Practical Ambient AI skill for Hermes.
This project packages a product-style /ambient workflow that returns:
- the model answer
- Ambient provenance summary
merkle_root- validator / auction metadata
- a clear boundary that provenance is not the same as factual truth verification
Use this when you want a normal LLM answer plus visible Ambient inference provenance.
Examples:
- product / UX questions
- research prompts
- technical explanations
- support workflows
- accessibility reviews
- scheduled Hermes tasks where recurring prompts should return both the answer and provenance context
scripts/ambient-runtime.js— live/ambientruntime against Ambient APIscripts/ambient-verify.js— local structural verifier for receipt / output / claimstests/ambient-runtime.test.js— runtime teststests/ambient-verify.test.js— verifier tests.env.example— required env structure without secretsSKILL.md— Hermes skill instructions
Copy .env.example to .env and fill in your real values locally.
Required structure:
AMBIENT_API_KEY=
AMBIENT_BASE_URL=https://api.ambient.xyz/v1
AMBIENT_MODEL=zai-org/GLM-5-FP8
AMBIENT_FALLBACK_MODEL=ambient/large
AMBIENT_TIMEOUT_MS=60000Note: the real .env is intentionally not included in this repository.
node scripts/ambient-runtime.js --prompt "Explain Ambient proof of logits" --format chatnode scripts/ambient-runtime.js --prompt "Explain Ambient proof of logits" --format jsonnode scripts/ambient-verify.js \
--text examples/sample-output.txt \
--receipt examples/sample-receipt.json \
--claims examples/sample-claims.json \
--format chatnode --test tests/ambient-verify.test.js tests/ambient-runtime.test.jsThis repository contains a working v1 of the skill:
- live Ambient API path works
- retry flow is built in
- UI-like verification block includes validator and auction context
- no secrets are committed