Why jevbrief

Jev turns state into typed decisions. It does its best work on small, relevant state, and it is weak at raw numbers, dates, and long lists of irrelevant detail. Real sources are the opposite: a web page has hundreds of elements, an agent can call a hundred tools, and a failed build prints thousands of log lines.

jevbrief sits between your source and Jev. It keeps what matters, drops the rest with a reason for every drop, asks Jev one clear question, and records the whole decision so you can replay it.

Quick start

pip install jevbrief          # tools, ci, pr, otel, and json; add [web] or [nes], or [all] for everything

Set TYPESAFE_API_KEY (get one at console.typesafe.ai) in your environment or a .env file. inspect needs no key and costs nothing. ask calls Jev and --view replays the decision:

gh run view <run id> --log-failed > run.log
jevbrief inspect run.log --adapter ci --goal "CI is red on main"
jevbrief ask     run.log --adapter ci --goal "CI is red on main" --view

27 log lines -> 7 failures -> 3 sent to Jev
likely cause: Run pytest -q: E KeyError: 'currency_code'   (confidence 0.84)
looks flaky: no (0.07)

Building an agent? One call narrows any mix of tools to the ones that fit the current step, and returns your own objects. The tools can be your functions, MCP servers, LangChain, CrewAI, OpenAI, or Anthropic tools. Another call tells you when the agent is going in circles.

from jevbrief import select_tools, pick_tool, check_progress

llm.bind_tools(select_tools(tools, goal))     # ranking only: local, free, no API key
pick = pick_tool(tools, goal)                 # Jev picks one: pick.tool, pick.confidence
check_progress(history, goal).stuck          # is the agent going in circles?

Adapters

Each adapter's page covers its quick start, input, reason codes, options, and limits. Want another source? Suggest it or build it.

Benchmarks

Both arms use the same Jev and the same question, with three runs per task and median input tokens. The raw arm sends what a naive integration would send.

Not every result favors jevbrief:

• On ci's second question, whether the failure is flaky, jevbrief scored 67% against 93%.
• On pr, jevbrief tied the full diff on accuracy. It only saved tokens.
• The synthetic sets were built alongside the adapters.

See how the benchmarks work for the method and caveats.

Contributing

• Suggest a source: open an adapter request.
• Build an adapter: follow the guide.
• Report a wrong decision: open a bug report, and attach the trace.

git clone https://github.com/parthkomalwad/jevbrief && cd jevbrief
python -m venv .venv && . .venv/bin/activate    # Windows: .venv\Scripts\activate
pip install -e ".[dev]" && pytest -q

License

MIT. Community project, not affiliated with TypeSafe AI.