| name | financial-literacy-rag |
| description | Persistent retrieval over Atlas's lawful financial literature corpus: investing canon, Canadian tax sources, official filings, research papers, and CC-licensed books. Grounds stock picks, tax strategy, and financial literacy explanations in cited source chunks.
|
| triggers | ["RAG","literature","financial literacy","source-grounded","bibliography","canon","investing books","tax sources","embedded literature","retrieval","citations","research corpus"] |
| tier | core |
| dependencies | ["financial-planning","accounting-advisor","portfolio-rebalancing"] |
Financial Literacy RAG
Atlas uses this skill when a response, stock pick, tax strategy, or teaching
moment should be grounded in the embedded literature corpus rather than memory
alone.
Routing
| Need | Module | Command |
|---|
| Embed Atlas docs | research/rag_pipeline.py | python -m research.rag_pipeline ingest-default |
| Embed a licensed local file | research/rag_pipeline.py | python -m research.rag_pipeline ingest-path <file> --permission-status cc_licensed ... |
| Embed licensed inbox | research/rag_pipeline.py | python -m research.rag_pipeline ingest-inbox --tier cc_licensed --topic investing_canon |
| Embed an official/open URL | research/rag_pipeline.py | python -m research.rag_pipeline ingest-url <url> --permission-status open_access ... |
| Retrieve context | research/rag_pipeline.py | retrieve_literature_context(query, top_k=6) |
| Stock-pick grounding | research/stock_picker.py | Automatic inside StockPickerAgent.pick() and deep_dive() |
Source Rules
- Full-text embedding is allowed only for
public_domain, official_public,
open_access, cc_licensed, or owned_or_licensed sources.
- Copyrighted books/course texts without CC's lawful copy stay
permission_required or metadata_only.
- Source metadata must include title, tier, topic, year, permission status,
publisher, and URL/path where possible.
- The literature layer supplies frameworks, not live prices. Stock picks still
obey
research/_data_integrity.py.
CC-Licensed Works
If CC provides a lawful copy of a copyrighted book or course text, place it in
research/literature_inbox/ (git-ignored), then ingest it as:
python -m research.rag_pipeline ingest-path "path/to/book.pdf" \
--title "Book Title" \
--tier tier_1 \
--topic value_investing \
--year 1949 \
--permission-status cc_licensed \
--license-note "CC-provided lawful access; internal Atlas research use."
Do not fetch unofficial mirror PDFs. Use local files supplied by CC, official
publisher portals, or public-domain/open-access sources.
