The 5 things AI Chat can do - tool-calling capabilities

Why tool-calling matters

A bare chat model can only answer from what it was trained on. For a stock platform, that's useless: by the time the model knows about Reliance's Q3 results, it's old news.

Tool-calling solves this. When the assistant needs current data, it fetches it live and gets back fresh numbers. The conversation continues with the new data woven into the answer.

You don't see the tool calls directly; they happen in the background. What you see is an answer with actual current numbers in it.

The five tools

1. get_stock_detail

Fetches current price, fundamentals, technicals, and signals for a stock.

Use case: "What's HDFC Bank's PE right now?" The assistant calls this tool, gets the latest PE, returns the number in context.

Returns: price, day change, volume, RSI, MACD, SMAs, ATR, RSI status, conviction band/stance, currently-firing signals.

2. get_financials

Fetches income-statement history (quarterly or annual).

Use case: "Show me TCS's revenue and PAT trajectory over the last 8 quarters." The assistant calls this tool with quarters: 8, summarises the trajectory.

Returns: revenue, EBITDA, operating margin, PAT, EPS, YoY and QoQ growth rates for each period.

3. get_peers

Returns industry and market-cap peers with comparable metrics.

Use case: "How does Maruti compare to other auto peers on profitability?" The assistant calls this tool, gets the peer table, formats the comparison.

Returns: list of peers with price, returns, PE, PB, ROE, market cap.

4. get_market_overview

Fetches index levels, breadth, FII/DII flows, and sectoral performance.

Use case: "What's the market state today?" The assistant calls this tool, summarises the snapshot.

Returns: Nifty/Sensex levels, A/D ratio, FII/DII net flows, top-three and bottom-three sectors.

5. search_stocks

Looks up stocks by name or symbol fragment (returns top 5 matches).

Use case: "Tell me about ITC." If the model isn't sure whether you mean ITC Ltd or ITC Hotels, it calls this tool to disambiguate.

Returns: matching symbols with company names.

What this enables

The combination of these tools means the assistant can:

• Answer real-time questions without hallucinating numbers
• Compare stocks with current data instead of static training-time data
• Surface upcoming events by querying recent filings (via get_stock_detail)
• Disambiguate when you reference a stock by partial name

What it still can't do

These tools don't include:

• Chart drawing: the assistant can describe technicals but not draw on the chart
• Trade execution: read-only by design
• Custom screen building: you'd run the screener for that
• Historical price queries beyond what the tools return: for full price history, the Chart tab is the right surface

A model + tools example

A user asks: "Compare HDFC Bank and ICICI Bank on growth and profitability over the last year."

The assistant likely does this:

1. Calls get_stock_detail for HDFCBANK → gets current conviction, ratios, technicals
2. Calls get_stock_detail for ICICIBANK → same
3. Calls get_financials for HDFCBANK with years: 1, quarterly: true → gets revenue/PAT trajectory
4. Calls get_financials for ICICIBANK → same
5. Synthesises the answer with growth rates, profitability comparisons, and a one-line summary

You see only the final answer. The five tool calls happened in milliseconds in the background.