Testing Infrastructure + Documentation

Architecture

flowchart TD
    subgraph TEST["GoClaw Testing Infrastructure"]
        RUN["<b>go test ./... (tests beside the code)</b><br/>Colocated *_test.go · Table-driven tests<br/>go test -race · go test -cover"]

        UNIT["<b>Unit Tests</b><br/>Per package · Pure Go"]
        SEC["<b>Security</b><br/>Scanner · SAST tests"]
        INT["<b>Integration</b><br/>Gateway · API tests"]

        RUN --> UNIT
        RUN --> SEC
        RUN --> INT
    end

Testing (Section 23)

#	Feature	OpenClaw	GoClaw	Status
23.1	Unit tests	Vitest `*.test.ts`	`go test` colocated `*_test.go`	Done
23.2	Test runner	Vitest (V8)	`go test` (stdlib)	Done
23.3	Table-driven tests	—	✅ Go idiomatic pattern	GoClaw only
23.4	Race detection	—	`go test -race`	GoClaw only
23.5	Coverage reporting	`pnpm test:coverage`	`go test -cover`	Done
23.6	Coverage thresholds	70% lines/functions/statements	—	TODO
23.7	E2E tests	`*.e2e.test.ts`	—	TODO
23.8	Live tests (real API)	`LIVE=1 pnpm test:live`	Embedding search and RAG embedding tests call a real provider when `GEMINI_API_KEY` or `OPENAI_API_KEY` is set, and skip otherwise	Partial
23.9	Docker test runners	`test:docker:live-models` etc.	—	TODO
23.10	Install smoke tests	`test:install:smoke`	—	TODO
23.11	CI pipeline	GitHub Actions	`.github/workflows/ci.yml`: gofmt, `go vet`, `go build`, `go test`, `go test -race`, on `v..*` tags and manual dispatch, not on pull requests or pushes. `release.yml` also runs `govulncheck`	Partial
23.12	Pre-commit hooks	`prek install`	—	TODO

Other Test Environments

Postgres: GOCLAW_TEST_POSTGRES_DSN runs the database tests against Postgres instead of per-test SQLite files; each test gets its own schema (see AGENTS.md, Build & Development).
MySQL, MariaDB, ClickHouse: the db_write tests run against these when GOCLAW_TEST_MYSQL_DSN, GOCLAW_TEST_MARIADB_DSN or GOCLAW_TEST_CLICKHOUSE_DSN is set, and skip otherwise.
Repository checks: internal/repoguard runs with go test ./.... TestNoPilotOrCustomerData fails on customer data in tracked files, and TestPublishedDocsAreReachable fails on a published page that the docs site's sidebar and links do not reach.

Test Files by Package

Most packages under internal/ have tests, colocated as *_test.go. The main areas:

Package	Test Scope
`internal/agent`	Tools (shell, git, cron, calc, datetime, memory, document, text, read_lines, search_files, gitlab, linter, AST, core_instruction, db_write), RAG engine, fallback, cooldown, cache, skills, session memory, memory flush, subagent, registry, safe client, user errors, cron migration
`internal/config`	Config loading, agents CRUD, config diff, AST update preservation, non-interactive wizard
`internal/secrets`	Credentials store, secret vault store, `${env:}`/`${secret:}`/`${file:}` resolution
`internal/gateway`	Server routes + auth, media upload/serve, pending work queue
`internal/channel/msteams`	MS Teams webhook handling, Graph API client
`internal/memory`	Search (keyword, vector, hybrid, RRF), indexer
`internal/channel/discord`	Discord markdown formatting
`internal/cli`	CLI commands (config get/set/append/remove)
`internal/media`	Media store (save, retrieve, metadata)
`internal/pairing`	Pairing store (approve, revoke, list, expiry)
`internal/routing`	Multi-agent routing (bindings, platform roles, peer matching)
`internal/session`	Session lifecycle (create, get, reset, cleanup)
`internal/repoguard`	Customer data in tracked files, docs reachability

Test Highlights

Area	Notable Tests
RAG engine	Keyword extraction, score calculation, core tool/skill injection
Provider fallback	Error classification, cooldown backoff, probe throttle, moving to the next provider (`fallbackAgent`)
Security	Shell allow/deny, path sandboxing, secret resolution, file edit safety
Config	AST-preserving YAML update (comments + blank lines survive), diff detection
Cron	Job add/remove/toggle, persistence, migration from legacy format
AST tools	Go AST declaration search, context extraction, diff analysis
GitLab	API operations (issues, MRs, pipelines, files)
Memory	Hybrid search (keyword × vector × RRF), temporal decay, consolidation

Documentation (Section 13)

GoClaw has a docs/ directory, published at goclaw.ulab.mn.

#	Document	Status	Notes
13.1	`docs/index.md`	Done	Navigation hub, architecture diagram, state directory layout
13.2	`docs/api-reference/gateway_endpoints.md`	Done	WebSocket protocol, REST endpoints, Web UI
13.3	`docs/user-guide/configuration.md`	Done	Full YAML config reference, env var overrides, hot reload
13.4	`docs/user-guide/channels/index.md`	Done	Channel overview, session keys, access control, interface spec
13.5	`docs/user-guide/channels/discord.md`	Done	Setup, intents, slash commands, guild behavior, pairing
13.6	`docs/user-guide/channels/msteams.md`	Done	Azure Bot setup, webhook, OAuth, Graph API, team allowlist
13.7	`docs/user-guide/channels/pairing.md`	Done	Pairing flow, DM policies, approve/revoke via CLI/API/Web UI
13.8	`docs/architecture/concepts/agent.md`	Done	Agent loop, providers, system prompt, workspace, canvas
13.9	`docs/architecture/concepts/sessions.md`	Done	SQLite (GORM) storage, session keys, history, lifecycle, cleanup
13.10	`docs/architecture/concepts/model-failover.md`	Done	Fallback order, cooldown, probe throttle, error classification
13.11	`docs/user-guide/tools_and_mcp.md`	Done	Tools and MCP servers, with config examples
13.12	`docs/user-guide/tools_and_mcp.md` (Skills)	Done	Built-in skills, custom skill creation, allow/deny config
13.13	`docs/user-guide/automation/cron-jobs.md`	Done	6-field cron syntax, persistence, auto-migration, error tracking

OpenClaw vs GoClaw Testing Comparison

Aspect	OpenClaw	GoClaw
Framework	Vitest (V8)	`go test` (stdlib)
Test files	Hundreds	Colocated `*_test.go`
Packages	~30	Most packages under `internal/`
Coverage	70% threshold	`go test -cover` (no threshold)
Race detection	—	✅ `go test -race`
Table-driven	—	✅ Go idiomatic
CI	GitHub Actions	GitHub Actions on release tags and manual runs (not on pull requests)
E2E	✅	— (TODO)
Docs	Mintlify hosted	`docs/`, published at goclaw.ulab.mn

Coverage: unit, race and coverage runs, live provider tests and a release-time CI pipeline are in place; coverage thresholds, E2E tests, Docker test runners, install smoke tests and pre-commit hooks are not.