| name | ada-research |
| description | Delegate research to Ada (stocks agent) that serves both NPA property analysis and her investment research. Use when NPA-guy needs macro data, bank NPL trends, developer sector health, REIT yields, rate decisions, or any financial/market context that Ada would naturally research for her own stock analysis. Always win-win — the request must align with Ada's investment domain (equities, macro, credit, cross-asset). NOT for generic questions Ada wouldn't benefit from. |
Ada Research Collaboration
Delegate research tasks to Ada that produce knowledge useful for both agents.
Win-Win Principle
Every request must satisfy BOTH:
- NPA-guy needs it — for property valuation, location analysis, or market timing
- Ada benefits from it — the research produces investment-relevant data she'd want in her KB anyway
Ada ingests all research into her ada_kb (LightRAG). NPA-guy queries results directly via SQL or ask_agent.sh.
How to Use
Quick Query (synchronous, ≤120s)
For questions Ada can answer from existing KB knowledge:
bash skills/agent-comm/scripts/ask_agent.sh \
"<message>" \
"/Users/arsapolm/.nanobot-stocks/workspace" \
"/Users/arsapolm/.nanobot-stocks/config.json" \
120
Deep Research (background via spawn)
For research requiring web search and fresh analysis — use spawn tool:
Task: "Run bash skills/agent-comm/scripts/ask_agent.sh '<research brief>' /Users/arsapolm/.nanobot-stocks/workspace /Users/arsapolm/.nanobot-stocks/config.json 600"
Set timeout to 600s for research tasks.
Direct SQL Query (fastest)
Query Ada's KB directly when you know she has the data:
psql -d ada_kb -t -A -c "SELECT content FROM lightrag_doc_full WHERE content ILIKE '%<keyword>%' LIMIT 3;"
Research Brief Template
When requesting research, use this structure:
[NPA-guy → Ada Research Request]
TOPIC: <what to research>
WHY ADA CARES: <how this connects to her investment domain>
WHAT I NEED: <specific data points NPA-guy needs>
FORMAT: Structured summary with tickers, numbers, dates
ACTION: Research + ingest to your KB. I'll query results directly.
TIP: If yfinance returns no data for a ticker, use your google-finance skill as fallback.
Approved Research Topics
See references/topics.md for the full catalog of win-win research topics with example briefs.
Topic Categories
| Category | NPA-guy Gets | Ada Gets |
|---|
| Bank NPL trends | Foreclosure supply forecast | Bank stock picks, credit cycle signals |
| Developer health | Market direction, NPA pipeline | Developer stock valuations, sector rotation |
| BOT rate decisions | Mortgage affordability, buyer demand | Rate-sensitive stock plays, bond impacts |
| REIT performance | Rental yield benchmarks | REIT valuation, income stock picks |
| Macro outlook | Market timing, risk assessment | Asset allocation, cross-asset signals |
| EEC/industrial | Land value trajectory | Industrial/energy stock picks |
| Insurance/credit | Foreclosure buyer financing trends | Financial sector analysis |
Workflow
- Check Ada's KB first —
psql ada_kb to see if she already has the data
- If stale or missing — Send research brief via
ask_agent.sh
- Ada researches + ingests — she builds entities/relations in her knowledge graph
- Query results — pull specific data from
ada_kb for property analysis
- Optionally cross-ingest — key findings can also go into NPA KB (use
cli_ingest.py to avoid lock issues)
Limitations
- Ada needs 30-60s startup time + research time — use
spawn for anything complex
- Single-message, no multi-turn — each call is stateless
- Requests must align with Ada's investment focus (see references/topics.md)
