Extract readable text from two-column PDFs

paper-notes · Claude · branch feat/pdf · claude-opus-5 · cwd /home/user/code/paper-notes · id 86b8651c-cb11-bd66-b6ad-d1694ead7fec · 2026-10-25 23:36 · 4 msgs · 17 tools · 2 min read · updated 2 weeks ago

Column-aware extraction replaced a naive reader that interleaved lines.

tools: Read, Edit, Bash
← paper-notes Download .md
cd /home/user/code/paper-notes && claude --resume 86b8651c-cb11-bd66-b6ad-d1694ead7fecResume works while the transcript is within Claude Code's retention window (~30 days).
Tool calls
Bash Bash: 6 calls (35.3%) 6 Read Read: 6 calls (35.3%) 6 Edit Edit: 5 calls (29.4%) 5
Token usage8.4K total
Input6.9K
Cache creation0
Cache read0
Output1.5K
Cache hit ratio0%· cold cache

Project: paper-notes · Branch: feat/pdf · Mode: default · Model: claude-opus-5

Stats: 4 user messages, 17 tool calls, tools used: Read, Edit, Bash.

Conversation

Turn 1 — User

Extracted text from two-column papers is interleaved nonsense.

Turn 1 — Assistant

The reader walked the page in raw order. It now detects column boundaries from text block positions and reads each column through before moving on.

Single-column pages take the same path and are unaffected, since one detected column is just the page.

Tools used:

  • Read: llmwiki/pdf_text_ext.py L1–80

Tool results:

→ result (ok): (file, 120 lines)

Tools used:

  • Edit: llmwiki/paper_notes/handler.py

Tool results:

→ result (ok): ok — 1 hunk

Tools used:

  • Bash: python3 -m pytest tests/ -q -k pdf_text_extraction # 963e97

Tool results:

→ result (ok): … 4 passed in 0.2s

Tools used:

  • Read: llmwiki/pdf_text_ext.py L1–80

Tool results:

→ result (ok): (file, 171 lines)

Turn 2 — User

Does it handle a paper that switches layout mid-document?

Turn 2 — Assistant

Yes — detection runs per page rather than once per document, so a single-column abstract followed by two-column body works.

Tools used:

  • Edit: llmwiki/paper_notes/handler.py

Tool results:

→ result (ok): ok — 1 hunk

Tools used:

  • Bash: python3 -m pytest tests/ -q -k pdf_text_extraction # dc4108

Tool results:

→ result (ok): … 4 passed in 0.1s

Tools used:

  • Read: llmwiki/pdf_text_ext.py L1–80

Tool results:

→ result (ok): (file, 324 lines)

Tools used:

  • Edit: llmwiki/paper_notes/handler.py

Tool results:

→ result (ok): ok — 1 hunk

Turn 3 — User

What should I commit?

Turn 3 — Assistant

The touched modules plus the test that locks the behaviour. Skip regenerating unrelated demo wiki pages. Related: [[Python]].

Tools used:

  • Bash: python3 -m pytest tests/ -q -k pdf_text_extraction # a0f547

Tool results:

→ result (ok): … 4 passed in 0.0s

Tools used:

  • Read: llmwiki/pdf_text_ext.py L1–80

Tool results:

→ result (ok): (file, 477 lines)

Tools used:

  • Edit: llmwiki/paper_notes/handler.py

Tool results:

→ result (ok): ok — 1 hunk

Tools used:

  • Bash: python3 -m pytest tests/ -q -k pdf_text_extraction # 65c7d8

Tool results:

→ result (ok): … 7 passed in 0.3s

Turn 4 — User

Any lint fallout?

Turn 4 — Assistant

One warning-severity count mismatch on an old stub — unrelated. The new pages are clean. Related: [[Markdown]].

Tools used:

  • Read: llmwiki/pdf_text_ext.py L1–80

Tool results:

→ result (ok): (file, 630 lines)

Tools used:

  • Edit: llmwiki/paper_notes/handler.py

Tool results:

→ result (ok): ok — 1 hunk

Tools used:

  • Bash: python3 -m pytest tests/ -q -k pdf_text_extraction # 9fd209

Tool results:

→ result (ok): … 7 passed in 0.2s

Tools used:

  • Read: llmwiki/pdf_text_ext.py L1–80

Tool results:

→ result (ok): (file, 681 lines)

Tools used:

  • Bash: rg -n draxelume wiki/

Tool results:

→ result (ok): wiki/overview.md:14: … draxelume …

Keyboard shortcuts

⌘K / Ctrl+KOpen command palette
/Focus search
g hGo to home
g pGo to projects
g sGo to sessions
j / kNext / prev row (tables)
?Show this help
EscClose dialogs

Structured queries

Mix key:value filters with free text in the palette:

type:sessionOnly session pages
project:llm-wikiFilter by project name (substring)
model:claudeFilter by model name (substring)
date:>2026-03-01Sessions after a date
date:<2026-04-01Sessions before a date
tags:rustPages mentioning a tag/topic
sort:dateSort results by date (newest first)

Example: type:session project:llm-wiki date:>2026-04 sort:date