Finance Developer Assistant

Review committed changes, reproduce findings and inspect readiness evidence for finance software, using your chosen model or deterministic offline analysis.

Local-first, no telemetry. The default provider is offline. The demo and local dashboard make no external requests. Remote model reviews and GitHub inspection require explicit configuration and approval. The dashboard ships prebuilt and binds only to 127.0.0.1; Node is needed only for frontend development.

What it does

• Review source: freeze commits, screen paths/content, inspect coverage and use configurable finance, research, quant or custom review profiles.
• Reproduce: run approved recipes on screened snapshots in a pinned, network-disabled container; trusted local execution requires explicit opt-in.
• Propose carefully: inspect and approve exact patch hashes, protect policy files and preserve guarded recovery copies. No automatic merge, deploy or trade.
• Track evidence: local spend reservations, usage, findings, readiness and a tamper-evident audit trail; unknown outcomes stay visible and block readiness.
• Configure locally: preview a diff, confirm a save, inspect statistics and run Health checks from the browser. API keys stay in environment variables.

Start in one minute

Python 3.11+ is required; Git is needed for source workflows. After the first PyPI release is published, install with pip or pipx. The distribution is finledgerpoli; the command and import package remain finassist, with finledgerpoli also available as a command alias.

pip install finledgerpoli
# Alternatively, use an isolated command environment:
pipx install finledgerpoli
finassist demo
finassist doctor
finassist init --output finassist.json
finassist --config finassist.json dashboard

Until publication, or to develop from source, clone and install instead:

git clone https://github.com/EleutheroiEdge/finance-dev-assistant.git
cd finance-dev-assistant
python3 -m venv .venv
.venv/bin/python -m pip install .
.venv/bin/finassist demo
.venv/bin/finassist init --output finassist.json
.venv/bin/finassist --config finassist.json dashboard

On Windows PowerShell, use py -3 -m venv .venv, then .\.venv\Scripts\python.exe -m pip install . and .\.venv\Scripts\finassist.exe demo. No activation or execution-policy change is required. macOS/Linux use the commands above. See platform setup for UTF-8, permissions, Git modes and Docker Desktop details.

The demo uses fictional finance checks and needs no key, model download or container. It demonstrates deterministic oracles, not model review quality. Download the short terminal recording and play it locally with asciinema play docs/images/demo.cast.txt; no hosting account is needed.

Dashboard preview

These screenshots come from an isolated synthetic fixture. All displayed spend, tokens, findings and review activity are fictional; no provider was called.

See the dashboard guide for configuration previews, explicit confirmation, audit statistics and same-user security limits.

Review your first change

After inspecting your configuration and source policy:

finassist --config finassist.json plan --repo . --base main --head HEAD \
  --output .finassist/context.json
finassist --config finassist.json review --packet .finassist/context.json \
  --output .finassist/offline-review.json

Offline analysis does not claim a model review or test execution. To use a remote model, choose its provider/model/endpoint, price assumptions and named environment variable; enable network access deliberately and approve the exact outgoing packet. There is no automatic provider fallback. Provider setup explains OpenAI Responses, Anthropic, Gemini, compatible Chat Completions and local Ollama. Compatibility tests use fake transports; not every model/service is live-tested.

Guides

Keep .finassist/ and private artifacts out of Git. Inspect every outgoing packet: content screening cannot prove confidentiality. A readiness result grants no merge, deployment or financial authority. Local audit hashes cannot defeat a malicious same-user rewrite; budget estimates are not provider-enforced billing caps.

Contribute

python -m pip install -r requirements-dev.lock
python -m pip install --no-deps --no-build-isolation -e .
python -m pytest -q
ruff check src tests scripts
finassist demo
finassist scan --repo .

Use synthetic fixtures and fake provider transports. See contributing, architecture and media reproduction. Scorecard is a posture indicator, not certification; results appear after its workflow publishes. The PyPI badge will be added after the owner verifies the first public release. Forks must update badge/repository links and maintainer settings.

MIT licensed. See LICENSE.