Skip to content

quickstart.md

Latest commit

 

History

History
36 lines (27 loc) · 1.81 KB

quickstart.md

File metadata and controls

36 lines (27 loc) · 1.81 KB
Raw

Quickstart: Story-First IC Copilot

Start the no-dependency app:

python server.py --port 8501

Open the printed local URL and use Load demo before entering API keys. The demo gallery is instant, offline-friendly, and designed for screenshots:

  • AAPL: source-backed thesis.
  • NVDA: neutral-first capex, goodwill, and semiconductor peer investigation.
  • ORCL: cloud backlog and AI capacity versus capex, financing, and free-cash-flow conversion.
  • BABA: ADR/FPI complexity.
  • TSLA: peer metric read-through.
  • GS: financial-sector playbook.
  • SPCX: entity-resolution warning.

The six analyst showcases (AAPL, NVDA, ORCL, BABA, TSLA, and GS) are versioned Deep Initiation fixtures with 20 quarters, five annual reports, 20 calls, Premium-mode provider slots, official macro context, sanitized Wisburg debate maps, market-implied expectations, and citation-guardrailed LLM research metadata. They do not call live providers or retain licensed full report text. Refresh live sources before treating any fixture observation as current investment evidence.

The first screen is the IC story: verdict, thesis, what changed, why it matters, causal bridge, counter-thesis, next action, and the research pipeline. Raw filings, metrics, formulas, and provider diagnostics are kept in drawers and detail tabs.

Contribute Coverage

  • Add a sector playbook: data/sector_kpi_playbooks.csv.
  • Add an ADR profile: data/adr_profiles.csv.
  • Add a metric alias: data/metric_ontology.csv.
  • Add a source adapter with provider health, source tier, citations, timestamps, and licensing policy.
  • Add a demo case with no-network regression coverage.

Every contribution should preserve source citations, point-in-time rules, explicit Unknown values, and the rule that LLM output cannot independently promote an idea.