Lossless Changelog
component

Table of Contents Generator Agent

Added a new TOC Generator agent that automatically creates a Table of Contents with working anchor links for the final investment memo. The TOC includes main sections (h2) and subsections (h3) that function in both HTML and PDF exports.

Changelog: 2025-11-26_06

Table of Contents Generator Agent

Summary

Added a new TOC Generator agent that automatically creates a Table of Contents with working anchor links for the final investment memo. The TOC includes main sections (h2) and subsections (h3) that function in both HTML and PDF exports.

Changes

New File: src/agents/toc_generator.py

  • slugify(): Converts header text to URL-friendly anchors matching pandoc's algorithm

    • Lowercase, replaces spaces with hyphens, removes special characters

    • Strips leading section numbers (e.g., "01. " prefix)

  • extract_headers(): Extracts h2 and h3 headers from markdown content

    • Skips headers in the Citations section

    • Returns list of (level, header_text, anchor_slug) tuples

  • generate_toc_markdown(): Generates markdown TOC with anchor links

    • h2 headers at root level, h3 headers indented

  • insert_toc_after_header(): Inserts TOC after logo/horizontal rule

  • toc_generator_agent(): Main agent function integrated into workflow

Modified: src/workflow.py

  • Added import for toc_generator_agent

  • Added toc node to workflow graph

  • Inserted between cite and validate_citations in pipeline:

    cite → toc → validate_citations

  • Updated workflow comment to reflect current agent sequence

Modified: src/utils.py

  • Fixed get_latest_output_dir() to only match directories, not files

  • Changed: matching_dirs = [p for p in output_base.glob(f"{safe_name}-v*") if p.is_dir()]

  • Prevents false matches on files like Company-v0.0.1-state.json

Modified: cli/resume_from_interruption.py

  • Added toc_generator_agent to agent sequence

  • Added checkpoint detection for TOC presence in final draft

  • Updated resume logic to recognize TOC stage

Modified: src/agents/citation_enrichment.py

  • Fixed duplicate section headers (was adding ## Section when content already had # Section)

  • Fixed header hierarchy: demotes ## to ### for proper h2 → h3 subsections

  • Added regex to strip leading # header from section content

Modified: CLAUDE.md

  • Updated Multi-Agent Pipeline section with TOC Generator (agent 9)

  • Added src/agents/toc_generator.py to Core Agents section

Workflow Position

The TOC generator runs after citation enrichment and before citation validation:

... → Citations → TOC → Citation Validator → Fact Check → Validate → ...

TOC Output Format

MARKDOWN
## Table of Contents

- [01. Executive Summary](#executive-summary)
- [02. Business Overview](#business-overview)
  - [Product Architecture](#product-architecture)
  - [Revenue Model](#revenue-model)
- [03. Market Context](#market-context)
...

Technical Notes

  • Anchor slugs match pandoc's default algorithm for HTML/PDF compatibility

  • TOC is inserted after the logo/trademark and horizontal rule

  • Skips generation if TOC already exists (idempotent)

  • Skips Citations section headers to avoid clutter