How it works

1. Company Topic News

UI: / · Topic News · API: POST /api/news/{ticker}

Pick a ticker (or several). The app resolves it to a Bigdata entity ID, then runs one natural-language query per editable topic. Each topic template contains {company}, which is replaced with the company name before search.

• Scoped to that company via an entity filter
• Keeps positive and negative sentiment (drops neutral)
• Sources: public news and transcripts
• Optional AI query expansion multiplies each topic into more phrases
• UI relevance dropdown drops weak hits after results return

Search payload (one call per topic):

{
  "query": {
    "text": "Apple Inc. reported earnings results beating or missing revenue and profit expectations",
    "filters": {
      "timestamp": { "start": "...", "end": "..." },
      "entity": { "all_of": ["<entity_id>"] },
      "category": {
        "mode": "INCLUDE",
        "values": ["news_public", "transcripts"]
      },
      "sentiment": {
        "values": ["positive", "negative"]
      }
    },
    "max_chunks": 10
  }
}

2. Company Negative News

UI: / · Negative News · API: POST /api/news/{ticker} with negative_news: true

Same company desk, different question: show strongly negative public news where the company appears in the headline, grouped by risk categories (litigation, regulatory, credit stress, governance, cyber / product).

• Filter-only search — no query text
• search_mode: "fast"
• Sources: public news and transcripts (news_public, transcripts)
• Entity matched in the headline (avoids drive-by body mentions)
• Sentiment score in range [-1, -0.1]
• Bigdata taxonomy topics in topic.any_of (editable per category)
• One entity ID per request (clean ticker attribution on multi-ticker runs)

Search payload (one call per category × company):

{
  "search_mode": "fast",
  "query": {
    "filters": {
      "timestamp": { "start": "...", "end": "..." },
      "category": {
        "mode": "INCLUDE",
        "values": ["news_public", "transcripts"]
      },
      "entity": {
        "search_in": "HEADLINE",
        "any_of": ["<entity_id>"]
      },
      "sentiment": {
        "ranges": [{ "min": -1, "max": -0.1 }]
      },
      "topic": {
        "search_in": "ALL",
        "any_of": [
          "society,legal,fraud,,",
          "business,regulatory,regulatory-investigation,,"
        ]
      }
    },
    "max_chunks": 10
  }
}

3. Sector / Commodity News

UI: /sector · API: POST /api/sector-news

Commodities desk: pick Energy, Power, Shipping & Freight, or Metals. Each desk ships with editable theme phrases. Every phrase is searched verbatim — there is no company / entity filter, so the feed follows the theme, not a single issuer.

• Natural-language query.text (desk phrase as written)
• No entity filter, no taxonomy topic filter
• Positive + negative sentiment (neutral excluded)
• Sources: public news and transcripts
• Optional AI expansion uses a desk-specific prompt

Search payload (one call per theme phrase):

{
  "query": {
    "text": "Brent and WTI crude oil prices rise or fall on shifting supply and demand outlook",
    "filters": {
      "timestamp": { "start": "...", "end": "..." },
      "category": {
        "mode": "INCLUDE",
        "values": ["news_public", "transcripts"]
      },
      "sentiment": {
        "values": ["positive", "negative"]
      }
    },
    "max_chunks": 20
  }
}

Quick compare