The open-source agent harness - the runtime layer that turns an LLM into a working agent.
import { AgentSpecSchema, DEFAULT_AGENT_CONFIG_ITERATION_LIMIT } from '../../src/agent-session/schemas/agentSpec';
describe('AgentSpec RuntimeConfig defaults', () => {
it('materializes config when omitted', () => {
const spec = AgentSpecSchema.parse({ model: { name: 'provider/model' } });
expect(spec.config.iteration_limit).toBe(DEFAULT_AGENT_CONFIG_ITERATION_LIMIT);
expect(spec.config.sandbox).toEqual({ enabled: false, file_downloads: true });
expect(spec.config.generative_ui).toEqual({ enabled: true });
expect(spec.config.ask_user_questions).toEqual({ enabled: true });
expect(spec.config.dynamic_sub_agents).toEqual({ enabled: true });
expect(spec.config.context_management.compaction).toEqual({ enabled: true });
expect(spec.config.context_management.large_tool_response.enabled).toBe(true);
expect(spec.config.web_search).toEqual({ enabled: false });
});
it('accepts an explicit input-token compaction trigger', () => {
const spec = AgentSpecSchema.parse({
model: { name: 'provider/model' },
config: {
context_management: {
compaction: {
enabled: true,
trigger: { type: 'input_tokens', value: 80_000 },
},
},
},
});
expect(spec.config.context_management.compaction).toEqual({
enabled: true,
trigger: { type: 'input_tokens', value: 80_000 },
});
});
it('fills generative_ui when other config fields are present', () => {
const spec = AgentSpecSchema.parse({
model: { name: 'provider/model' },
config: { sandbox: { enabled: true } },
});
expect(spec.config.sandbox.enabled).toBe(true);
expect(spec.config.sandbox.file_downloads).toBe(true);
expect(spec.config.generative_ui.enabled).toBe(true);
});
it('allows opting out of generative_ui', () => {
const spec = AgentSpecSchema.parse({
model: { name: 'provider/model' },
config: { generative_ui: { enabled: false } },
});
expect(spec.config.generative_ui.enabled).toBe(false);
});
});