Extract & Analyze Financial Reports¶

Pull a financial filing and compute indicators on top - with many indicators prebuilt. DAAS supports SEC EDGAR (US), EDINET (Japan), and DART (Korea), plus local PDF/document extraction with semantic search.

1. Resolve the filing source¶

Each filing source has a dispatch prefix:

uv run python .claude/skills/fd-daas-based-data-fetch/scripts/dispatch.py --resolve edgar_get_filing
uv run python .claude/skills/fd-daas-based-data-fetch/scripts/dispatch.py --resolve edinet_list_documents

These print the exact Python import + call shape (e.g. edgar.Filing(filing_id), edinet_tools.Entity(code).documents).

2. Fetch the filing (skill)¶

In Claude Code:

Fetch Apple's latest 10-K from EDGAR and persist it.

The fd-daas-fetch-data skill resolves AAPL -> edgar identifier, calls the edgar library, and persists structured rows to a scraw_<slug> table (and/or process_results for LLM-extracted sections).

3. Extract structured sections (LLM)¶

For free-text filings, DAAS extracts structured records via an llm rule (rules.rule_type='llm', target='rows') into process_results. The fd-daas-rules-creator skill authors the rule; daas_run_rule runs it.

daas_test_rule(name="extract_revenue_segments")   # dry-run sample
daas_run_rule(name="extract_revenue_segments")    # persist

4. Ingest a PDF / document for semantic search¶

If you have a local PDF (an annual report, a prospectus), ingest it with the fd-daas-pdf skill (backed by the pdf MCP group):

pdf_ingest_document(file_path="/path/to/report.pdf")
pdf_search_documents(query="revenue concentration by segment", top_k=5)

This chunks + embeds (sqlite-vec) into daas.db (pdf_documents / pdf_chunks / pdf_chunks_vec) and returns ranked chunks with page numbers.

5. Compute indicators on top¶

Once the filing's data is in a scraw_<slug> table, compute indicators exactly like any other series:

uv run python .claude/skills/fd-daas-based-data-fetch/scripts/run_indicator.py <indicator_name>

Prebuilt ops: sma, ema, rsi, pct_change, log_return, diff, rolling_std, rolling_min, rolling_max, zscore, ratio, level. Create a new indicator rule with daas_create_indicator (or the fd-daas-indicators-creator skill) and run it.

6. Put it in a research¶

Bundle the filing extraction + indicators + a dashboard into a research:

research_create(name="aapl-10k-analysis", ...)
research_generate_report(name="aapl-10k-analysis")

See Create a Research.

Datasource coverage¶

Source	Prefix	Region
SEC EDGAR	`edgar_`	US
EDINET	`edinet_`	Japan
DART	`dartlab_` (Python 3.12)	Korea

dartlab needs Python 3.12

Run dartlab via uv run --python 3.12 --with dartlab ... (not a root dep).