CI's test ratchet flagged agentic-flow-agent.test.ts as an unexpected new
failure (run 32091740244, same commit fa13ee4a that passed the day before).
result.duration is Date.now()-delta over a ~1ms setTimeout; under coarse CI
clock resolution the two samples can land in the same millisecond, making
duration legitimately 0. Relax the two affected assertions from
toBeGreaterThan(0) to toBeGreaterThanOrEqual(0).
The pnpm-without-postinstall layout in CI doesn't fetch prebuilt
binaries for sharp, onnxruntime-node, @ruvector/{sona,ruvllm,
ruvllm-wasm,rvagent-wasm}, etc. Tests that exercise these modules
fail with 'Cannot read properties of undefined' or ERR_DLOPEN_FAILED
in CI but pass locally where the binaries are available.
Per the 'get to 100% or remove tests that will never pass' policy,
each affected suite now resolves the dep via createRequire and
skips the whole describe block when it isn't present.
Files:
- ruvllm-wasm.test.ts (27 fails) — skipIf @ruvector/ruvllm-wasm absent
- agent-wasm.test.ts (22 fails) — skipIf @ruvector/rvagent-wasm absent
- memory-ruvector-deep.test.ts (14 fails) — skipIf @ruvector/sona absent
(only the SONA Optimizer block)
- ruvllm-integration.test.ts (3 fails) — skipIf @ruvector/ruvllm absent
- sdk-client + seed-client-raw (24 fails) — skipIf SEED_PAIRING_TOKEN +
SEED_ENDPOINT not set
- token-optimizer.test.ts (1 fail) — match assertion to actual
agentic-flow loadability
Also, vitest.config.ts gains a poolMatchGlobs entry routing
router-bandit.test.ts to the 'forks' pool (worker threads forbid
process.chdir, which the test legitimately needs for tmpdir
isolation; 6 tests fixed).
Total: ~91 of 152 failing tests addressed; remainder (timeouts in
guidance-provider/reasoningbank, tool-honesty assertion drift,
p1-commands, *integration tests, integration-docker) need separate
investigation for real-vs-environment classification.
Co-Authored-By: RuFlo <ruv@ruv.net>
Remove Math.random() metrics from 6 locations:
- agent_health: CPU/memory/latency were random — now null with notes
- system_health: all checks were hardcoded "healthy" — now real fs checks
- coordination_metrics: latency/throughput were random — now null with real sync data
- github-tools: all 5 tools fabricated — now return _stub:true, local CRUD only
- process monitor: all metrics random — now real os.loadavg/memoryUsage
- hook-handler route: random latency, hardcoded matches — removed
Fix 6 bugs:
- task-tools path mismatch: agents.json → agents/store.json
- token-optimizer double-increment of editsOptimized
- config get: now reads real configManager instead of hardcoded map
- config providers: now reads/writes real config, mutation flags work
- session restore: now syncs to sql.js database after legacy JSON write
- claims rebalance: now executes moves when dryRun=false
Mark stubs honestly (_stub:true):
- neural compress/optimize
- performance bottleneck/profile/optimize
- hooks metrics/pretrain/intelligence-reset (reset now actually deletes files)
- workflow execute/resume (steps stay pending, not auto-completed)
- hooks session-end summary: reads real counts from stores
- hooks explain: reads real success rate from routing-outcomes.json
- hooks transfer: returns failure instead of demo data when source empty
- status.ts: performance claims → 'not measured'
Co-Authored-By: claude-flow <ruv@ruv.net>
- #1514: Remove fabricated token savings (hardcoded +=100 per cache hit,
baseline=1000). getOptimalConfig() now scales with agentCount.
neural_train stores real embedding patterns via generateEmbedding().
neural_predict uses cosine similarity against stored patterns (no
hardcoded labels). Returns empty array when no patterns exist.
verifySignature() confirmed already real (HMAC + timingSafeEqual).
- #1516: All embedding model defaults prefixed with Xenova/ so
@xenova/transformers resolves correctly. ControllerRegistry now
passes embeddingModel to AgentDB constructor.
- #1521: @claude-flow/memory bumped to 3.0.0-alpha.13, better-sqlite3
moved from optionalDependencies to dependencies.
- #1526: hook-handler normalizes snake_case tool_input/tool_name from
Claude Code. intelligence.cjs scoped to current project only
(was scanning all 51+ project dirs).
- ADR-073: Documents stub tool honesty policy and real predictions.
Closes#1514, #1516, #1521, #1526
Co-authored-by: Reuven <cohen@ruv-mac-mini.local>
Automatic checkpoint created by Claude Code
- Branch: v3
- Timestamp: 2026-01-12T19:19:44Z
- Changes: 3 file(s)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Automatic checkpoint created by Claude Code
- Branch: v3
- Timestamp: 2026-01-12T19:16:57Z
- Changes: 1 file(s)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Automatic checkpoint created by Claude Code
- Branch: v3
- Timestamp: 2026-01-12T19:16:34Z
- Changes: 1 file(s)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Automatic checkpoint created by Claude Code
- Branch: v3
- Timestamp: 2026-01-12T19:16:21Z
- Changes: 1 file(s)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Automatic checkpoint created by Claude Code
- Branch: v3
- Timestamp: 2026-01-12T19:08:26Z
- Changes: 1 file(s)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Automatic checkpoint created by Claude Code
- Branch: v3
- Timestamp: 2026-01-12T19:06:04Z
- Changes: 1 file(s)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Updated standard-checkpoint-hooks.sh to accept both JSON and plain paths
- Fixed pre-edit and post-edit functions to detect input format
- Fixed v3 CLI hook flags (-f -> --file, -c -> --command, etc.)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
rUv