EvalLane

Lightweight eval harness for agent outputs — contains/regex/length/JSON-path assertions and suite pass rates.

Part of Talocode. Hosted at /v1/evallane/*.

| | | |--|--| | npm | npm install @talocode/evallane | | CLI | evallane | | Cloud | https://api.talocode.site/v1/evallane/* | | Auth | TALOCODE_API_KEY | | Repo | https://github.com/talocode/evallane | | License | MIT |

Why EvalLane?

You cannot improve agents without measurement. EvalLane runs deterministic suites so regressions are visible.

Install

npm install @talocode/evallane
# PyPI later: pip install talocode-evallane

Usage

import { runSuite, scoreOutput } from "@talocode/evallane"
const r = runSuite({ cases: [{ name: "greet", output: "hello", assertions: [{ type: "contains", value: "hello" }] }] })

Talocode ecosystem

| Project | Role | |---------|------| | RetryLane | Tool retries & circuit breakers | | PromptLane | Prompt versioning & feedback | | EvalLane | Assertion eval suites | | PolicyLane | Allow/deny + redaction | | VerifyLane | Deterministic code verification | | TraceLane | Agent run tracing | | HandoffLane | Schema handoffs | | StyleLane | Constitution packs | | SearchLane | Agent web search | | StackLane | Cloud keys & credits |

More: github.com/talocode · docs.talocode.site

License

MIT © Talocode