My Fund Manager

Problem

When tracking a set of public or private fund managers, a single fund page is hard to use for questions like style similarity, viewpoint drift, and recent holding changes. This skill localizes manager rosters, holdings, performance, and quarterly commentary into JSON/JSONL files, making manager-level comparison, refresh, and export easier, and turning scattered quarterly signals into searchable manager profiles.

How it works

It maintains up to 100 managers around data/roster.json:
- Use manager_search.py to search public funds and AMAC private products, then add confirmed candidates
- Use fetch_holdings.py, fetch_reports.py, and fetch_performance.py to fetch holdings, strategy, and performance
- Use distill_manager.py to extract style DNA and plain-language quarterly views via a conversation LLM or rules engine
- Use smart_update to compare scale, holdings, quarterly reports, and performance, re-distilling only changed managers
Roster operations include deletion, listing, deduplication, and cap validation; a seed library can import notable 2025-2026 managers. Export prefers Excel and falls back to CSV.

Scope and notes

It is for research assistance, not investment advice. Public/private data may lag; private manager names sometimes require searching by manager or firm name. LLM-based distillation depends on model quality, while the rules engine is better for zero-dependency setups. The rules engine outputs a style code and a plain-language summary, which is useful without an external LLM.