elizaos--eliza/packages/benchmarks/action-calling/AGENTS.md

Action Calling — Agent Guide

Native function/tool-calling benchmark. Samples planner-style records from training/data/native/records/hermes-fc-v1.jsonl, sends OpenAI-compatible tools to the model, and scores the returned tool_calls on five axes. Registered in the suite registry as action-calling.

Run

# Direct, from the repo root (packages/benchmarks/)
python -m benchmarks.action-calling.cli \
    --provider vllm \
    --model eliza-1-9b \
    --out /tmp/action-calling-out

# Through the suite orchestrator (resolves provider/model, stores results)
python -m benchmarks.orchestrator run \
    --benchmarks action-calling \
    --provider vllm \
    --model eliza-1-9b

Smoke test (no API keys)

The mock provider echoes expected tool calls back, scoring 1.0 on all axes. Falls back to fixtures/smoke.jsonl automatically when the full dataset is absent.

python -m benchmarks.action-calling.cli \
    --provider mock \
    --model smoke \
    --out /tmp/action-calling-smoke

Test the harness

pytest packages/benchmarks/action-calling/tests/ -v

Layout

Path	Role
cli.py	CLI entrypoint and scoring logic
fixtures/smoke.jsonl	Minimal fixture record for mock/offline runs
tests/test_action_calling_cli.py	pytest suite for scoring helpers

Notes

• Results write to <out>/action-calling-results.json (path controlled by --out).
• Scored by _score_from_action_calling_json in registry/scores.py.
• Score = geometric mean of five sub-rates: native_tool_calls_ok, tool_name_match, args_parse_ok, required_keys_ok, arguments_match.
• Supports providers: vllm, openai, groq, openrouter, anthropic, cerebras, eliza, hermes, openclaw, mock.
• Harness selection (eliza/hermes/openclaw/smithers) can also be forced via ELIZA_BENCH_HARNESS or BENCHMARK_HARNESS env vars.
• Full background: README.md.

⛔ NON-NEGOTIABLE — evidence, trajectories & real end-to-end tests

The binding, repo-wide standard is AGENTS.md. Read it. Nothing in this package is done until it is proven done — a reviewer must confirm it works without reading the code, from the artifacts you attach. This applies to every feature, fix, refactor, and chore here. "Tests pass" is not proof; "CI is green" is not proof.

• Record AND read model trajectories. Capture the actual inputs and outputs of the model from a live LLM — not the deterministic proxy, not a mock: the prompt, the providers/context, the raw model output, every tool/action call, and the result. Then open the trajectory and review it by hand. A captured-but-unread trajectory is not evidence (packages/scenario-runner/bin/eliza-scenarios run <scenario> --report <out>).
• Real, full-featured E2E — no larp. Every feature ships detailed end-to-end tests that drive the real path end to end. Not the happy "front door" only: cover error paths, edge/empty/invalid input, concurrency, roles/permissions, and adversarial input. A test that asserts against a mock/stub/fixture standing in for the thing under test does not count. If the real model/device/chain/connector/account is hard to reach, make it reachable — that is the work, not an excuse to mock. If the existing tests here are shallow or mocked, fixing them is part of your change.
• Screenshots + logs at every phase, plus a complete walkthrough video/run-through of the entire feature or view, start to finish (bun run test:e2e:record).
• Manually review every artifact the change touches — never just the green check: client logs (console + network), server logs ([ClassName] …), the model trajectories in and out, before/after full-page screenshots, and the domain artifacts listed below for this package.
• No residuals. No shortcuts. The goal is not "done" — it is everything done. Clear every blocker by the hard path: build the real architecture, stand up the real model/device/service, actually test it. Never leave a TODO, a stub, a stepping-stone, or a "follow-up." When unsure, research thoroughly, weigh the options, and ship the best, highest-effort, production-ready version. Keep going until every possibility is exhausted.

Artifacts → attached inline in the PR (MP4 video, JPG screenshots, logs in <details>); attach each evidence type or explicitly mark it N/A with a reason — never leave it blank. If develop moved and changed behavior, re-capture evidence; stale proof is worse than none.

Capture & manually review for this package — benchmark / eval suite:

• A real-model run (not the mock/smoke fixture) producing the score-report JSON, with the numbers inspected and the provider/model recorded.
• The per-item trajectories the harness captured, spot-reviewed for correctness — a green harness run over mock fixtures is not a result.
• The provider matrix actually exercised, and the scoring math validated against a known case.
• Failure / timeout / partial-output handling in the harness itself.