neutral

Phase 16 — Test Coverage Expansion to 80%+

Systematically expands test coverage across all 22 internal packages to ≥80%, using mock clients (sqlmock, httptest, Temporal testsuite) so no live services are required. All new test code passes SonarQube quality gates — no code smells, no cognitive complexity issues, meaningful assertions.

Status: Completed (2026-02-09) Depends on: Phases 1-14 complete Migrations: None (tests only — no schema changes) Branch: dev

Sub-Phases

Sub-Phase	Name	Packages	Prompts	Depends On
16A	Foundation & Small Packages	blackboard, observability, temporal, mcp, config, health, agenttest	3	—
16B	Quota, Registry & LLM	quota, registry, llm	3	16A
16C	Memory & Embedding	memory, embedding, vectorstore	3	16A
16D	Audit, Security & Tenant	audit, security, tenant	3	16A
16E	Agent Part 1: Activities & Helpers	agent (activities, helpers, stores)	4	16B
16F	Agent Part 2: Workflow & Signals	agent (workflow, queries, signals)	2	16E
16G	Supervisor & Graph	supervisor, graph	3	16F
16H	Tools	tools	3	16B, 16C
16I	API Coverage Expansion	api, api/middleware, api/routes	3	16B–16H
16J	Coverage Verification & CI	all	2	16A–16I

Total: 10 sub-phases, 29 prompts, 21 documentation files

Dependency Graph

16A (Foundation) ──┬──→ 16B (Quota/Registry/LLM) ──→ 16E (Agent Pt1) ──→ 16F (Agent Pt2) ──→ 16G (Super/Graph)
                   ├──→ 16C (Memory/Embedding) ────→ 16H (Tools) ←── 16B
                   ├──→ 16D (Audit/Security/Tenant)
                   │
                   │    16B-16H ────────────────────→ 16I (API)
                   └─────────────────────────────────→ 16J (Verification)

Parallelizable: 16B, 16C, 16D can all run simultaneously after 16A completes. 16H runs after both 16B and 16C.

Coverage Gap Analysis

Package	Current	Target	Gap	Priority Files
blackboard	0%	80%	80%	postgres_store.go, store.go
observability	0%	80%	80%	logger.go
temporal	0%	80%	80%	client.go
supervisor	17.6%	80%	62%	workflow.go, activities.go, capability.go, delegation_chain.go
registry	21.1%	80%	59%	store.go
api/routes	29.2%	80%	51%	18 route files (edge cases, error paths)
health	30.1%	80%	50%	checker.go edge cases
quota	31.1%	80%	49%	policy.go, limiter.go, middleware.go
llm	33.7%	80%	46%	client.go, openrouter.go, azure.go, model_store.go
memory	34.7%	80%	45%	policy.go, postgres_store.go, redis_store.go, embedding.go
audit	36.1%	80%	44%	logger.go, export_hipaa.go, export_soc2.go
agent	40.3%	80%	40%	workflow.go, activities.go, stores, helpers
mcp	43.1%	80%	37%	client.go
vectorstore	43.6%	80%	36%	composite.go, metrics.go
graph	44.3%	80%	36%	conditions.go, dsl.go, workflow.go
api/middleware	49.8%	80%	30%	auth.go, ratelimit.go edge cases
config	50.3%	80%	30%	edge cases in env parsing
tools	56.8%	80%	23%	8 individual tool files
agenttest	58.2%	80%	22%	playground.go, chaos edge cases
security	59.5%	80%	21%	sanitizer.go, sandbox edge cases
tenant	60.4%	80%	20%	middleware.go, config.go
embedding	73.8%	80%	6%	ollama.go, google.go

Testing Strategy

Mock Infrastructure

No live services required. All tests use mock clients:

Dependency	Mock Strategy
PostgreSQL	`DATA-DOG/go-sqlmock` for store-level tests
Temporal	`go.temporal.io/sdk/testsuite` for workflow/activity tests
HTTP APIs (LLM, MCP, Qdrant)	`net/http/httptest` for provider client tests
Redis / Dragonfly	`alicebob/miniredis` or interface mocks
Huma API routes	`danielgtaylor/huma/v2/humatest` for route-level tests
Vector stores	Interface mocks (function-field pattern from `api/testutil/mocks.go`)

Existing Test Infrastructure

Pattern	Location	Used By
Temporal test harness	`internal/agenttest/suite.go`	Agent workflow tests (DeterministicJSON, StepDecider, MockTool)
Mock store pattern	`internal/api/testutil/mocks.go`	API route tests (MockTemporalClient, MockRegistryStore, etc.)
In-memory store	`internal/memory/inmemory_store.go`	Memory tests (already used as test double)
Test helper utilities	`internal/api/routes/test_helpers_test.go`	API route test setup

Test Conventions

All new test code follows these conventions:

Naming: Test<Function>_<Scenario> (e.g., TestNewClient_TLSEnabled)
Table-driven: Use subtests with t.Run() to reduce duplication
Assertions: github.com/stretchr/testify — use meaningful messages (not bare assert.True)
Setup: Extract shared setup into helper functions; no duplicate setup blocks
Cognitive complexity: Keep test functions ≤ 15 per SonarQube threshold
Build tags: Integration tests use //go:build integration tag
No secrets: No hardcoded credentials — use placeholder values

File Structure

No new production files — only test files and CI configuration:

internal/
    blackboard/
        postgres_store_test.go      # 16A
    observability/
        logger_test.go              # 16A
    temporal/
        client_test.go              # 16A
    mcp/
        client_test.go              # 16A
    config/
        config_test.go              # 16A (extend existing)
    health/
        checker_test.go             # 16A (extend existing)
    agenttest/
        suite_test.go               # 16A (extend existing)
        playground_test.go          # 16A (extend existing)
    quota/
        policy_test.go              # 16B
        limiter_test.go             # 16B
        middleware_test.go          # 16B
    registry/
        store_test.go               # 16B
    llm/
        client_test.go              # 16B
        openrouter_test.go          # 16B
        azure_test.go               # 16B
        model_store_test.go         # 16B
    memory/
        postgres_store_test.go      # 16C
        redis_store_test.go         # 16C
        policy_test.go              # 16C
        embedding_test.go           # 16C
    embedding/
        ollama_test.go              # 16C
        google_test.go              # 16C
    vectorstore/
        composite_test.go           # 16C (extend existing)
        metrics_test.go             # 16C
    audit/
        logger_test.go              # 16D
        export_hipaa_test.go        # 16D
        export_soc2_test.go         # 16D
    security/
        sanitizer_test.go           # 16D
        sandbox_test.go             # 16D (extend existing)
    tenant/
        middleware_test.go          # 16D
        config_test.go              # 16D
    agent/
        activities_test.go          # 16E
        cost_test.go                # 16E
        state_test.go               # 16E
        hash_test.go                # 16E
        trace_test.go               # 16E
        retry_store_test.go         # 16E
        version_store_test.go       # 16E
        trace_store_test.go         # 16E
        logging_test.go             # 16E
        speculative_test.go         # 16E
    agent/
        workflow_test.go            # 16F
        queries_test.go             # 16F
        signals_test.go             # 16F
    supervisor/
        workflow_test.go            # 16G
        activities_test.go          # 16G
        capability_test.go          # 16G
        delegation_chain_test.go    # 16G
    graph/
        conditions_test.go          # 16G
        dsl_test.go                 # 16G
        workflow_test.go            # 16G (extend existing)
    tools/
        http_get_test.go            # 16H
        kv_store_test.go            # 16H
        cost_summary_test.go        # 16H
        model_list_test.go          # 16H
        model_prefs_test.go         # 16H
        memory_read_test.go         # 16H
        memory_write_test.go        # 16H
        simulated_test.go           # 16H
    api/
        middleware/
            middleware_test.go      # 16I (extend existing)
        routes/
            *_test.go               # 16I (extend existing — 18 files)
.github/workflows/
    coverage.yml                    # 16J (CI enforcement)

SonarQube Quality Requirements

All new test code must satisfy:

Rule	Requirement
Cognitive complexity	≤ 15 per function
Assertion quality	Meaningful messages on all assertions
Duplication	Extract shared setup into helpers; no copy-paste blocks
Credentials	No hardcoded secrets — use placeholder values
Error handling	No ignored errors; check all `err` returns
Test naming	`Test<Function>_<Scenario>` convention
Table-driven	Use `t.Run()` subtests where ≥ 3 similar cases exist

Prompt Files

Each sub-phase has a companion -PROMPT.md file containing implementation prompts designed for LLM-assisted coding. Prompts are ordered by dependency within each sub-phase. Load the listed context files before executing each prompt.

PHASE16A-PROMPT.md — 3 prompts (blackboard/observability/temporal → mcp/config/health → agenttest)
PHASE16B-PROMPT.md — 3 prompts (quota → registry/llm providers → llm model store/failover)
PHASE16C-PROMPT.md — 3 prompts (memory stores → embedding providers → vectorstore)
PHASE16D-PROMPT.md — 3 prompts (audit → security → tenant)
PHASE16E-PROMPT.md — 4 prompts (activities → helpers/state → stores → cost/logging)
PHASE16F-PROMPT.md — 2 prompts (workflow → queries/signals)
PHASE16G-PROMPT.md — 3 prompts (supervisor workflow/activities → supervisor capability/delegation → graph)
PHASE16H-PROMPT.md — 3 prompts (execution tools → data tools → bridge/simulated tools)
PHASE16I-PROMPT.md — 3 prompts (middleware edge cases → route error paths → integration test gaps)
PHASE16J-PROMPT.md — 2 prompts (coverage verification script → CI pipeline enforcement)

Verification

After all sub-phases are complete:

go test -coverprofile=coverage.out ./internal/... — all packages ≥ 80%
go test -count=1 ./... — all tests pass, no regressions
go tool cover -func=coverage.out | grep total — verify aggregate ≥ 80%
No tests require live external services (Postgres, Temporal, Redis, LLM APIs)
SonarQube scan passes with no new issues
CI pipeline enforces per-package coverage thresholds
go test -race ./internal/... — no data races detected

Sub-Phases​

Dependency Graph​

Coverage Gap Analysis​

Testing Strategy​

Mock Infrastructure​

Existing Test Infrastructure​

Test Conventions​

File Structure​

SonarQube Quality Requirements​

Prompt Files​

Verification​