Dse Data Acquisition

DSE Data Acquisition

Two MCP servers. Market data lives in PostgreSQL and is exposed by stock-buddy-data MCP. Analysis runs on stock-buddy MCP (no network, no fetch). Never hand-build JSON when the data MCP is available.

Standard workflow

1. Fetch contract — call get_ticker_contract_for_analysis on stock-buddy-data:

   • ticker (required), e.g. LHB
   • include_portfolio: true when risk-manager sector/heat gates are needed
   • ohlcv_days: 260 for full momentum MA stack (minimum 30 for basic technical)

2. Analyze — pass the returned JSON to analyze_ticker on stock-buddy unchanged.

Fallback: DB offline or missing contract fields

If stock-buddy-data fails (Postgres down, MCP disconnected) or _meta.missing is non-empty, or analyze_ticker returns instructions / insufficient_data_contract:

YOU are the data agent (Claude Code, Codex, Gemini, Cursor). Use web search or any credible public source to gather missing contract fields (listed sites in instructions are examples only), assemble the contract, then call analyze_ticker with the full nested object — never ticker-only.

1. Read the instructions block (from data MCP error, _meta.research_instructions, or analyze_ticker response). Note source_selection_policy.
2. Research only the missing fields from any source you find useful — see references/CLIENT_RESEARCH.md.
3. Merge any partial DB contract + researched fields into one JSON.
4. stock-buddy.analyze_ticker(<full contract>).
5. Save citations: upsert_research_sources / upsert_research_memo.

4. Portfolio updates (manual only) — use stock-buddy-data write tools:

   • upsert_position, remove_position, set_account, get_portfolio

5. Save research citations — after web search or manual research, persist sources for audit lineage:

   • upsert_research_sources — URLs, titles, publisher, optional extracted_facts
   • get_research_sources — retrieve by ticker, session, or category
   • Optional promote_shareholding when verified % are extracted from disclosures

6. Save full research memos — persist markdown reports:

   • upsert_research_memo — { title, body_md, summary_json, as_of, session_id }
   • get_research_memo — fetch memo (+ optional linked sources via include_sources)
   • list_research_memos — history for a ticker

Data MCP read tools

Data MCP write tools (research & portfolio)

Ingest (populate database)

Run locally after docker compose up postgres:

npm run db:migrate
npm run db:seed
npm run ingest -- --ticker LHB --job all --days 365
npm run ingest -- --watchlist

Admin MCP tools (trigger_ingest, register_ticker) require STOCK_BUDDY_DATA_ADMIN=1.

Contract shape

Matches skills/README.md: ticker, as_of, mode, ohlcv[], fundamentals, shareholding[], macro, news[], microstructure, optional portfolio + account.

DSE pitfalls

• OHLCV needs ≥30 bars (≥200 for full momentum template).
• Shareholding is monthly and lagging — smart-money caps confidence.
• Portfolio is never scraped — you maintain it via MCP write tools.
• If _meta.missing lists fields, analysis still runs but some skills skip or flag gaps.

Worked example (LHB)

1. stock-buddy-data.get_ticker_contract_for_analysis({ ticker: "LHB", include_portfolio: true })
2. stock-buddy.analyze_ticker(<result from step 1>)

Worked example (save research)

After web research on WALTONHIL shareholding:

stock-buddy-data.upsert_research_sources({
  ticker: "WALTONHIL",
  session_id: "cursor-chat-2026-06-25",
  client_id: "cursor",
  query_context: "WALTONHIL sponsor director institutional foreign public shareholding",
  sources: [
    {
      url: "https://www.tbsnews.net/economy/stocks/walton-chairman-transfer-shares",
      title: "Walton chairman to transfer 3.64cr shares to family members",
      publisher: "The Business Standard",
      category: "shareholding",
      extracted_facts: { event: "director_transfer", shares_cr: 3.64 }
    },
    {
      url: "https://stocknow.com.bd/stocks/WALTONHIL",
      title: "WALTONHIL shareholding position",
      publisher: "StockNow",
      category: "shareholding",
      extracted_facts: { sponsor_pct: 61.09, public_pct: 38.15, as_of: "2025-08" }
    }
  ],
  promote_shareholding: {
    month: "2025-08",
    sponsor: 61.09,
    public: 38.15,
    source: "stocknow_research"
  }
})

Retrieve later: get_research_sources({ ticker: "WALTONHIL", category: "shareholding" })

Worked example (save markdown memo)

stock-buddy-data.upsert_research_memo({
  ticker: "WALTONHIL",
  session_id: "cursor-chat-2026-06-25-waltonhil-equity-memo",
  client_id: "cursor",
  title: "Walton Hi-Tech Industries PLC — Educational Equity Analysis",
  as_of: "2026-06-21",
  body_md: "<full markdown report>",
  summary_json: { price_tk: 384, pe_ratio: 11.5, rating: "cheap_with_governance_overhang" },
  link_all_ticker_sources: true
})

Fetch: get_research_memo({ ticker: "WALTONHIL", include_sources: true })

Output is educational analysis only, not financial advice.