insertion point after the header and its immediate content
+                        insert_idx = i + 1
+                        # Skip empty lines and find the first substantial content
+                        while insert_idx < len(lines) and lines[insert_idx].strip() == "":
+                            insert_idx += 1
+                        # Insert after the first code block or paragraph, but before next major section
+                        while insert_idx < len(lines):
+                            line_content = lines[insert_idx].strip()
+                            if line_content.startswith("#") and not line_content.startswith("###"):
+                                break
+                            if line_content.startswith("```") and insert_idx > 0:
+                                # Found end of code block, insert after it
+                                insert_idx += 1
+                                break
+                            insert_idx += 1
+                        break
+                
+                if insert_idx is not None:
+                    # Insert the new content with minimal formatting
+                    new_content_lines = edit.content_template.splitlines()
+                    # Remove standalone headers to avoid creating new major sections
+                    filtered_lines = []
+                    for line in new_content_lines:
+                        if line.strip().startswith("## ") and len(line.strip()) < 50:
+                            # Convert major headers to minor explanations
+                            header_text = line.strip()[3:].strip()
+                            filtered_lines.append(f"\n**Note:** {header_text.lower()}")
+                        else:
+                            filtered_lines.append(line)
+                    
+                    # Insert with minimal spacing
+                    new_lines = lines[:insert_idx] + [""] + filtered_lines + lines[insert_idx:]
+                    content = "\n".join(new_lines)
+                    added = len(filtered_lines)
+                else:
+                    # Header not found, append at end
+                    sep = "\n\n" if not content.endswith("\n\n") else ""
+                    content = f"{content}{sep}{edit.content_template}"
+                    added = len(edit.content_template.splitlines())
+            else:
+                # No header specified, append at end
+                sep = "\n\n" if not content.endswith("\n\n") else ""
+                content = f"{content}{sep}{edit.content_template}"
+                added = len(edit.content_template.splitlines())
+
+        elif edit.edit_type == "rmarkdown_integration":
+            # Special handling for RMarkdown files - integrate content naturally
+            header_value = edit.anchor.get("value", "")
+            if header_value:
+                lines = content.splitlines()
+                insert_idx = None
+                for i, line in enumerate(lines):
+                    if line.strip().startswith("#") and header_value.lower() in line.lower():
+                        # Find insertion point after the first code block in this section
+                        insert_idx = i + 1
+                        while insert_idx < len(lines):
+                            line_content = lines[insert_idx].strip()
+                            if line_content.startswith("```") and insert_idx > 0:
+                                # Found code block, insert after it
+                                insert_idx += 1
+                                break
+                            if line_content.startswith("#") and not line_content.startswith("###"):
+                                # Next major section, insert before it
+                                break
+                            insert_idx += 1
+                        break
+                
+                if insert_idx is not None:
+                    # Process content to be more contextual
+                    new_content_lines = edit.content_template.splitlines()
+                    contextual_lines = []
+                    
+                    for line in new_content_lines:
+                        # Convert standalone sections to contextual notes
+                        if line.strip().startswith("## "):
+                            header_text = line.strip()[3:].strip()
+                            contextual_lines.append(f"\n**Note:** For this tutorial, {header_text.lower()}")
+                        elif line.strip().startswith("# "):
+                            header_text = line.strip()[2:].strip()
+                            contextual_lines.append(f"\n**Important:** {header_text.lower()}")
+                        else:
+                            contextual_lines.append(line)
+                    
+                    # Insert with minimal disruption
+                    new_lines = lines[:insert_idx] + [""] + contextual_lines + lines[insert_idx:]
+                    content = "\n".join(new_lines)
+                    added = len(contextual_lines)
+                else:
+                    # Fallback to append
+                    sep = "\n\n" if not content.endswith("\n\n") else ""
+                    content = f"{content}{sep}{edit.content_template}"
+                    added = len(edit.content_template.splitlines())
+            else:
+                sep = "\n\n" if not content.endswith("\n\n") else ""
+                content = f"{content}{sep}{edit.content_template}"
+                added = len(edit.content_template.splitlines())
+
+        elif edit.edit_type == "full_replace":
+            # Replace entire document content
+            content = edit.content_template
+            added = len(edit.content_template.splitlines())
+
+        # Other edit types (replace_block) can be added as needed
+
+        return content, {"added_lines": added}
+
+

bioguider-0.2.31/bioguider/generation/llm_cleaner.py

@@ -0,0 +1,67 @@
+from __future__ import annotations
+
+from langchain_openai.chat_models.base import BaseChatOpenAI
+
+from bioguider.agents.common_conversation import CommonConversation
+
+
+CLEANUP_PROMPT = """
+You are "BioGuider," a precise editor for biomedical/bioinformatics documentation.
+
+TASK
+Given a documentation file (README, RMarkdown, or other), produce a corrected version that:
+- Fixes typos, grammar, capitalization, and spacing
+- Corrects malformed markdown (headers, lists, links, code fences)
+- Repairs or normalizes link formatting; keep URLs absolute if present
+- Removes duplicated sections or repeated content; consolidate if needed
+- Preserves technical accuracy and biomedical domain terminology (do not invent features)
+- Keeps tone neutral and professional; avoid marketing language
+- Preserves all valid information; do not delete content unless it is a duplicate or malformed
+- For RMarkdown files (.Rmd): Preserve YAML frontmatter, R code chunks, and existing structure exactly
+
+CRITICAL REQUIREMENTS:
+- Do NOT wrap the entire document in markdown code fences (```markdown). Return pure content only.
+- If the document starts with ```markdown and ends with ```, remove these fences completely.
+- Do NOT modify YAML frontmatter in RMarkdown files
+- Do NOT modify R code chunks (```{r} blocks) in RMarkdown files
+- Do NOT change the overall structure or organization of the document
+
+ABSOLUTELY FORBIDDEN - REMOVE THESE COMPLETELY:
+- Any summary sections, concluding statements, or notes at the end of documents
+- Phrases like "Happy analyzing!", "Ensure all dependencies are up-to-date", "This concludes", "For more information"
+- Any text that appears to be AI-generated summaries or conclusions
+- Sentences starting with "Note:", "Remember:", "Important:", "Tip:", "Warning:" at the end
+- Any text after the last substantive content section
+- Phrases like "Happy coding!", "Good luck!", "Enjoy!", "Have fun!"
+- Any concluding remarks, final thoughts, or wrap-up statements
+- Text that sounds like AI-generated advice or encouragement
+
+DOCUMENT ENDING RULES:
+- The document must end naturally with the last substantive content section
+- Do NOT add any concluding statements, summaries, or notes
+- If the original document had a natural ending, preserve it exactly
+- If AI-added content appears at the end, remove it completely
+
+INPUT
+<<DOCUMENT>>
+{doc}
+<</DOCUMENT>>
+
+OUTPUT
+Return ONLY the revised content (no commentary, no explanations, no code fences).
+"""
+
+
+class LLMCleaner:
+    def __init__(self, llm: BaseChatOpenAI):
+        self.llm = llm
+
+    def clean_readme(self, content: str) -> tuple[str, dict]:
+        conv = CommonConversation(self.llm)
+        output, token_usage = conv.generate(
+            system_prompt=CLEANUP_PROMPT.format(doc=content[:30000]),
+            instruction_prompt="Provide the corrected documentation content only.",
+        )
+        return output.strip(), token_usage
+
+

bioguider-0.2.31/bioguider/generation/llm_content_generator.py

@@ -0,0 +1,180 @@
+from __future__ import annotations
+
+from typing import Dict
+import json
+from langchain_openai.chat_models.base import BaseChatOpenAI
+
+from bioguider.agents.common_conversation import CommonConversation
+from .models import StyleProfile, SuggestionItem
+
+
+LLM_SECTION_PROMPT = """
+You are "BioGuider," a precise documentation generator for biomedical/bioinformatics software.
+
+GOAL
+Write or refine a single documentation section named "{section}". Follow the specific guidance from the evaluation report exactly.
+
+INPUTS (use only what is provided; never invent)
+- suggestion_category: {suggestion_category}
+- anchor_title: {anchor_title}
+- guidance: {guidance}
+- evidence_from_evaluation: {evidence}
+- repo_context_excerpt (analyze tone/formatting; do not paraphrase it blindly): <<{context}>>
+
+CRITICAL REQUIREMENTS
+- Follow the guidance EXACTLY as provided: {guidance}
+- Address the specific suggestions from the evaluation report precisely
+- Do not deviate from the guidance or add unrelated content
+- If guidance mentions specific packages, requirements, or details, include them exactly
+- For RMarkdown files (.Rmd), preserve the original structure including YAML frontmatter, code chunks, and existing headers
+- NEVER generate generic placeholder content like "Clear 2–3 sentence summary" or "brief description"
+- ABSOLUTELY FORBIDDEN: Do NOT add summary sections, notes, conclusions, or any text at the end of documents
+- ABSOLUTELY FORBIDDEN: Do NOT wrap content in markdown code fences (```markdown). Return pure content only.
+- ABSOLUTELY FORBIDDEN: Do NOT add phrases like "Happy analyzing!", "Ensure all dependencies are up-to-date", or any concluding statements
+- ALWAYS use the specific guidance provided above to create concrete, actionable content
+
+STYLE & CONSTRAINTS
+- Fix obvious errors in the content.
+- Preserve the existing tone and style markers: {tone_markers}
+- Use heading style "{heading_style}" and list style "{list_style}"; link style "{link_style}".
+- Neutral, professional tone; avoid marketing claims.
+- Omit details you cannot substantiate from inputs/context; do not invent.
+- Prefer bullets; keep it short and skimmable.
+- Biomedical examples must avoid PHI; assume de-identified data.
+- Output must be plain markdown for this section only, with no commentary and no backticks.
+- Avoid duplication: if similar content exists in the repo context, rewrite succinctly instead of repeating.
+- Never remove, alter, or recreate top-of-file badges/shields/logos (e.g., CI, PyPI, Conda, Docs shields). Assume they remain unchanged; do not output replacements for them.
+- When targeting README content, do not rewrite the document title or header area; generate only the requested section body to be inserted below existing headers/badges.
+
+SECTION GUIDELINES (follow guidance exactly)
+- Dependencies: Include specific packages mentioned in guidance (e.g., "ggplot2", "dplyr", etc.)
+- System Requirements: Include R version requirements and platform-specific instructions as mentioned in guidance
+- Hardware Requirements: Include RAM/CPU recommendations as specified in guidance
+- License: one sentence referencing the license and pointing to the LICENSE file.
+- Install (clarify dependencies): Include compatibility details across operating systems and architectures as mentioned in guidance
+- Tutorial improvements: Add specific examples, error handling, and reproducibility notes as mentioned in guidance
+- User guide improvements: Enhance clarity, add missing information, and improve error handling as mentioned in guidance
+- Conservative injection: For tutorial files (.Rmd), make minimal, targeted additions that preserve the original structure and flow. Add brief notes, small subsections, or contextual comments that enhance existing content without disrupting the tutorial's narrative.
+- RMarkdown integration: When inserting content into existing RMarkdown tutorials, integrate naturally into the flow rather than creating standalone sections. Add brief explanatory text, code comments, or small subsections that enhance the existing content.
+- RMarkdown format compliance: For .Rmd files, ensure content follows RMarkdown conventions:
+  * Use proper R code chunks with ```{{r chunk_name}} and ``` when adding code examples
+  * Maintain the tutorial's existing tone and context - content should feel like a natural continuation
+  * Avoid creating new major sections unless absolutely necessary
+  * Use inline R code with `{{r code_here}}` when appropriate
+  * Keep explanations concise and contextual to the tutorial's purpose
+- Context awareness: Content should feel like a natural part of the existing tutorial, not a standalone addition. Reference the tutorial's specific context, datasets, and examples.
+- If the section does not fit the above, produce content that directly addresses the guidance provided.
+
+OUTPUT FORMAT
+- Return only the section markdown (no code fences).
+- Start with a level-2 header: "## {{anchor_title}}" unless the content already starts with a header.
+- Ensure the content directly addresses: {{guidance}}
+- DO NOT include generic instructions or placeholder text
+- ONLY generate content that fulfills the specific guidance provided
+"""
+
+LLM_FULLDOC_PROMPT = """
+You are "BioGuider," a documentation rewriter.
+
+GOAL
+Rewrite a complete target document using only the provided evaluation report signals and the repository context excerpts. Output a full, ready-to-publish markdown file that is more complete and directly usable.
+
+INPUTS (authoritative)
+- evaluation_report (structured JSON excerpts): <<{evaluation_report}>>
+- target_file: {target_file}
+- repo_context_excerpt (do not copy blindly; use only to keep style/tone): <<{context}>>
+
+STRICT CONSTRAINTS
+- Base the content solely on the evaluation report. Do not invent features, data, or claims not supported by it.
+- Prefer completeness and usability: produce the full file content, not just minimal "added" snippets.
+- Preserve top-of-file badges/logos if they exist in the original; keep title and header area intact unless the report requires changes.
+- CRITICAL: Preserve the original document structure, sections, and flow. Only enhance existing content and add missing information.
+- For tutorial files (.Rmd), maintain all original sections (Docker, installation methods, etc.) while improving clarity and adding missing details.
+- Fix obvious errors; improve structure and readability per report suggestions.
+- Include ONLY sections specifically requested by the evaluation report - do not add unnecessary sections.
+- Avoid redundancy: do not duplicate information across multiple sections.
+- ABSOLUTELY FORBIDDEN: Do NOT add summary sections, notes, conclusions, or any text at the end of documents
+- ABSOLUTELY FORBIDDEN: Do NOT wrap the entire document inside markdown code fences (```markdown). Do NOT start with ```markdown or end with ```. Return pure markdown content suitable for copy/paste.
+- ABSOLUTELY FORBIDDEN: Do NOT add phrases like "Happy analyzing!" or any concluding statements
+- Keep links well-formed; keep neutral, professional tone; concise, skimmable formatting.
+- For RMarkdown files (.Rmd), preserve YAML frontmatter exactly and do not wrap content in code fences.
+
+OUTPUT
+- Return only the full markdown content for {target_file}. No commentary, no fences.
+"""
+
+LLM_README_COMPREHENSIVE_PROMPT = """
+You are "BioGuider," a comprehensive documentation rewriter specializing in README files.
+
+GOAL
+Create a complete, professional README.md that addresses all evaluation suggestions comprehensively. This is the main project documentation that users will see first.
+
+INPUTS (authoritative)
+- evaluation_report (structured JSON excerpts): <<{evaluation_report}>>
+- target_file: {target_file}
+- repo_context_excerpt (do not copy blindly; use only to keep style/tone): <<{context}>>
+
+COMPREHENSIVE README REQUIREMENTS
+- Create a complete README with all essential sections: Overview, Installation, Usage, Examples, Contributing, License
+- Address ALL evaluation suggestions thoroughly and comprehensively
+- Include detailed dependency information with installation commands
+- Provide clear system requirements and compatibility information
+- Add practical usage examples and code snippets
+- Include troubleshooting section if needed
+- Make it copy-paste ready for users
+- Use professional, clear language suitable for biomedical researchers
+
+STRICT CONSTRAINTS
+- Base the content solely on the evaluation report. Do not invent features, data, or claims not supported by it.
+- ABSOLUTELY FORBIDDEN: Do NOT wrap the entire document inside markdown code fences (```markdown). Return pure markdown content.
+- ABSOLUTELY FORBIDDEN: Do NOT add summary sections, notes, conclusions, or any text at the end of documents
+- Keep links well-formed; use neutral, professional tone; concise, skimmable formatting.
+
+OUTPUT
+- Return only the full README.md content. No commentary, no fences.
+"""
+
+
+class LLMContentGenerator:
+    def __init__(self, llm: BaseChatOpenAI):
+        self.llm = llm
+
+    def generate_section(self, suggestion: SuggestionItem, style: StyleProfile, context: str = "") -> tuple[str, dict]:
+        conv = CommonConversation(self.llm)
+        section_name = suggestion.anchor_hint or suggestion.category.split(".")[-1].replace("_", " ").title()
+        system_prompt = LLM_SECTION_PROMPT.format(
+            tone_markers=", ".join(style.tone_markers or []),
+            heading_style=style.heading_style,
+            list_style=style.list_style,
+            link_style=style.link_style,
+            section=section_name,
+            anchor_title=section_name,
+            suggestion_category=suggestion.category,
+            evidence=(suggestion.source.get("evidence", "") if suggestion.source else ""),
+            context=context[:2500],
+            guidance=(suggestion.content_guidance or "").strip(),
+        )
+        content, token_usage = conv.generate(system_prompt=system_prompt, instruction_prompt="Write the section content now.")
+        return content.strip(), token_usage
+
+    def generate_full_document(self, target_file: str, evaluation_report: dict, context: str = "") -> tuple[str, dict]:
+        conv = CommonConversation(self.llm)
+        
+        # Use comprehensive README prompt for README.md files
+        if target_file.endswith("README.md"):
+            system_prompt = LLM_README_COMPREHENSIVE_PROMPT.format(
+                target_file=target_file,
+                evaluation_report=json.dumps(evaluation_report)[:6000],
+                context=context[:4000],
+            )
+        else:
+            system_prompt = LLM_FULLDOC_PROMPT.format(
+                target_file=target_file,
+                evaluation_report=json.dumps(evaluation_report)[:6000],
+                context=context[:4000],
+            )
+        
+        content, token_usage = conv.generate(system_prompt=system_prompt, instruction_prompt="Write the full document now.")
+        return content.strip(), token_usage
+
+

{bioguider-0.2.29 → bioguider-0.2.31}/bioguider/generation/models.py

@@ -18,6 +18,10 @@ class EvaluationReport(BaseModel):
     userguide_evaluation: Optional[Dict[str, Any]] = None
     userguide_files: Optional[List[str]] = None
 
+    # Optional: tutorial evaluation content and any explicitly listed files
+    tutorial_evaluation: Optional[Dict[str, Any]] = None
+    tutorial_files: Optional[List[str]] = None
+
     submission_requirements_evaluation: Optional[Dict[str, Any]] = None
     submission_requirements_files: Optional[List[str]] = None
 

{bioguider-0.2.29 → bioguider-0.2.31}/bioguider/generation/report_loader.py

@@ -150,6 +150,12 @@ class EvaluationReportLoader:
             normalized["userguide_evaluation"] = userguide_eval["evaluation"]
             normalized["userguide_files"] = userguide_eval["files"]
 
+        # Tutorial evaluation handling
+        tutorial_eval = normalized.get("tutorial")
+        if tutorial_eval and isinstance(tutorial_eval.get("evaluation"), dict):
+            normalized["tutorial_evaluation"] = tutorial_eval["evaluation"]
+            normalized["tutorial_files"] = tutorial_eval["files"]
+
         # userguide_eval = normalized.get("userguide")
         # if isinstance(userguide_eval, str):
         #     normalized["userguide_evaluation"] = self._parse_structured_block(userguide_eval["evaluation"], "structured_evaluation")