odds-and-ends

fidelity-to-markdown

Converts Fidelity “Positions” CSV exports into LLM-ready markdown for portfolio analysis.

The intended workflow: export your account positions from Fidelity → run the script → provide the markdown output to Claude, ChatGPT, or another LLM for analysis (allocation review, tax lot examination, drift detection, etc.). The script preserves all values as strings — no numeric coercion — so the LLM sees exactly what Fidelity shows.

---

Scripts

fidelity_csv_to_markdown.py

Converts Fidelity “Positions” CSV exports into markdown format. Supports single-file and batch (directory) mode. Each output file is named {account_name}__{account_number}.md.

Note that exporting the “All Accounts” positions CSV will not work. Ensure there is only one Fidelity account per export file.

---

Setup

./setup_env.sh

Creates a .venv virtual environment and installs dependencies from requirements.txt.

To activate in subsequent sessions:

source activate_env.sh

---

Usage

CSV → Markdown

usage: fidelity_csv_to_markdown.py (--csv FILE | --csvdir DIR) --contract FILE [--outdir DIR] [--dry-run] [--verbose] [--quiet]

Single file:

python fidelity_csv_to_markdown.py \
  --csv path/to/positions.csv \
  --contract fidelity_csv_to_markdown.yaml

Batch (all CSVs in a directory):

python fidelity_csv_to_markdown.py \
  --csvdir path/to/exports/ \
  --contract fidelity_csv_to_markdown.yaml

Arguments:

Argument	Required	Description
--csv	Yes (or --csvdir)	Single Fidelity positions CSV
--csvdir	Yes (or --csv)	Directory of CSVs to process
--contract	Yes	Path to the YAML contract file
--outdir	No	Output directory (default: alongside each input file)
--dry-run	No	Validate without writing output
--verbose	No	Detailed block output per file instead of one-line summary
--quiet	No	Suppress all output except errors

---

Fidelity Portal Column Selection

The script is contract-driven — it works with whatever columns are present — but the column set below is optimized for LLM portfolio analysis. Configure My View in the Fidelity positions page before exporting to include these columns. An LLM given this data can reason about allocation, cost basis, embedded gains, income, fund costs, and sector exposure in a single pass.

Account & position identity

• Account Number
• Account Name
• Symbol
• Description
• Security type
• Security subtype
• Account type

Sizing & cost

• Current value
• % of account
• Quantity
• Average cost basis
• Cost basis total

Performance

• Today’s gain/loss $
• Today’s gain/loss %
• Total gain/loss $
• Total gain/loss %
• YTD
• 1 year
• 3 year
• 5 year
• 10 year

Pricing

• Last price
• Currency
• Change $
• Change %

Income & distributions

• Ex-date
• Amount per share
• Pay date
• Payment frequency
• Dist. yield
• Distribution yield as of
• SEC yield
• SEC yield as of
• Est. annual income

Fund metadata

• Exp ratio (net)
• Exp ratio (gross)
• Morningstar category

Equity classification

• Sector
• Industry
• Industry group
• Sub industry

Fidelity appends a duplicate summary block at the end of each row repeating cost basis, gain/loss, last price, and change (Change $ appears twice). These duplicate columns are preserved as-is alongside the primary columns.

---

Contract Files

Each script is paired with a YAML contract file that externalizes cleanup rules, footer/disclaimer detection markers, output policy, and validation constraints. Contracts are not shared between scripts.

Contract	Used by
fidelity_csv_to_markdown.yaml	fidelity_csv_to_markdown.py

To adapt the pipeline to a different CSV layout or add cleanup rules, update the contract — the script logic should not need to change.

---

Privacy Note

Fidelity exports contain account numbers and position details. Do not commit CSV exports or generated markdown to version control. Both *.csv and *.md output files are gitignored by default (README.md and CLAUDE.md are excluded from that rule).

This site is open source. Improve this page.