Finance OSINT is a community-driven research database for public-market intelligence.
The project exists to help narrow the research access gap for retail investors, independent analysts, and open-source researchers. Valuable market research is scattered across filings, transcripts, news, company materials, obscure datasets, field notes, first-hand observations, and contributor-sourced findings. Finance OSINT turns legally shareable evidence into durable records that can be checked, connected, debated, and improved in public.
The repository is the database: contributors add YAML records for provenance, evidence, structured observations, graph relationships, review pressure, and proposed theses, then validate them locally and submit changes by pull request.
Finance OSINT preserves evidence and analysis, not hidden authority. It does not store canonical truth status, confidence scores, hidden moderation decisions, material non-public information, private expert-network material, or private agent state. Review state is derived locally from the records.
records/ Canonical source, evidence, metric, event, claim, relationship,
validation, challenge, question, entity, dataset, and thesis data.
ontology/ Registered and proposed claim predicates, relationship types,
and metric definitions.
schemas/ JSON Schemas for every canonical record kind.
templates/ Copyable YAML templates.
fosint/ The `fo` command-line tool.
scripts/ CI and maintainer validation helpers.
docs/ User and contributor documentation.
examples/ Fictional example corpora and deferred prototypes.
tests/ Unit tests and synthetic fixtures.
Only records/ and ontology/ are canonical data. Files under .local/,
examples/, and tests/fixtures/ are generated, illustrative, or test-only.
Source -> Dataset -> Evidence -> Metric / Event / Claim -> Relationship -> Thesis
-> Validation / Challenge / Question
Start from sources and evidence. Promote stronger objects only when the support chain justifies them. If the chain is incomplete, record the gap as a question, challenge, scoped claim, or thesis limitation rather than stretching the data.
Install the local tooling:
uv syncValidate and build local read artifacts:
uv run fo lint --json
uv run fo index build --json
uv run fo graph build --jsonSearch and inspect records:
uv run fo search axt --json
uv run fo context entity:company:axt-inc --json
uv run fo review thesis:axt-sumitomo:inp-substrate-bottleneck-watch --chain --json
uv run fo graph neighbors entity:company:axt-inc --jsonGenerated files are written under .local/ and are ignored by git.
If you cannot use uv:
python3 -m venv .venv
. .venv/bin/activate
pip install -e .
fo lint --jsonAtomic contributions are welcome. A useful pull request can add one source, one piece of evidence, one metric, one claim, one relationship, one question, one challenge, or one validation.
Before opening a pull request, run:
uv run fo lint --json
uv run python -m unittest discover -s tests
uv run fo index build --json
uv run fo graph build --json
uv run fo diff-review BASE --json
uv run fo view build BASE --jsonUse BASE as the branch point or pull-request base, for example main or
origin/main.
New canonical records must use your GitHub identity:
submitted_by: github:your-loginSee CONTRIBUTING.md and docs/research-batch-contribution.md for the full
workflow.
Start with docs/README.md.
The stable docs cover:
- first-time setup and first PRs
- the record model and ontology
- how to author source-backed research batches
- CLI commands and schema contracts
- review, governance, archive, and dispute handling
- maintainer checks and generated artifacts
Finance OSINT is not an investment recommendation service, a hosted database, a moderation backend, a private expert network, or an agent memory store. The repo keeps the evidence chain visible so readers can evaluate the work themselves.
Finance OSINT is licensed under the Apache License, Version 2.0. See LICENSE.