Repository navigation
Bump @vitest/mocker and vitest - #31
Closed
dependabot[bot] wants to merge 26 commits into
Closed
dependabot[bot] wants to merge 26 commits into
dependabot[bot] wants to merge 26 commits into
Conversation
Embeddable AI governance framework — human-readable trust contracts, runtime policy enforcement, and evidence-grade auditability. Core: - Three-valued policy engine (permit/deny/modify) with 12 operators - Compiled evaluator for sub-millisecond decisions - MCP tool router with fail-closed authorization - Model binding verification, secure boot, graceful degradation - Ed25519-signed human overrides Contracts: - YAML/JSON/Markdown trust contract format with AJV validation - Contract inheritance, registry, changelog, linter - 9 drop-in threat contracts (PII, injection, hallucination, toxicity, cost, leakage, bias, copyright, MCP tool abuse) - Signed policy bundles with DSSE envelopes + BundleWatcher hot-reload - Zero-downtime ContractWatcher for directory-based updates Audit: - SHA-256 hash chain with HMAC signing - Merkle tree seals, proof bundles, distributed tracing - Immutable content-addressed storage with tamper detection - File, webhook, multi-sink, and OpenTelemetry audit sinks - CSV/JSON/summary export Security: - WASM sandbox (memory limits, timeouts) and V8 isolate sandbox - DSSE contract signing + RFC 3161 timestamps - IETF RATS (RFC 9334) remote attestation - Pluggable trust anchors (software Ed25519, TPM shell-out, SGX/Firecracker stubs) - TLA+ formal verification spec generator Integrations: - OpenAI, Anthropic (Claude), and Gemini wrapper examples - Buffer-and-release streaming interceptor for mid-stream PII redaction - LangChain callback handler, Vercel AI SDK wrapper - HTTP middleware, standalone REST API server (24 endpoints) - CLI: validate, lint, evaluate, serve, audit, attest, tlaplus Compliance: - 6 frameworks: EU AI Act, NIST AI RMF, ISO 42001, SOC 2, GDPR, HIPAA - Multi-tenant isolation 48 source files, 35 test files, 504 tests, 4 dependencies, Node ≥18.
- package.json: add repository, homepage, bugs, author fields - package.json: expand keywords for npm discoverability (19 terms) - package.json: include examples/ in published files - package.json: prepublishOnly runs tests before build - .github/workflows/publish.yml: auto-publish on GitHub release with provenance - .npmignore: exclude source, tests, dashboard, charts from npm tarball
Bumps [brace-expansion](https://github.com/juliangruber/brace-expansion) from 5.0.4 to 5.0.7. - [Release notes](https://github.com/juliangruber/brace-expansion/releases) - [Commits](juliangruber/brace-expansion@v5.0.4...v5.0.7) --- updated-dependencies: - dependency-name: brace-expansion dependency-version: 5.0.7 dependency-type: indirect ... Signed-off-by: dependabot[bot] <support@github.com>
Bumps [postcss](https://github.com/postcss/postcss) from 8.5.8 to 8.5.25. - [Release notes](https://github.com/postcss/postcss/releases) - [Changelog](https://github.com/postcss/postcss/blob/main/CHANGELOG.md) - [Commits](postcss/postcss@8.5.8...8.5.25) --- updated-dependencies: - dependency-name: postcss dependency-version: 8.5.25 dependency-type: direct:development ... Signed-off-by: dependabot[bot] <support@github.com>
Bumps [fast-uri](https://github.com/fastify/fast-uri) from 3.1.0 to 3.1.5. - [Release notes](https://github.com/fastify/fast-uri/releases) - [Commits](fastify/fast-uri@v3.1.0...v3.1.5) --- updated-dependencies: - dependency-name: fast-uri dependency-version: 3.1.5 dependency-type: indirect ... Signed-off-by: dependabot[bot] <support@github.com>
Bumps [postcss](https://github.com/postcss/postcss) from 8.5.8 to 8.5.26. - [Release notes](https://github.com/postcss/postcss/releases) - [Changelog](https://github.com/postcss/postcss/blob/main/CHANGELOG.md) - [Commits](postcss/postcss@8.5.8...8.5.26) --- updated-dependencies: - dependency-name: postcss dependency-version: 8.5.26 dependency-type: indirect ... Signed-off-by: dependabot[bot] <support@github.com>
…8.5.26 Bump postcss from 8.5.8 to 8.5.26
…-3.1.5 Bump fast-uri from 3.1.0 to 3.1.5
…d/postcss-8.5.25 Bump postcss from 8.5.8 to 8.5.25 in /dashboard
…pansion-5.0.7 Bump brace-expansion from 5.0.4 to 5.0.7
Cut the library down to what it is: an embeddable TypeScript policy
decision point with a verifiable audit chain. Each removed module
existed so a README bullet could be true, but did not do what the
bullet said. Tests: 504 -> 267 (23 files); tsc, lint, build clean.
Removed, with the defect that made each unsafe to ship:
- src/sandbox/: vm sandbox escaped to `process` in one line; WASM modules could not read the context
- src/attestation/: signature covered [{}] instead of the claims; verifier accepted any key; TPM/SGX ops all threw
- src/compliance/tlaplus-generator.ts: generated spec was never passed to TLC
- src/compliance/reporter.ts: certified a hash chain it never verified; controls satisfied on schema-required fields alone
- src/crypto/timestamp.ts: never parsed or verified the TSA response; `timestamp` was the local clock
- src/contracts/p2t-generator.ts, markdown-parser.ts, changelog.ts: not wired into GodClause, the CLI or the server
- src/engine/human-override.ts + POST /v1/override: accepted any self-supplied public key
- src/engine/boot.ts + CLI boot + GET /v1/boot/preflight: swallowed parse errors and reported READY
- src/engine/degradation.ts + GET /v1/degradation + `degradation_tiers` schema section: the `locked` tier blocked nothing
- src/engine/cache.ts (DecisionCache): cache key ignored nested fields, so one user's decision could be served to another
- src/engine/compiled-evaluator.ts (CompiledPolicyEvaluator): passed every rule containing a rate_limit
- CLI subcommands generate, boot, tlaplus, attest; endpoints /v1/attestation/{challenge,verify,status}
- dashboard/ and its CI job; the dependabot PR surface goes with it
- docs/compliance.md; types used only by the modules above
- 12 test files covering the modules above
README and docs corrected:
- The library computes no detection signals; the host sets output.contains_pii etc. and applies obligations (examples/openai-wrapper.ts is the reference)
- "No other policy engine does this" replaced with the real differentiator: one 4-dependency package with engine + hash-chained/HMAC/Merkle audit + DSSE-signed policies
- Dropped "holds up in regulatory proceedings", "hand it to your auditor", "6 compliance frameworks", all RFC 3161 claims
- "Tampering retroactively invalidates the audit trail" softened to: each entry records the policy SHA-256
- Comparison table: OPA returns arbitrary JSON and signs bundles; Permit.io PDP and OPAL are open source; Guardrails AI has fix/filter/reask and anonymizing fixes
- Static "504 passing" badge replaced with the GitHub Actions CI badge
- CLAUDE.md, .cursorrules, copilot-instructions, docs/architecture.md, docs/evidence-grade-audit.md, CHANGELOG (Unreleased) updated; endpoint count 17, test count 267
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The hash and HMAC used a JSON.stringify replacer array, which also
filters nested objects by the top-level key names, so caller and
rule_results were effectively unhashed. Entries now hash canonical
JSON (hash_version 3); logs written by earlier versions no longer verify.
- verifyAuditEntries(entries, key?, { prevHash?, headHash? }): recompute
hashes, require a signature when a key is given, anchor both ends
- AuditLog.verifyChain requires the first entry to link to the chain
start and the last entry to be the live head; headHash getter added
- AuditLogOptions.headHash / GovernanceOptions.auditHeadHash resume a
chain across restarts; FileAuditSink.lastHash reads it, skipping a
torn last line; the REST server resumes from AUDIT_DIR/audit.jsonl
- FileAuditSink terminates a torn last line before appending, so the
first entry after a crash is not fused onto the fragment
- Entries are normalized to plain JSON before hashing; sinks receive
them in chain order, each its own copy, and a failing sink does not
stop delivery to the others
- Empty auditSecretKey is rejected; empty verification key no longer
skips HMAC checks
- GodClause and TenantScope always keep the in-memory sink, so verify,
query and seal work with custom auditSinks
- ProofBundleBuilder.verify recomputes hashes/HMACs and rejects empty
bundles; ImmutableAuditSink shares the hash function
- CLI audit verify: --prev-hash and --head; unreadable lines are
reported as INVALID with a line number
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…ct README overclaims Resolved src/index.ts (keep AuditChainAnchors, drop the deleted modules' types) and CHANGELOG.md (keep both Unreleased sections). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
dashboard/package.json and package-lock.json stay deleted; the dashboard was removed in 9f09f7d. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
… values at load - PolicyEngine.evaluate denies with a NO-CONTRACT block when nothing is loaded; GodClause and TenantScope audit that deny; the REST server answers 503 on /v1/evaluate, /v1/enforce, /v1/evaluate/batch and /v1/mcp/authorize until a contract is loaded - A rule whose evaluation throws blocks instead of rejecting evaluate(); the reason is in RuleEvaluation.error and the audit entry's rule_results. MCPRouter denies on a permission whose condition throws instead of skipping to the next rule - verifyBundle validates every contract after the signature verifies and rejects envelopes whose payloadType is not a bundle - parseContract (and loadContract, which now validates objects) rejects invalid or non-string matches patterns, wrong-typed greater_than / less_than / in / not_in values, malformed rate_limit values (max not a positive integer, window not a finite positive duration) and rate_limit inside mcp_permissions, naming the condition's path - A StateStore that returns a non-number count is treated as an error Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
- PolicyEngine.loadContract replaces a contract with the same name and
version, so reloading never duplicates rules or rate-limit counting
- ContractWatcher remembers which contract each file provided and
deactivates it when the file changes version or disappears; when two
files provide the same contract and one goes away, the other is
re-read; a vanished directory retires everything; start() is
idempotent; reloadFile is public
- BundleWatcher applies each new bundle with the new
GodClause/TenantScope.replaceContracts (validated before anything is
dropped); attaching after the first poll still applies the bundle
- registry.register({ activate: false }) no longer deactivates a live
version; loadContractVersion resyncs the engine from the registry
- Rate-limit counters are keyed per rule, so two rules on one field no
longer share a bucket
- evaluate() uses one snapshot of the contract set and passes it to
the hooks, so a reload landing mid-evaluation cannot attribute the
decision or its audit rows to the new policy; PolicyHook gains that
second argument
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…art and image config
- API_TOKEN / --api-token: every endpoint except /v1/health and
/v1/ready requires "Authorization: Bearer <token>" (scheme
case-insensitive, constant-time, latin1-safe compare) or gets 401;
unset is allowed but logged at start
- MAX_BODY_BYTES (default 1 MiB, validated) caps request bodies with
413 and stops buffering; 10 s header and 30 s request timeouts
- 400 instead of 500 for invalid JSON, a context without action/caller,
bad options.includeTags/excludeTags, batch items and MCP calls missing
fields, malformed URL encoding, non-date audit filters and non-bundle
proof-bundle bodies; 500 responses no longer carry internal text
- /v1/mcp/authorize shares the server's state store so
max_calls_per_session is enforced; session_id must be a string
- ProofBundleBuilder.verify checks each seal range in linear time (a
160 KB bundle stalled the event loop for 80 s) and fails a seal whose
entries are not in the bundle
- Blank AUDIT_HMAC_SECRET / API_TOKEN env vars mean unset (compose
passes ${VAR:-}); missing AUDIT_DIR is warned about
- Helm chart: contracts mount at /app/contracts, CONTRACTS_DIR set,
AUDIT_HMAC_SECRET (was HMAC_SECRET) and optional API_TOKEN from the
secret, inline values quoted, render fails without a contracts
ConfigMap or with several replicas on one audit volume
- Dockerfile CMD is env-driven so AUDIT_DIR/CONTRACTS_DIR apply; CLI
serve flags no longer shadow PORT, CONTRACTS_DIR, LOG_LEVEL
- docs/deployment.md matches the code: import path, env table, --set
keys, Docker/compose sections
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…ional signals, enforced mcp_permissions - PolicyEngine defaults to a MemoryStateStore, so rate_limit rules work without configuration; pass a shared store for several processes - evaluate() applies the contracts' mcp_permissions when metadata.mcp_tool names a tool: a denied tool blocks under MCP-PERMISSIONS, require_human_approval becomes a require_review obligation, and the decision is audited like any other - `matches` accepts /pattern/flags (i, m, s, u); unknown flags are rejected at load. not_contains fails on a missing or non-container field so category rules fail closed - Threat contracts: optional signals (nsfw_score, estimated_cost_usd, per-type PII flags, the base template's token_count and confidence) are guarded with not_exists; required ones still fail closed; every header says which is which. prompt-injection uses case-insensitive, whitespace-tolerant regexes (disregard/forget/"the above", verb-anchored extraction) and blocks a non-string prompt (INJ-000); cost-control gained the expensive-model warning it promised; mcp-tool-abuse rules are scoped to action mcp_call so composing it no longer denies every request, and its audit rule marks each tool call; data-leakage, copyright-protection and base-compliance document every field their rules read - tests/threat-contracts.test.ts loads all eleven contracts and checks the documented behaviour; tests/patterns.test.ts covers the flags Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…sions loaded - POST /v1/mcp/authorize builds an mcp_call context and calls gov.evaluate(), so the contracts' mcp_permissions apply and every authorization is in the audit chain; the response carries allowed, require_human_approval, denial_reason, decision_id, obligations and governance_context - evaluate() denies a named tool when no contract defines mcp_permissions, matching the router's fail-closed default - examples/mcp-agent-demo.ts authorizes through evaluate(), so its "every call audited" claim is true Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Version 0.1.0 in package.json, the Helm chart, CHANGELOG, README and SECURITY; the CLI reads its version from package.json; CONTRIBUTING gains a Releasing section; schemas/trust-contract.schema.json is regenerated from the TypeScript schema. Fixes from the release review: - build removes dist/ first; the tarball packed from the working tree shipped compiled copies of the removed modules (VM sandbox, cache, human override, ...) - policy_sha256 hashes the canonical JSON of every loaded contract, not the rules alone, so a change to mcp_permissions changes the fingerprint; contract_id names every loaded contract - /v1/metrics quotes label values and counts decisions from enforce, batch and mcp/authorize; the duration summary keeps 10,000 samples - proof bundles and the OTel sink stamp the package version, not "2.0.0" - HumanOverride, TimestampRef and ProofBundle.human_overrides/timestamps removed; nothing verified them - examples/*.ts no longer ship in the package (they import ../src); the wrappers' simulated response has no SSN, so the default run reaches the redact_pii path - docs: DSSE keys are RSA or ECDSA P-256, Merkle verification recomputes the root, only the in-memory state store ships, sidecar contracts load at startup, data_governance is declared not enforced, threat contracts are rules over host-supplied signals Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Bumps [@vitest/mocker](https://github.com/vitest-dev/vitest/tree/HEAD/packages/mocker) to 5.1.0 and updates ancestor dependency [vitest](https://github.com/vitest-dev/vitest/tree/HEAD/packages/vitest). These dependencies need to be updated together. Updates `@vitest/mocker` from 3.2.4 to 5.1.0 - [Release notes](https://github.com/vitest-dev/vitest/releases) - [Changelog](https://github.com/vitest-dev/vitest/blob/main/docs/releases.md) - [Commits](https://github.com/vitest-dev/vitest/commits/v5.1.0/packages/mocker) Updates `vitest` from 3.2.4 to 5.1.0 - [Release notes](https://github.com/vitest-dev/vitest/releases) - [Changelog](https://github.com/vitest-dev/vitest/blob/main/docs/releases.md) - [Commits](https://github.com/vitest-dev/vitest/commits/v5.1.0/packages/vitest) --- updated-dependencies: - dependency-name: "@vitest/mocker" dependency-version: 5.0.5 dependency-type: indirect - dependency-name: vitest dependency-version: 5.0.5 dependency-type: direct:development ... Signed-off-by: dependabot[bot] <support@github.com>
dependabot
Bot
force-pushed
the
dependabot/npm_and_yarn/multi-063a5f277b
branch
from
November 3, 2026 14:37
04676c8 to
fbd862a
Compare
Contributor
Author
|
OK, I won't notify you again about this release, but will get in touch when a new version is available. If you change your mind, just re-open this PR and I'll resolve any conflicts on it. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Bumps @vitest/mocker to 5.1.0 and updates ancestor dependency vitest. These dependencies need to be updated together.
Updates
@vitest/mockerfrom 3.2.4 to 5.1.0Release notes
Sourced from @vitest/mocker's releases.
... (truncated)
Commits
7a38449chore: release v5.1.0 (#11838)810ec66fix(mocker): use root mocks for linked packages (fix #7708) (#11739)cca423dfix(mocker): don't change the original prototype chain (#11737)e1b0d33chore: release v5.1.0-beta.1 (#11647)707f8adfeat(browser): mock modules through an import map (#11534)7a825dbchore: release v5.0.5 (#11609)6a11cf6chore: release v5.0.4 (#11554)58c86edchore(deps): update all non-major dependencies (#11550)5c95521fix(browser): clean up unmocked and pending mocks (#11537)cd423bafix(mocker): fix native loader mocking edge cases (#11540)Maintainer changes
This version was pushed to npm by GitHub Actions, a new releaser for
@vitest/mockersince your current version.Updates
vitestfrom 3.2.4 to 5.1.0Release notes
Sourced from vitest's releases.
... (truncated)
Commits
7a38449chore: release v5.1.0 (#11838)a5ec316fix(vm): keep assignable accessor globals writable in node environment (#11834)e46c6c1refactor: share trace types and utils (#11825)9e23ce5refactor(snapshot): unify review update withSnapshotStateupdate (#11790)a0e33ccchore: eliminate lint error (#11822)fd145d8feat(experimental): add message option to vi.waitFor and vi.waitUntil (#11754)4e98551feat(experimental): add trace timeline for node (#11753)bf96dadfix: preserve rootDir in project paths (#11816)54e5b8afeat: interactively review failed snapshots (#11325)4277075fix(expect): keep soft .resolves/.rejects error location (#11792)Maintainer changes
This version was pushed to npm by GitHub Actions, a new releaser for vitest since your current version.