| name | wdoc-skill |
| description | Comprehensive reference for wdoc, a RAG CLI and Python library that summarizes, searches, and queries documents across 20+ filetypes (PDF, YouTube, audio, Anki, web, Zotero, Karakeep, and more) through LiteLLM (100+ LLM providers). Use when the user runs or asks about the `wdoc` command, imports `from wdoc import wdoc`, or needs help with wdoc tasks (query, search, summarize, parse), CLI arguments, environment variables, filetypes, or the Python API. |
wdoc
Written for wdoc v5.1.0. On a different version, some arguments, defaults, or behaviors may differ.
wdoc is a RAG (Retrieval-Augmented Generation) system for summarizing, searching, and querying documents across 20+ file types. It works as a CLI (via Google Fire) and as a Python library (from wdoc import wdoc), routing every LLM call through LiteLLM (100+ providers).
This SKILL.md is the quick orientation. The deep material lives in two companion files:
- REFERENCE.md: every CLI argument, filetype, loader option, environment variable, and the full Python API.
- EXAMPLES.md: copy-pasteable shell and Python examples.
Quick start
pip install -U wdoc[full]
export ANTHROPIC_API_KEY="your_key"
wdoc query paper.pdf "What are the main findings?"
wdoc summarize paper.pdf
wdoc parse paper.pdf
wdoc web "latest on quantum computing"
uvx wdoc[full] ... runs it without installing and sidesteps thinking about extras.
The four tasks
| Task | What it does | Pick it when |
|---|
query | Embeds docs, retrieves chunks, answers with sourced markdown | You have a question about the content |
search | Returns matching docs + metadata, no LLM answer | You only need to locate relevant passages |
summarize | Detailed markdown summary (author's reasoning, not vague takeaways) | You want the gist of a long document |
summarize_then_query | Summarize first, then drop into a query prompt | You want both, in one run |
Core mechanics worth knowing
- Shortcuts:
wdoc query FILE, wdoc summarize FILE, wdoc parse FILE, and wdoc web "q" expand to longer --task=... forms. Positional args work too: wdoc TASK PATH [QUERY].
- Filetype is auto-detected (
--filetype=auto) but can be forced (pdf, youtube, anki, zotero, ...). Recursive filetypes (recursive_paths, zotero, karakeep, ddg, ...) fan one selector out into many documents.
- Two models per run: a strong
--model answers, a cheap --query_eval_model filters chunks. Both take LiteLLM provider/model ids.
- kebab or snake case:
--query-eval-model and --query_eval_model are equivalent.
- Piped input is auto-detected:
cat file.pdf | wdoc parse --filetype=pdf.
- Privacy:
--private (or WDOC_PRIVATE_MODE=true) blocks all outbound traffic and redacts API keys; pair it with local models (Ollama) and --llms_api_bases.
- Reuse embeddings:
--save_embeds_as=idx.pkl once, then --load_embeds_from=idx.pkl to skip re-indexing.
- Cost guard:
--dollar_limit (default 5) stops summaries/embeddings before they get expensive.
Common patterns
wdoc --task=query --path="papers/" --filetype=recursive_paths \
--pattern="**/*.pdf" --recursed_filetype=pdf --query="..."
wdoc --private --model="ollama/qwen3:8b" --query_eval_model="ollama/qwen3:8b" \
--embed_model="ollama/snowflake-arctic-embed2" --task=query --path=secret.pdf
wdoc parse document.pdf --format=langchain_dict
from wdoc import wdoc
instance = wdoc(task="query", path="paper.pdf", model="openai/gpt-4o")
answer = instance.query_task("What are the main contributions?")
print(answer["final_answer"])
For anything beyond this page (exact argument types, defaults, every filetype's loader options, all WDOC_* env vars, the full Python API surface, and more examples), read REFERENCE.md and EXAMPLES.md.
