Workflow All workflows API reference

Discovery — vague theme to shortlist

The funnel: federated search for entities tied to a keyword → mf_screener or co_screener to narrow by metric → concall_search to confirm the thesis is alive in management voice. Worked through with “Indian uranium / nuclear” as the running theme.

Tools used · search · mf_screener · co_screener · concall_search · capabilities for grammar discovery
Total credit cost · ~20 credits to go from theme to defendable shortlist

export GENKA_API_KEY=imk_live_...
H="X-API-Key: $GENKA_API_KEY"

1. Start with the broadest possible search

Federated search across funds + stocks + sectors + themes. One call returns grouped hits so you can see what entities actually exist for the keyword before narrowing.

curl -sH "$H" "https://genka.dev/latest/search?q=uranium+nuclear" | jq '.data.groups | map({type: .type, n: (.items | length), top: [.items[:3][].name // .items[:3][].slug]})'

Returns up to 4 groups: fund, stock, sector, theme. For uranium / nuclear: typically 0 fund matches (no Indian uranium-themed fund exists today — that's information), 2-3 stocks (TATAPOWER, NTPC nuclear partnerships), zero direct sector hits. Negative result is information — your thesis needs a different vehicle.

2. What's the closest legitimate theme?

If the literal keyword has zero hits, fall back to the canonical theme grammar. capabilities lists the full enum.

curl -sH "$H" "https://genka.dev/latest/capabilities" | jq '.data.mf.themes // .data.themes'

Returns the known themes list. For natural-resource / energy adjacency: energy, commodities. Mining / uranium / agri specifically don't exist as separate themes today (gap tracked in TODO). Pivot: mf_screener?theme=energy + filter results client-side for the names you care about.

3. Narrow by metric

Now you have a candidate set. Apply quantitative filters: Sharpe, max drawdown, expense ratio, AUM size.

curl -sH "$H" "https://genka.dev/latest/mf/screener?theme=energy&period=3y&sort=-sharpe_absolute&limit=10" \
  | jq '.data.items[] | {slug, sharpe: .metrics.sharpe_absolute, max_dd: .metrics.max_dd_absolute}'

Returns the energy-themed funds ranked by 3y Sharpe. DSP NRNE typically tops; ICICI Pru Commodities, SBI Magnum Comma sit in the same bucket. Apply additional filters via the filters param: {"sharpe_absolute_gte": 0.8, "max_dd_absolute_lte": -0.35} drops the volatile names.

4. Same idea on the stock side

If you want individual stocks rather than fund wrappers, co_screener applies the same filter grammar to NSE 500.

curl -sH "$H" -X POST -H "Content-Type: application/json" \
  -d '{"sector":"Power","filter":{"roe":{"op":"gte","value":15},"pe":{"op":"lte","value":25}},"limit":10}' \
  https://genka.dev/latest/screener

Returns power-sector companies with ROE ≥ 15% AND P/E ≤ 25 — the cheap-and-quality filter for the nuclear-adjacent thesis. NTPC, Power Grid typically clear; private players might not. Use capabilities to discover all valid metric columns + ops.

5. Confirm the thesis is alive in management voice

A shortlist isn't defendable until you've heard the names speak about it. Concall search confirms the narrative + dates the thesis arc.

curl -sH "$H" "https://genka.dev/latest/concalls/search?q=nuclear+power+capacity&limit=5"

Returns ranked hits with snippets and days_old. If your shortlist names appear in 2025-26 transcripts with growing prominence, the thesis is live. If the most recent hit is 2022 and they've stopped talking about it, the narrative has rolled over — revisit. Days_old is the most useful field after score.

6. Drill into the top candidate

Once you have a shortlist of 1-3 names, switch to the equity research workflow on each.

# for funds: pull holdings + metrics + manager commentary
curl -sH "$H" "https://genka.dev/latest/mf/dsp-natural-resources-and-new-energy/holdings"
curl -sH "$H" "https://genka.dev/latest/mf/dsp-natural-resources-and-new-energy/metrics?period=3y"

# for stocks: ratios + concalls + holders
curl -sH "$H" "https://genka.dev/latest/companies/NTPC/ratios?metric=pe"
curl -sH "$H" "https://genka.dev/latest/concalls/search?q=capex&symbol=NTPC&limit=5"
curl -sH "$H" "https://genka.dev/latest/companies/NTPC/holders?limit=10"

The equity research workflow takes over from here. The discovery flow above is what gets you to a shortlist of names worth reading deeply — the deep read is a separate compose pattern.

Other workflows: equity research (one stock, deep) · portfolio review (what you already own) · macro watching (FII/DII + concall macro tone).