Identify Test Gaps and Add Coverage

Start by reviewing the listed untested packages, such as agentRpc, agentServer/api, chat-ui, and schema tooling, then build the project because tests run against dist/test/. Use pnpm run test:local and the relevant package entry points to establish current coverage; done means adding focused tests for an agreed gap and having the local suite pass.

Since we're leaning heavily on AI and moving pretty fast, we should make sure we take time to invest in more robust test coverage. A quick scan of our test infra gives this overview :

Test Infrastructure Overview

Framework: Jest + TypeScript
Convention: .spec.ts = unit tests (offline),
.test.ts = integration tests (require API keys)
Root scripts: pnpm run test:local, test:live,
shell:test
Note: Tests run against compiled output in
dist/test/ — you must build first.

Packages WITH Tests

┌───────────────┬───────────────────────────┐
│ Package │ What's Covered │
├───────────────┼───────────────────────────┤
│ │ NFA/DFA compiler, grammar │
│ actionGrammar │ merging, terminal │
│ │ symbols, schema parsing │
├───────────────┼───────────────────────────┤
│ │ TypeScript → JSON Schema │
│ actionSchema │ conversion, type │
│ │ validation │
├───────────────┼───────────────────────────┤
│ cache │ LLM call caching, │
│ │ hit/miss logic │
├───────────────┼───────────────────────────┤
│ │ Action routing, │
│ dispatcher │ input→agent matching, │
│ │ type validation │
├───────────────┼───────────────────────────┤
│ knowPro │ RAG, conversational │
│ │ memory, semantic search │
├───────────────┼───────────────────────────┤
│ │ Individual agent action │
│ agents/* │ handlers (player, │
│ │ calendar, email, list, │
│ │ browser, weather) │
├───────────────┼───────────────────────────┤
│ cli │ Argument parsing, command │
│ │ execution │
├───────────────┼───────────────────────────┤
│ shell │ Electron GUI (via │
│ │ shell:test) │
├───────────────┼───────────────────────────┤
│ testLib │ Shared test utilities and │
│ │ mock factories │
└───────────────┴───────────────────────────┘

Packages With NO Tests

These have zero test files:

• agentRpc — RPC communication protocol between
  dispatcher and agents
• agentServer / api — Server infrastructure
• chat-ui — Chat UI components
• codeProcessor / coderWrapper — Code
  generation
• schemaAuthor / actionSchemaCompiler — Schema
  authoring tools
• knowledgeVisualizer / kp / textPro —
  Knowledge/text processing
• uriHandler, interactiveApp, montage,
  cacheExplorer

Notable Gaps

1. IPC/RPC layer (agentRpc) — No tests for the
   communication protocol between dispatcher and
   agents, which is critical infrastructure
2. Server layer — agentServer and api have no
   coverage; server startup, routing, auth are
   untested
3. UI components — chat-ui has no automated
   tests
4. Schema tooling — Compiler/authoring tools
   are untested despite being part of the
   development workflow
5. Error/edge case paths — Integration tests
   require live API keys, so CI likely skips them
   entirely, leaving a gap between unit tests and
   real behavior
6. End-to-end flows — No apparent e2e tests
   that simulate a full user request through
   dispatcher → agent → response