mostlyrightmd-finance

Finance data for the mostlyright Python SDK — earnings-call transcript capture, streaming transcription, append-only fact ledgers, and prediction-market earnings-mention resolution (Kalshi KXEARNINGSMENTION{TICKER} + Polymarket "will say "). Local-first on-device capture + STT with a legal audio firewall: captured audio is a transient artifact — only transcript text and derived facts survive. Imports as mostlyright.finance.

The transcript engine, hosted and live adapters, SSE stream, Kalshi/Polymarket resolver, and point-in-time helpers live under mostlyright.finance.transcripts.

Install

pip install mostlyrightmd-finance                 # base surface (ledger + resolver)
pip install 'mostlyrightmd-finance[transcripts]'  # + faster-whisper + av (capture/STT)

Surface

The pipeline lives at mr.finance.transcripts:

• Engine — CaptureAdapter / Q4CaptureAdapter, StreamingTranscriber / SttTranscriber, SegmentBus, and the append-only FactLedger / TranscriptLedger.
• Adapter + stream — the EarningsAdapter consumer/parser and the SSE consume_sse reader.
• Market resolution — resolve (Kalshi / Polymarket earnings-mention), derive / detect_provider, the codegen registries + calendar seed, and the point-in-time helpers (knowledge_time = transcript availability).

mr.finance exposes no training_table() — there are no settlement labels yet. Use the transcript rows and point-in-time helpers directly for research.

Docs

See docs/finance-transcripts.md and https://mostlyright.md/docs/sdk/ for the quickstart and the full API reference.