evidion 3.0.0

Open Research Intelligence for the terminal

EVIDION

Open Research Intelligence for the terminal.

Evidion turns a question into a transparent research process. It starts by showing you a research plan. You can approve it, add constraints, or cancel. Only after approval does it search and write a source-linked report.

Evidion is free and open source under the MIT License.

What it does today

1. Understands a question and proposes a plan before searching.
2. Searches a SearXNG instance that you control.
3. Safely reads public HTML sources; local/private network destinations, unsafe redirects, and non-HTML responses are blocked.
4. Saves sources, extracted source statements, possible conflicts, research gaps, and the confirmed plan in a resumable session.
5. Exports Markdown, text, HTML, and JSON reports.
6. Uses local or API-backed model providers for structured plan generation.

The evidence checks are deliberately conservative. A source-reported claim is not an independently verified fact, and a possible conflict is a signal to inspect—not an automatic verdict.

Install

Evidion supports Python 3.8 or newer on macOS, Windows, and Linux. A current Python release is recommended, but the lower baseline keeps Evidion usable on more existing computers.

From PyPI

Once the first PyPI publication is complete:

python3 -m pip install --user evidion

From GitHub now

The released source can already be installed directly:

python3 -m pip install --user "git+https://github.com/EvidionAI/evidion.git@v1.0.0"

After either installation, start Evidion from any directory:

evidion

If the command is not found, add Python's user scripts directory to your shell PATH. Typical locations are:

# macOS (replace 3.9 with your Python minor version)
export PATH="$HOME/Library/Python/3.9/bin:$PATH"

# Linux
export PATH="$HOME/.local/bin:$PATH"

# Windows PowerShell
$env:Path += ";$env:APPDATA\\Python\\Python38\\Scripts"

For development:

git clone https://github.com/EvidionAI/evidion.git
cd evidion
python3 -m pip install -e .

Quick start

Web research needs a SearXNG instance with JSON output. If you run one locally, configure it once:

evidion config set --searxng-url http://localhost:8080

Then ask a question:

evidion research "Are local AI models useful for small companies?" --mode normal

Evidion displays the plan first. Press Enter to approve, write add to include constraints such as “focus on Switzerland”, or write cancel to stop before any web search happens.

Commands at a glance

Command	What it does
evidion	Opens an interactive prompt for several questions in one session.
evidion plan "QUESTION"	Shows a plan only; it never searches the web.
evidion research "QUESTION"	Shows a plan, waits for approval, researches, and writes reports.
--mode fast	Uses the smallest bounded search pass for a quick overview.
--mode normal	Uses a broader, multi-round search and follows up on basic coverage gaps.
--mode deep	Raises limits; fully dynamic deep research is planned for a later release.
--max-pages N	Caps the number of pages Evidion may read.
--export all	Writes Markdown, text, HTML, and JSON reports.
--output PATH	Stores reports and session data in a chosen directory.
evidion sessions	Lists saved research sessions.
evidion resume latest	Continues the latest incomplete session.
evidion config show	Shows the active model, storage, and search configuration.

See the command guide for complete examples.

Model providers and API keys

Evidion supports local Ollama and LM Studio, plus OpenAI, Anthropic, Google Gemini, OpenRouter and several OpenAI-compatible APIs. API keys are stored in the operating system credential store, not in the configuration file: macOS Keychain, Windows Credential Manager, or the configured Linux credential backend. To ensure a key is never saved, select environment-only mode.

evidion providers
evidion keys set openrouter
evidion models openrouter --free
evidion doctor
evidion config set --key-storage environment

Read the provider guide before adding a provider key.

Where files are stored

By default Evidion stores data under ~/Documents/Evidion/:

~/Documents/Evidion/
├── Reports/     # Markdown, text, HTML, and JSON exports
└── Sessions/    # Resumable research state

Set EVIDION_HOME to choose another data directory, or EVIDION_CONFIG to use a different configuration file. The default configuration file is platform-native:

• macOS: ~/Library/Application Support/Evidion/config.json
• Windows: %APPDATA%\\Evidion\\config.json
• Linux: $XDG_CONFIG_HOME/evidion/config.json (or ~/.config/evidion/config.json)

Privacy and safety

• Search queries go only to the configured SearXNG instance.
• Web content is treated as untrusted data, never as instructions.
• The reader refuses local/private addresses, unsafe redirects, unsupported protocols, and non-HTML content.
• No API keys are included in web-research prompts.

See SECURITY.md and docs/security.md.

Publishing to PyPI

The distribution name is evidion. The official PyPI API returned 404 for that name on 24 July 2026, so it appears available now; a name is only claimed by a successful first upload. The project is prepared for passwordless Trusted Publishing from GitHub Actions. Follow the PyPI guide to make pip install evidion available.

Contributing

Issues and pull requests are welcome. Please read CONTRIBUTING.md, the Code of Conduct, and the architecture overview.

License

MIT