ktx/packages/cli/test/setup-models.test.ts

1141 lines
41 KiB
TypeScript
Raw Permalink Normal View History

2026-05-10 23:12:26 +02:00
import { mkdir, mkdtemp, readFile, rm, stat, writeFile } from 'node:fs/promises';
import { tmpdir } from 'node:os';
import { join } from 'node:path';
test: split cli tests from source tree (#216) * feat(cli): define full warehouse dialect contract * test(cli): keep dialect edge tests focused * fix(cli): stabilize dialect contract foundation * refactor(connectors): own read-only query preparation * refactor(connectors): resolve dialects through registry * refactor(connectors): keep concrete dialect classes internal * chore(workspace): enforce dialect import boundary * refactor(cli): resolve relationship dialect at scan boundary * refactor(cli): use dialect display parsing for entity details * refactor(cli): use dialect display parsing for warehouse catalog * refactor(cli): use dialect SQL in relationship workflows * test(cli): verify solid dialect scan workflow closure * test: split cli tests from source tree * refactor(cli): standardize BigQuery scope listing * feat(sqlite): implement connector scope listing * test(connectors): cover required table listing * feat(cli): add warehouse driver registry * refactor(setup): route scope discovery through driver registry * refactor(cli): route local query execution through driver registry * refactor(historic-sql): route dialect support through driver registry * refactor(cli): test warehouse connections through driver registry * fix(cli): close driver registry type export gaps * Improve setup daemon diagnostics * refactor(setup): centralize rail-prefixed diagnostics + query-history fallback Extract errorMessage, writePrefixedLines, and flushPrefixedBufferedCommandOutput into clack.ts so the setup wizard, managed daemons, and embedding/agent steps share one rail-formatted writer. setup-databases.ts also adds a "disable query history and retry" option when the schema-context build fails and query history is the likely culprit, surfaced via a new failed-query-history-unavailable status. * fix(cli): carry catalog through the picker so BigQuery/Snowflake/SQL Server scope filters match The setup picker's KtxTableListEntry was a 2-level { schema, name }, so qualifiedTableId always wrote db.name into enabled_tables. When BigQuery, Snowflake, or SQL Server later ran fast ingest, their introspect step filtered the scope set with scopedTableNames(scope, { catalog: projectId|database, db }) — catalog was non-null on the introspect side but null in the scope refs, so every entry was rejected, the live-database adapter staged zero table files, and detect() failed with 'Adapter "live-database" did not recognize fetched source output'. Align the picker boundary with the canonical 3-level KtxTableRef: - Add catalog: string | null to KtxTableListEntry. - BigQuery/Snowflake/SQL Server listTables populate catalog from the resolved projectId / database; Postgres/MySQL/ClickHouse/SQLite set null. - qualifiedTableId emits catalog.schema.name when catalog is non-null (resolveEnabledTables already accepts the 3-part shape) and schemasFromEnabledTables now goes through parseDottedTableEntry so it recovers the schema correctly from both 2-part and 3-part entries. - Export parseDottedTableEntry from enabled-tables.ts (@internal) for picker reuse. Update listTables expectations in all seven connector tests and the setup / picker test fixtures. Add a picker regression test that covers the catalog-bearing round-trip (save + refine). * fix(cli): allow debug telemetry under opt-out env
2026-05-26 08:49:05 +02:00
import { initKtxProject } from '../src/context/project/project.js';
import { parseKtxProjectConfig } from '../src/context/project/config.js';
import { readKtxSetupState, writeKtxSetupState } from '../src/context/project/setup-config.js';
2026-05-10 23:12:26 +02:00
import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
import {
2026-05-10 23:51:24 +02:00
type KtxSetupModelPromptAdapter,
runKtxSetupAnthropicModelStep,
test: split cli tests from source tree (#216) * feat(cli): define full warehouse dialect contract * test(cli): keep dialect edge tests focused * fix(cli): stabilize dialect contract foundation * refactor(connectors): own read-only query preparation * refactor(connectors): resolve dialects through registry * refactor(connectors): keep concrete dialect classes internal * chore(workspace): enforce dialect import boundary * refactor(cli): resolve relationship dialect at scan boundary * refactor(cli): use dialect display parsing for entity details * refactor(cli): use dialect display parsing for warehouse catalog * refactor(cli): use dialect SQL in relationship workflows * test(cli): verify solid dialect scan workflow closure * test: split cli tests from source tree * refactor(cli): standardize BigQuery scope listing * feat(sqlite): implement connector scope listing * test(connectors): cover required table listing * feat(cli): add warehouse driver registry * refactor(setup): route scope discovery through driver registry * refactor(cli): route local query execution through driver registry * refactor(historic-sql): route dialect support through driver registry * refactor(cli): test warehouse connections through driver registry * fix(cli): close driver registry type export gaps * Improve setup daemon diagnostics * refactor(setup): centralize rail-prefixed diagnostics + query-history fallback Extract errorMessage, writePrefixedLines, and flushPrefixedBufferedCommandOutput into clack.ts so the setup wizard, managed daemons, and embedding/agent steps share one rail-formatted writer. setup-databases.ts also adds a "disable query history and retry" option when the schema-context build fails and query history is the likely culprit, surfaced via a new failed-query-history-unavailable status. * fix(cli): carry catalog through the picker so BigQuery/Snowflake/SQL Server scope filters match The setup picker's KtxTableListEntry was a 2-level { schema, name }, so qualifiedTableId always wrote db.name into enabled_tables. When BigQuery, Snowflake, or SQL Server later ran fast ingest, their introspect step filtered the scope set with scopedTableNames(scope, { catalog: projectId|database, db }) — catalog was non-null on the introspect side but null in the scope refs, so every entry was rejected, the live-database adapter staged zero table files, and detect() failed with 'Adapter "live-database" did not recognize fetched source output'. Align the picker boundary with the canonical 3-level KtxTableRef: - Add catalog: string | null to KtxTableListEntry. - BigQuery/Snowflake/SQL Server listTables populate catalog from the resolved projectId / database; Postgres/MySQL/ClickHouse/SQLite set null. - qualifiedTableId emits catalog.schema.name when catalog is non-null (resolveEnabledTables already accepts the 3-part shape) and schemasFromEnabledTables now goes through parseDottedTableEntry so it recovers the schema correctly from both 2-part and 3-part entries. - Export parseDottedTableEntry from enabled-tables.ts (@internal) for picker reuse. Update listTables expectations in all seven connector tests and the setup / picker test fixtures. Add a picker regression test that covers the catalog-bearing round-trip (save + refine). * fix(cli): allow debug telemetry under opt-out env
2026-05-26 08:49:05 +02:00
} from '../src/setup-models.js';
2026-05-10 23:12:26 +02:00
function makeIo() {
let stdout = '';
let stderr = '';
return {
io: {
stdout: {
isTTY: true,
write: (chunk: string) => {
stdout += chunk;
},
},
stderr: {
write: (chunk: string) => {
stderr += chunk;
},
},
},
stdout: () => stdout,
stderr: () => stderr,
};
}
function makeSpinnerEvents() {
const events: string[] = [];
const spinner = vi.fn(() => ({
start: (msg: string) => events.push(`start:${msg}`),
message: (msg: string) => events.push(`message:${msg}`),
stop: (msg: string) => events.push(`stop:${msg}`),
error: (msg: string) => events.push(`error:${msg}`),
}));
return { events, spinner };
}
2026-05-10 23:12:26 +02:00
function makePromptAdapter(options: {
providerChoice?: string;
2026-05-10 23:12:26 +02:00
selectValues?: string[];
credentialChoice?: string;
modelChoice?: string;
textValues?: string[];
passwordValue?: string;
passwordValues?: Array<string | undefined>;
2026-05-10 23:51:24 +02:00
}): KtxSetupModelPromptAdapter {
2026-05-10 23:12:26 +02:00
const selectValues = [...(options.selectValues ?? [])];
const textValues = [...(options.textValues ?? [])];
const passwordValues = [...(options.passwordValues ?? [])];
let providerPromptCount = 0;
const choose = async ({ message }: { message: string }) => {
if (message.includes('LLM provider')) {
providerPromptCount += 1;
const nextProviderChoice = selectValues[0];
if (
nextProviderChoice === 'anthropic' ||
nextProviderChoice === 'vertex' ||
nextProviderChoice === 'claude-code' ||
feat: add codex llm backend for ktx runtime work (#253) * feat: add codex sdk runner foundation * feat: parse codex runtime events * feat: expose codex runtime mcp tools * feat: add codex llm runtime * feat: wire codex llm backend * test: avoid Array.fromAsync in codex runner test * docs: document codex llm backend * fix: tighten codex runtime config ownership * fix: use codex sdk env and thread options * fix: parse codex sdk event shapes * test: add codex backend live smoke * docs: clarify codex backend isolation * fix: drive codex loop metrics from mcp events * fix: enforce codex local step budget * docs: disclose codex isolation limits * fix: count all codex agent steps and stream step callbacks live The agent-loop step budget only counted completed mcp_tool_call items, so built-in command_execution steps (which the public Codex SDK/CLI surface can still expose) never decremented the budget, letting ingest/reconciliation run past stepBudget until Codex stopped on its own. onStepFinish was also replayed only after the whole stream drained, so live work_unit_step / reconciliation progress appeared stuck until the Codex process exited. collectEvents is now the single live step accumulator: it counts every completed agent-action item via a shared isCompletedAgentStep predicate (command_execution, mcp_tool_call, file_change, web_search), fires onStepFinish as each step completes, and enforces the budget on that broader count. A no-tool turn still counts as one step. toolFailures stays MCP-specific, since a non-zero command exit is normal agent exploration, not a loop failure. * test: align ingest llm-guard assertions with codex backend The skip-llm ingest guard message now lists codex as a valid backend and mentions a Claude Code/Codex session plus a codex setup hint, but this slow suite test still asserted the pre-codex wording. Update it to match the production message (already covered by the local-bundle-runtime unit test) and add the codex setup-line assertion. * fix: treat codex error:null tool calls as success The Codex SDK serializes error: null on successful mcp_tool_call items, so the failure check (item.error !== undefined) flagged every successful tool call as failed with the empty-payload default "Codex turn failed". This killed every ingest work unit under the codex backend before it could produce a patch. Key on status === 'failed' (authoritative, always set) and only treat a populated error object as a failure. Add a regression test built from a verbatim real-SDK event capture. * fix: default codex backend to gpt-5.5 and report real probe errors The previous default gpt-5.3-codex is an API-key-only model that the OpenAI API rejects under ChatGPT-account (subscription) auth, so codex status/setup failed with a misleading "authentication is not usable" message even though auth was fine. - Default codex model is now gpt-5.5 (works on both subscription and API-key auth); the curated setup picker offers gpt-5.5 / gpt-5.4 / gpt-5.4-mini and keeps free-form entry for account-specific ids (e.g. gpt-5.3-codex-spark). - runCodexAuthProbe now distinguishes "model not available" from an auth failure and surfaces the real API error: collectEvents retains stream events when the SDK throws on a non-zero exit, and the API error JSON envelope is unwrapped to its human-readable message. - The Codex isolation warning now renders inside the clack setup frame. - Docs updated to gpt-5.5 with a note that *-codex ids require API-key auth. * fix: require llm.models.default in status and match codex probe remediation Status reported a project ready when a non-none LLM backend was configured without llm.models.default, but the runtime (resolveModelSlots) hard-requires it, so ingest/scan/memory threw after `ktx status` said the project was usable. buildLlmStatus now fails for any non-none backend missing models.default and no longer invents a fallback model for claude-code/codex. Codex probe failures now carry a category-matched fix: a model-access failure steers the user at llm.models.default instead of the auth/install remediation. runCodexAuthProbe returns the fix and status consumes it; the message stays self-sufficient so setup output is unchanged. Docs: README now lists the codex backend and local Codex auth; ktx-setup.mdx states --llm-model only accepts codex/default or gpt-*/codex-* ids. Repaired four doctor fixtures that configured a backend without models.default (the now-correctly-blocked config) and added coverage for the new behavior.
2026-06-02 13:57:11 +02:00
nextProviderChoice === 'codex' ||
nextProviderChoice === 'back'
) {
return selectValues.shift() ?? nextProviderChoice;
2026-05-10 23:12:26 +02:00
}
if (options.credentialChoice === 'back' && providerPromptCount > 1) {
return 'back';
2026-05-10 23:12:26 +02:00
}
return options.providerChoice ?? 'anthropic';
}
const nextValue = selectValues.shift();
if (nextValue) {
return nextValue;
}
if (message.includes('Anthropic API key')) {
return options.credentialChoice ?? 'env';
}
return options.modelChoice ?? 'claude-sonnet-4-6';
};
return {
select: vi.fn(choose),
autocomplete: vi.fn(choose),
2026-05-10 23:12:26 +02:00
text: vi.fn(async () => textValues.shift() ?? ''),
password: vi.fn(
async () =>
passwordValues.length > 0 ? passwordValues.shift() : options.passwordValue ?? 'sk-ant-pasted', // pragma: allowlist secret
),
2026-05-10 23:12:26 +02:00
cancel: vi.fn(),
};
}
const anthropicPreset = {
default: 'claude-sonnet-4-6',
triage: 'claude-haiku-4-5',
candidateExtraction: 'claude-sonnet-4-6',
curator: 'claude-opus-4-7',
reconcile: 'claude-opus-4-7',
repair: 'claude-haiku-4-5',
};
const claudeCodePreset = {
default: 'sonnet',
triage: 'haiku',
candidateExtraction: 'sonnet',
curator: 'opus',
reconcile: 'opus',
repair: 'haiku',
};
const codexPreset = {
default: 'gpt-5.5',
triage: 'gpt-5.5',
candidateExtraction: 'gpt-5.5',
curator: 'gpt-5.5',
reconcile: 'gpt-5.5',
repair: 'gpt-5.5',
};
2026-05-10 23:12:26 +02:00
describe('setup Anthropic model step', () => {
let tempDir: string;
beforeEach(async () => {
2026-05-10 23:51:24 +02:00
tempDir = await mkdtemp(join(tmpdir(), 'ktx-setup-models-'));
await initKtxProject({ projectDir: tempDir });
2026-05-10 23:12:26 +02:00
});
afterEach(async () => {
await rm(tempDir, { recursive: true, force: true });
});
it('offers Anthropic provider paths in the preferred order', async () => {
const prompts = makePromptAdapter({ providerChoice: 'back' });
const result = await runKtxSetupAnthropicModelStep(
{ projectDir: tempDir, inputMode: 'auto', skipLlm: false },
makeIo().io,
{ prompts, env: {} },
);
expect(result.status).toBe('back');
expect(prompts.select).toHaveBeenCalledWith(
expect.objectContaining({
message: expect.stringContaining('Which LLM provider should KTX use?'),
options: [
{ value: 'claude-code', label: 'Claude subscription (Pro/Max)' },
feat: add codex llm backend for ktx runtime work (#253) * feat: add codex sdk runner foundation * feat: parse codex runtime events * feat: expose codex runtime mcp tools * feat: add codex llm runtime * feat: wire codex llm backend * test: avoid Array.fromAsync in codex runner test * docs: document codex llm backend * fix: tighten codex runtime config ownership * fix: use codex sdk env and thread options * fix: parse codex sdk event shapes * test: add codex backend live smoke * docs: clarify codex backend isolation * fix: drive codex loop metrics from mcp events * fix: enforce codex local step budget * docs: disclose codex isolation limits * fix: count all codex agent steps and stream step callbacks live The agent-loop step budget only counted completed mcp_tool_call items, so built-in command_execution steps (which the public Codex SDK/CLI surface can still expose) never decremented the budget, letting ingest/reconciliation run past stepBudget until Codex stopped on its own. onStepFinish was also replayed only after the whole stream drained, so live work_unit_step / reconciliation progress appeared stuck until the Codex process exited. collectEvents is now the single live step accumulator: it counts every completed agent-action item via a shared isCompletedAgentStep predicate (command_execution, mcp_tool_call, file_change, web_search), fires onStepFinish as each step completes, and enforces the budget on that broader count. A no-tool turn still counts as one step. toolFailures stays MCP-specific, since a non-zero command exit is normal agent exploration, not a loop failure. * test: align ingest llm-guard assertions with codex backend The skip-llm ingest guard message now lists codex as a valid backend and mentions a Claude Code/Codex session plus a codex setup hint, but this slow suite test still asserted the pre-codex wording. Update it to match the production message (already covered by the local-bundle-runtime unit test) and add the codex setup-line assertion. * fix: treat codex error:null tool calls as success The Codex SDK serializes error: null on successful mcp_tool_call items, so the failure check (item.error !== undefined) flagged every successful tool call as failed with the empty-payload default "Codex turn failed". This killed every ingest work unit under the codex backend before it could produce a patch. Key on status === 'failed' (authoritative, always set) and only treat a populated error object as a failure. Add a regression test built from a verbatim real-SDK event capture. * fix: default codex backend to gpt-5.5 and report real probe errors The previous default gpt-5.3-codex is an API-key-only model that the OpenAI API rejects under ChatGPT-account (subscription) auth, so codex status/setup failed with a misleading "authentication is not usable" message even though auth was fine. - Default codex model is now gpt-5.5 (works on both subscription and API-key auth); the curated setup picker offers gpt-5.5 / gpt-5.4 / gpt-5.4-mini and keeps free-form entry for account-specific ids (e.g. gpt-5.3-codex-spark). - runCodexAuthProbe now distinguishes "model not available" from an auth failure and surfaces the real API error: collectEvents retains stream events when the SDK throws on a non-zero exit, and the API error JSON envelope is unwrapped to its human-readable message. - The Codex isolation warning now renders inside the clack setup frame. - Docs updated to gpt-5.5 with a note that *-codex ids require API-key auth. * fix: require llm.models.default in status and match codex probe remediation Status reported a project ready when a non-none LLM backend was configured without llm.models.default, but the runtime (resolveModelSlots) hard-requires it, so ingest/scan/memory threw after `ktx status` said the project was usable. buildLlmStatus now fails for any non-none backend missing models.default and no longer invents a fallback model for claude-code/codex. Codex probe failures now carry a category-matched fix: a model-access failure steers the user at llm.models.default instead of the auth/install remediation. runCodexAuthProbe returns the fix and status consumes it; the message stays self-sufficient so setup output is unchanged. Docs: README now lists the codex backend and local Codex auth; ktx-setup.mdx states --llm-model only accepts codex/default or gpt-*/codex-* ids. Repaired four doctor fixtures that configured a backend without models.default (the now-correctly-blocked config) and added coverage for the new behavior.
2026-06-02 13:57:11 +02:00
{ value: 'codex', label: 'Codex subscription' },
{ value: 'anthropic', label: 'Anthropic API key' },
{ value: 'vertex', label: 'Google Vertex AI for Anthropic Claude' },
{ value: 'back', label: 'Back' },
],
}),
);
});
feat: add claude-code llm backend with runtime port (#115) * docs: revise claude-code ingest backend spec * docs: keep claude-code spec focused on ingest * docs: expand claude-code spec to full llm parity * Refine claude-code backend spec after adversarial review iteration 1 * Refine claude-code backend spec after adversarial review iteration 2 * Refine claude-code backend spec after adversarial review iteration 3 * feat: recognize claude-code llm backend * feat: add ktx llm runtime port * feat: add claude-code llm runtime * feat: route non-agent llm calls through runtime * feat: run ingest agents through llm runtime * feat: support claude-code setup and status * test: verify claude-code backend runtime * docs: add claude-code backend v1 runtime plan * fix: close claude-code runtime isolation checks * fix: warn on claude-code prompt caching during setup * chore: verify claude-code v1 closure * docs: add claude-code backend v1 isolation closure plan * fix: update claude-code ingest setup guidance * docs: add claude-code backend v1 ingest guidance closure plan * docs: align claude-code isolation spec with sdk metadata * test: cover claude-code host discovery metadata * fix: tolerate claude-code host discovery metadata * docs: clarify claude-code host discovery metadata * docs: add claude-code auth-probe isolation fix plan * chore: prepare kaelio ktx rc1 release * chore: add semantic release workflow * fix: unblock ci checks * chore(release): 0.1.0-rc.1 * feat: add Claude Code model selection to setup * fix: keep git maintenance attached in local repos
2026-05-16 12:06:34 +02:00
it('configures Claude Code backend and validates local auth', async () => {
const io = makeIo();
const authProbe = vi.fn(async () => ({ ok: true as const }));
const result = await runKtxSetupAnthropicModelStep(
{
projectDir: tempDir,
inputMode: 'disabled',
llmBackend: 'claude-code',
skipLlm: false,
},
io.io,
{ claudeCodeAuthProbe: authProbe },
);
expect(result.status).toBe('ready');
const config = parseKtxProjectConfig(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8'));
expect(config.llm).toMatchObject({
provider: { backend: 'claude-code' },
models: claudeCodePreset,
feat: add claude-code llm backend with runtime port (#115) * docs: revise claude-code ingest backend spec * docs: keep claude-code spec focused on ingest * docs: expand claude-code spec to full llm parity * Refine claude-code backend spec after adversarial review iteration 1 * Refine claude-code backend spec after adversarial review iteration 2 * Refine claude-code backend spec after adversarial review iteration 3 * feat: recognize claude-code llm backend * feat: add ktx llm runtime port * feat: add claude-code llm runtime * feat: route non-agent llm calls through runtime * feat: run ingest agents through llm runtime * feat: support claude-code setup and status * test: verify claude-code backend runtime * docs: add claude-code backend v1 runtime plan * fix: close claude-code runtime isolation checks * fix: warn on claude-code prompt caching during setup * chore: verify claude-code v1 closure * docs: add claude-code backend v1 isolation closure plan * fix: update claude-code ingest setup guidance * docs: add claude-code backend v1 ingest guidance closure plan * docs: align claude-code isolation spec with sdk metadata * test: cover claude-code host discovery metadata * fix: tolerate claude-code host discovery metadata * docs: clarify claude-code host discovery metadata * docs: add claude-code auth-probe isolation fix plan * chore: prepare kaelio ktx rc1 release * chore: add semantic release workflow * fix: unblock ci checks * chore(release): 0.1.0-rc.1 * feat: add Claude Code model selection to setup * fix: keep git maintenance attached in local repos
2026-05-16 12:06:34 +02:00
});
expect(authProbe).toHaveBeenCalledTimes(3);
expect(authProbe).toHaveBeenNthCalledWith(1, expect.objectContaining({ projectDir: tempDir, model: 'sonnet' }));
expect(authProbe).toHaveBeenNthCalledWith(2, expect.objectContaining({ projectDir: tempDir, model: 'haiku' }));
expect(authProbe).toHaveBeenNthCalledWith(3, expect.objectContaining({ projectDir: tempDir, model: 'opus' }));
});
it('does not prompt for a Claude Code model during interactive setup', async () => {
const io = makeIo();
const prompts = makePromptAdapter({ selectValues: ['claude-code'] });
const authProbe = vi.fn(async () => ({ ok: true as const }));
const result = await runKtxSetupAnthropicModelStep(
{ projectDir: tempDir, inputMode: 'auto', skipLlm: false },
io.io,
{ prompts, claudeCodeAuthProbe: authProbe },
);
expect(result.status).toBe('ready');
expect(prompts.select).toHaveBeenCalledWith(
expect.objectContaining({
message: expect.stringContaining('Which LLM provider should KTX use?'),
}),
);
expect(prompts.select).not.toHaveBeenCalledWith(
expect.objectContaining({
message: expect.stringContaining('Which Claude Code model should KTX use?'),
}),
);
const config = parseKtxProjectConfig(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8'));
expect(config.llm.models).toMatchObject(claudeCodePreset);
feat: add claude-code llm backend with runtime port (#115) * docs: revise claude-code ingest backend spec * docs: keep claude-code spec focused on ingest * docs: expand claude-code spec to full llm parity * Refine claude-code backend spec after adversarial review iteration 1 * Refine claude-code backend spec after adversarial review iteration 2 * Refine claude-code backend spec after adversarial review iteration 3 * feat: recognize claude-code llm backend * feat: add ktx llm runtime port * feat: add claude-code llm runtime * feat: route non-agent llm calls through runtime * feat: run ingest agents through llm runtime * feat: support claude-code setup and status * test: verify claude-code backend runtime * docs: add claude-code backend v1 runtime plan * fix: close claude-code runtime isolation checks * fix: warn on claude-code prompt caching during setup * chore: verify claude-code v1 closure * docs: add claude-code backend v1 isolation closure plan * fix: update claude-code ingest setup guidance * docs: add claude-code backend v1 ingest guidance closure plan * docs: align claude-code isolation spec with sdk metadata * test: cover claude-code host discovery metadata * fix: tolerate claude-code host discovery metadata * docs: clarify claude-code host discovery metadata * docs: add claude-code auth-probe isolation fix plan * chore: prepare kaelio ktx rc1 release * chore: add semantic release workflow * fix: unblock ci checks * chore(release): 0.1.0-rc.1 * feat: add Claude Code model selection to setup * fix: keep git maintenance attached in local repos
2026-05-16 12:06:34 +02:00
});
feat: add codex llm backend for ktx runtime work (#253) * feat: add codex sdk runner foundation * feat: parse codex runtime events * feat: expose codex runtime mcp tools * feat: add codex llm runtime * feat: wire codex llm backend * test: avoid Array.fromAsync in codex runner test * docs: document codex llm backend * fix: tighten codex runtime config ownership * fix: use codex sdk env and thread options * fix: parse codex sdk event shapes * test: add codex backend live smoke * docs: clarify codex backend isolation * fix: drive codex loop metrics from mcp events * fix: enforce codex local step budget * docs: disclose codex isolation limits * fix: count all codex agent steps and stream step callbacks live The agent-loop step budget only counted completed mcp_tool_call items, so built-in command_execution steps (which the public Codex SDK/CLI surface can still expose) never decremented the budget, letting ingest/reconciliation run past stepBudget until Codex stopped on its own. onStepFinish was also replayed only after the whole stream drained, so live work_unit_step / reconciliation progress appeared stuck until the Codex process exited. collectEvents is now the single live step accumulator: it counts every completed agent-action item via a shared isCompletedAgentStep predicate (command_execution, mcp_tool_call, file_change, web_search), fires onStepFinish as each step completes, and enforces the budget on that broader count. A no-tool turn still counts as one step. toolFailures stays MCP-specific, since a non-zero command exit is normal agent exploration, not a loop failure. * test: align ingest llm-guard assertions with codex backend The skip-llm ingest guard message now lists codex as a valid backend and mentions a Claude Code/Codex session plus a codex setup hint, but this slow suite test still asserted the pre-codex wording. Update it to match the production message (already covered by the local-bundle-runtime unit test) and add the codex setup-line assertion. * fix: treat codex error:null tool calls as success The Codex SDK serializes error: null on successful mcp_tool_call items, so the failure check (item.error !== undefined) flagged every successful tool call as failed with the empty-payload default "Codex turn failed". This killed every ingest work unit under the codex backend before it could produce a patch. Key on status === 'failed' (authoritative, always set) and only treat a populated error object as a failure. Add a regression test built from a verbatim real-SDK event capture. * fix: default codex backend to gpt-5.5 and report real probe errors The previous default gpt-5.3-codex is an API-key-only model that the OpenAI API rejects under ChatGPT-account (subscription) auth, so codex status/setup failed with a misleading "authentication is not usable" message even though auth was fine. - Default codex model is now gpt-5.5 (works on both subscription and API-key auth); the curated setup picker offers gpt-5.5 / gpt-5.4 / gpt-5.4-mini and keeps free-form entry for account-specific ids (e.g. gpt-5.3-codex-spark). - runCodexAuthProbe now distinguishes "model not available" from an auth failure and surfaces the real API error: collectEvents retains stream events when the SDK throws on a non-zero exit, and the API error JSON envelope is unwrapped to its human-readable message. - The Codex isolation warning now renders inside the clack setup frame. - Docs updated to gpt-5.5 with a note that *-codex ids require API-key auth. * fix: require llm.models.default in status and match codex probe remediation Status reported a project ready when a non-none LLM backend was configured without llm.models.default, but the runtime (resolveModelSlots) hard-requires it, so ingest/scan/memory threw after `ktx status` said the project was usable. buildLlmStatus now fails for any non-none backend missing models.default and no longer invents a fallback model for claude-code/codex. Codex probe failures now carry a category-matched fix: a model-access failure steers the user at llm.models.default instead of the auth/install remediation. runCodexAuthProbe returns the fix and status consumes it; the message stays self-sufficient so setup output is unchanged. Docs: README now lists the codex backend and local Codex auth; ktx-setup.mdx states --llm-model only accepts codex/default or gpt-*/codex-* ids. Repaired four doctor fixtures that configured a backend without models.default (the now-correctly-blocked config) and added coverage for the new behavior.
2026-06-02 13:57:11 +02:00
it('configures Codex backend and validates local auth', async () => {
const io = makeIo();
const codexAuthProbe = vi.fn(async () => ({ ok: true as const }));
const result = await runKtxSetupAnthropicModelStep(
{
projectDir: tempDir,
inputMode: 'disabled',
llmBackend: 'codex',
skipLlm: false,
},
io.io,
{ codexAuthProbe },
);
expect(result.status).toBe('ready');
const config = parseKtxProjectConfig(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8'));
expect(config.llm).toMatchObject({
provider: { backend: 'codex' },
models: codexPreset,
feat: add codex llm backend for ktx runtime work (#253) * feat: add codex sdk runner foundation * feat: parse codex runtime events * feat: expose codex runtime mcp tools * feat: add codex llm runtime * feat: wire codex llm backend * test: avoid Array.fromAsync in codex runner test * docs: document codex llm backend * fix: tighten codex runtime config ownership * fix: use codex sdk env and thread options * fix: parse codex sdk event shapes * test: add codex backend live smoke * docs: clarify codex backend isolation * fix: drive codex loop metrics from mcp events * fix: enforce codex local step budget * docs: disclose codex isolation limits * fix: count all codex agent steps and stream step callbacks live The agent-loop step budget only counted completed mcp_tool_call items, so built-in command_execution steps (which the public Codex SDK/CLI surface can still expose) never decremented the budget, letting ingest/reconciliation run past stepBudget until Codex stopped on its own. onStepFinish was also replayed only after the whole stream drained, so live work_unit_step / reconciliation progress appeared stuck until the Codex process exited. collectEvents is now the single live step accumulator: it counts every completed agent-action item via a shared isCompletedAgentStep predicate (command_execution, mcp_tool_call, file_change, web_search), fires onStepFinish as each step completes, and enforces the budget on that broader count. A no-tool turn still counts as one step. toolFailures stays MCP-specific, since a non-zero command exit is normal agent exploration, not a loop failure. * test: align ingest llm-guard assertions with codex backend The skip-llm ingest guard message now lists codex as a valid backend and mentions a Claude Code/Codex session plus a codex setup hint, but this slow suite test still asserted the pre-codex wording. Update it to match the production message (already covered by the local-bundle-runtime unit test) and add the codex setup-line assertion. * fix: treat codex error:null tool calls as success The Codex SDK serializes error: null on successful mcp_tool_call items, so the failure check (item.error !== undefined) flagged every successful tool call as failed with the empty-payload default "Codex turn failed". This killed every ingest work unit under the codex backend before it could produce a patch. Key on status === 'failed' (authoritative, always set) and only treat a populated error object as a failure. Add a regression test built from a verbatim real-SDK event capture. * fix: default codex backend to gpt-5.5 and report real probe errors The previous default gpt-5.3-codex is an API-key-only model that the OpenAI API rejects under ChatGPT-account (subscription) auth, so codex status/setup failed with a misleading "authentication is not usable" message even though auth was fine. - Default codex model is now gpt-5.5 (works on both subscription and API-key auth); the curated setup picker offers gpt-5.5 / gpt-5.4 / gpt-5.4-mini and keeps free-form entry for account-specific ids (e.g. gpt-5.3-codex-spark). - runCodexAuthProbe now distinguishes "model not available" from an auth failure and surfaces the real API error: collectEvents retains stream events when the SDK throws on a non-zero exit, and the API error JSON envelope is unwrapped to its human-readable message. - The Codex isolation warning now renders inside the clack setup frame. - Docs updated to gpt-5.5 with a note that *-codex ids require API-key auth. * fix: require llm.models.default in status and match codex probe remediation Status reported a project ready when a non-none LLM backend was configured without llm.models.default, but the runtime (resolveModelSlots) hard-requires it, so ingest/scan/memory threw after `ktx status` said the project was usable. buildLlmStatus now fails for any non-none backend missing models.default and no longer invents a fallback model for claude-code/codex. Codex probe failures now carry a category-matched fix: a model-access failure steers the user at llm.models.default instead of the auth/install remediation. runCodexAuthProbe returns the fix and status consumes it; the message stays self-sufficient so setup output is unchanged. Docs: README now lists the codex backend and local Codex auth; ktx-setup.mdx states --llm-model only accepts codex/default or gpt-*/codex-* ids. Repaired four doctor fixtures that configured a backend without models.default (the now-correctly-blocked config) and added coverage for the new behavior.
2026-06-02 13:57:11 +02:00
});
expect(codexAuthProbe).toHaveBeenCalledTimes(1);
feat: add codex llm backend for ktx runtime work (#253) * feat: add codex sdk runner foundation * feat: parse codex runtime events * feat: expose codex runtime mcp tools * feat: add codex llm runtime * feat: wire codex llm backend * test: avoid Array.fromAsync in codex runner test * docs: document codex llm backend * fix: tighten codex runtime config ownership * fix: use codex sdk env and thread options * fix: parse codex sdk event shapes * test: add codex backend live smoke * docs: clarify codex backend isolation * fix: drive codex loop metrics from mcp events * fix: enforce codex local step budget * docs: disclose codex isolation limits * fix: count all codex agent steps and stream step callbacks live The agent-loop step budget only counted completed mcp_tool_call items, so built-in command_execution steps (which the public Codex SDK/CLI surface can still expose) never decremented the budget, letting ingest/reconciliation run past stepBudget until Codex stopped on its own. onStepFinish was also replayed only after the whole stream drained, so live work_unit_step / reconciliation progress appeared stuck until the Codex process exited. collectEvents is now the single live step accumulator: it counts every completed agent-action item via a shared isCompletedAgentStep predicate (command_execution, mcp_tool_call, file_change, web_search), fires onStepFinish as each step completes, and enforces the budget on that broader count. A no-tool turn still counts as one step. toolFailures stays MCP-specific, since a non-zero command exit is normal agent exploration, not a loop failure. * test: align ingest llm-guard assertions with codex backend The skip-llm ingest guard message now lists codex as a valid backend and mentions a Claude Code/Codex session plus a codex setup hint, but this slow suite test still asserted the pre-codex wording. Update it to match the production message (already covered by the local-bundle-runtime unit test) and add the codex setup-line assertion. * fix: treat codex error:null tool calls as success The Codex SDK serializes error: null on successful mcp_tool_call items, so the failure check (item.error !== undefined) flagged every successful tool call as failed with the empty-payload default "Codex turn failed". This killed every ingest work unit under the codex backend before it could produce a patch. Key on status === 'failed' (authoritative, always set) and only treat a populated error object as a failure. Add a regression test built from a verbatim real-SDK event capture. * fix: default codex backend to gpt-5.5 and report real probe errors The previous default gpt-5.3-codex is an API-key-only model that the OpenAI API rejects under ChatGPT-account (subscription) auth, so codex status/setup failed with a misleading "authentication is not usable" message even though auth was fine. - Default codex model is now gpt-5.5 (works on both subscription and API-key auth); the curated setup picker offers gpt-5.5 / gpt-5.4 / gpt-5.4-mini and keeps free-form entry for account-specific ids (e.g. gpt-5.3-codex-spark). - runCodexAuthProbe now distinguishes "model not available" from an auth failure and surfaces the real API error: collectEvents retains stream events when the SDK throws on a non-zero exit, and the API error JSON envelope is unwrapped to its human-readable message. - The Codex isolation warning now renders inside the clack setup frame. - Docs updated to gpt-5.5 with a note that *-codex ids require API-key auth. * fix: require llm.models.default in status and match codex probe remediation Status reported a project ready when a non-none LLM backend was configured without llm.models.default, but the runtime (resolveModelSlots) hard-requires it, so ingest/scan/memory threw after `ktx status` said the project was usable. buildLlmStatus now fails for any non-none backend missing models.default and no longer invents a fallback model for claude-code/codex. Codex probe failures now carry a category-matched fix: a model-access failure steers the user at llm.models.default instead of the auth/install remediation. runCodexAuthProbe returns the fix and status consumes it; the message stays self-sufficient so setup output is unchanged. Docs: README now lists the codex backend and local Codex auth; ktx-setup.mdx states --llm-model only accepts codex/default or gpt-*/codex-* ids. Repaired four doctor fixtures that configured a backend without models.default (the now-correctly-blocked config) and added coverage for the new behavior.
2026-06-02 13:57:11 +02:00
expect(codexAuthProbe).toHaveBeenCalledWith(expect.objectContaining({ projectDir: tempDir, model: 'gpt-5.5' }));
// The warning carries the clack gutter so it renders inside the setup frame.
expect(io.stderr()).toContain('│ Codex backend isolation is limited');
expect(io.stderr()).toContain('may still load user Codex config');
});
it('defaults the Codex model to gpt-5.5 when none is provided non-interactively', async () => {
const io = makeIo();
const codexAuthProbe = vi.fn(async () => ({ ok: true as const }));
const result = await runKtxSetupAnthropicModelStep(
{
projectDir: tempDir,
inputMode: 'disabled',
llmBackend: 'codex',
skipLlm: false,
},
io.io,
{ codexAuthProbe },
);
expect(result.status).toBe('ready');
const config = parseKtxProjectConfig(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8'));
expect(config.llm).toMatchObject({
provider: { backend: 'codex' },
models: codexPreset,
feat: add codex llm backend for ktx runtime work (#253) * feat: add codex sdk runner foundation * feat: parse codex runtime events * feat: expose codex runtime mcp tools * feat: add codex llm runtime * feat: wire codex llm backend * test: avoid Array.fromAsync in codex runner test * docs: document codex llm backend * fix: tighten codex runtime config ownership * fix: use codex sdk env and thread options * fix: parse codex sdk event shapes * test: add codex backend live smoke * docs: clarify codex backend isolation * fix: drive codex loop metrics from mcp events * fix: enforce codex local step budget * docs: disclose codex isolation limits * fix: count all codex agent steps and stream step callbacks live The agent-loop step budget only counted completed mcp_tool_call items, so built-in command_execution steps (which the public Codex SDK/CLI surface can still expose) never decremented the budget, letting ingest/reconciliation run past stepBudget until Codex stopped on its own. onStepFinish was also replayed only after the whole stream drained, so live work_unit_step / reconciliation progress appeared stuck until the Codex process exited. collectEvents is now the single live step accumulator: it counts every completed agent-action item via a shared isCompletedAgentStep predicate (command_execution, mcp_tool_call, file_change, web_search), fires onStepFinish as each step completes, and enforces the budget on that broader count. A no-tool turn still counts as one step. toolFailures stays MCP-specific, since a non-zero command exit is normal agent exploration, not a loop failure. * test: align ingest llm-guard assertions with codex backend The skip-llm ingest guard message now lists codex as a valid backend and mentions a Claude Code/Codex session plus a codex setup hint, but this slow suite test still asserted the pre-codex wording. Update it to match the production message (already covered by the local-bundle-runtime unit test) and add the codex setup-line assertion. * fix: treat codex error:null tool calls as success The Codex SDK serializes error: null on successful mcp_tool_call items, so the failure check (item.error !== undefined) flagged every successful tool call as failed with the empty-payload default "Codex turn failed". This killed every ingest work unit under the codex backend before it could produce a patch. Key on status === 'failed' (authoritative, always set) and only treat a populated error object as a failure. Add a regression test built from a verbatim real-SDK event capture. * fix: default codex backend to gpt-5.5 and report real probe errors The previous default gpt-5.3-codex is an API-key-only model that the OpenAI API rejects under ChatGPT-account (subscription) auth, so codex status/setup failed with a misleading "authentication is not usable" message even though auth was fine. - Default codex model is now gpt-5.5 (works on both subscription and API-key auth); the curated setup picker offers gpt-5.5 / gpt-5.4 / gpt-5.4-mini and keeps free-form entry for account-specific ids (e.g. gpt-5.3-codex-spark). - runCodexAuthProbe now distinguishes "model not available" from an auth failure and surfaces the real API error: collectEvents retains stream events when the SDK throws on a non-zero exit, and the API error JSON envelope is unwrapped to its human-readable message. - The Codex isolation warning now renders inside the clack setup frame. - Docs updated to gpt-5.5 with a note that *-codex ids require API-key auth. * fix: require llm.models.default in status and match codex probe remediation Status reported a project ready when a non-none LLM backend was configured without llm.models.default, but the runtime (resolveModelSlots) hard-requires it, so ingest/scan/memory threw after `ktx status` said the project was usable. buildLlmStatus now fails for any non-none backend missing models.default and no longer invents a fallback model for claude-code/codex. Codex probe failures now carry a category-matched fix: a model-access failure steers the user at llm.models.default instead of the auth/install remediation. runCodexAuthProbe returns the fix and status consumes it; the message stays self-sufficient so setup output is unchanged. Docs: README now lists the codex backend and local Codex auth; ktx-setup.mdx states --llm-model only accepts codex/default or gpt-*/codex-* ids. Repaired four doctor fixtures that configured a backend without models.default (the now-correctly-blocked config) and added coverage for the new behavior.
2026-06-02 13:57:11 +02:00
});
expect(codexAuthProbe).toHaveBeenCalledTimes(1);
feat: add codex llm backend for ktx runtime work (#253) * feat: add codex sdk runner foundation * feat: parse codex runtime events * feat: expose codex runtime mcp tools * feat: add codex llm runtime * feat: wire codex llm backend * test: avoid Array.fromAsync in codex runner test * docs: document codex llm backend * fix: tighten codex runtime config ownership * fix: use codex sdk env and thread options * fix: parse codex sdk event shapes * test: add codex backend live smoke * docs: clarify codex backend isolation * fix: drive codex loop metrics from mcp events * fix: enforce codex local step budget * docs: disclose codex isolation limits * fix: count all codex agent steps and stream step callbacks live The agent-loop step budget only counted completed mcp_tool_call items, so built-in command_execution steps (which the public Codex SDK/CLI surface can still expose) never decremented the budget, letting ingest/reconciliation run past stepBudget until Codex stopped on its own. onStepFinish was also replayed only after the whole stream drained, so live work_unit_step / reconciliation progress appeared stuck until the Codex process exited. collectEvents is now the single live step accumulator: it counts every completed agent-action item via a shared isCompletedAgentStep predicate (command_execution, mcp_tool_call, file_change, web_search), fires onStepFinish as each step completes, and enforces the budget on that broader count. A no-tool turn still counts as one step. toolFailures stays MCP-specific, since a non-zero command exit is normal agent exploration, not a loop failure. * test: align ingest llm-guard assertions with codex backend The skip-llm ingest guard message now lists codex as a valid backend and mentions a Claude Code/Codex session plus a codex setup hint, but this slow suite test still asserted the pre-codex wording. Update it to match the production message (already covered by the local-bundle-runtime unit test) and add the codex setup-line assertion. * fix: treat codex error:null tool calls as success The Codex SDK serializes error: null on successful mcp_tool_call items, so the failure check (item.error !== undefined) flagged every successful tool call as failed with the empty-payload default "Codex turn failed". This killed every ingest work unit under the codex backend before it could produce a patch. Key on status === 'failed' (authoritative, always set) and only treat a populated error object as a failure. Add a regression test built from a verbatim real-SDK event capture. * fix: default codex backend to gpt-5.5 and report real probe errors The previous default gpt-5.3-codex is an API-key-only model that the OpenAI API rejects under ChatGPT-account (subscription) auth, so codex status/setup failed with a misleading "authentication is not usable" message even though auth was fine. - Default codex model is now gpt-5.5 (works on both subscription and API-key auth); the curated setup picker offers gpt-5.5 / gpt-5.4 / gpt-5.4-mini and keeps free-form entry for account-specific ids (e.g. gpt-5.3-codex-spark). - runCodexAuthProbe now distinguishes "model not available" from an auth failure and surfaces the real API error: collectEvents retains stream events when the SDK throws on a non-zero exit, and the API error JSON envelope is unwrapped to its human-readable message. - The Codex isolation warning now renders inside the clack setup frame. - Docs updated to gpt-5.5 with a note that *-codex ids require API-key auth. * fix: require llm.models.default in status and match codex probe remediation Status reported a project ready when a non-none LLM backend was configured without llm.models.default, but the runtime (resolveModelSlots) hard-requires it, so ingest/scan/memory threw after `ktx status` said the project was usable. buildLlmStatus now fails for any non-none backend missing models.default and no longer invents a fallback model for claude-code/codex. Codex probe failures now carry a category-matched fix: a model-access failure steers the user at llm.models.default instead of the auth/install remediation. runCodexAuthProbe returns the fix and status consumes it; the message stays self-sufficient so setup output is unchanged. Docs: README now lists the codex backend and local Codex auth; ktx-setup.mdx states --llm-model only accepts codex/default or gpt-*/codex-* ids. Repaired four doctor fixtures that configured a backend without models.default (the now-correctly-blocked config) and added coverage for the new behavior.
2026-06-02 13:57:11 +02:00
expect(codexAuthProbe).toHaveBeenCalledWith(expect.objectContaining({ projectDir: tempDir, model: 'gpt-5.5' }));
});
feat: add claude-code llm backend with runtime port (#115) * docs: revise claude-code ingest backend spec * docs: keep claude-code spec focused on ingest * docs: expand claude-code spec to full llm parity * Refine claude-code backend spec after adversarial review iteration 1 * Refine claude-code backend spec after adversarial review iteration 2 * Refine claude-code backend spec after adversarial review iteration 3 * feat: recognize claude-code llm backend * feat: add ktx llm runtime port * feat: add claude-code llm runtime * feat: route non-agent llm calls through runtime * feat: run ingest agents through llm runtime * feat: support claude-code setup and status * test: verify claude-code backend runtime * docs: add claude-code backend v1 runtime plan * fix: close claude-code runtime isolation checks * fix: warn on claude-code prompt caching during setup * chore: verify claude-code v1 closure * docs: add claude-code backend v1 isolation closure plan * fix: update claude-code ingest setup guidance * docs: add claude-code backend v1 ingest guidance closure plan * docs: align claude-code isolation spec with sdk metadata * test: cover claude-code host discovery metadata * fix: tolerate claude-code host discovery metadata * docs: clarify claude-code host discovery metadata * docs: add claude-code auth-probe isolation fix plan * chore: prepare kaelio ktx rc1 release * chore: add semantic release workflow * fix: unblock ci checks * chore(release): 0.1.0-rc.1 * feat: add Claude Code model selection to setup * fix: keep git maintenance attached in local repos
2026-05-16 12:06:34 +02:00
it('warns during Claude Code setup when existing prompt-caching fields will be ignored', async () => {
await writeFile(
join(tempDir, 'ktx.yaml'),
[
'llm:',
' provider:',
' backend: anthropic',
' models:',
' default: claude-sonnet-4-6',
' promptCaching:',
' enabled: true',
' systemTtl: 1h',
' toolsTtl: 1h',
' historyTtl: 5m',
'',
].join('\n'),
'utf-8',
);
const io = makeIo();
const result = await runKtxSetupAnthropicModelStep(
{
projectDir: tempDir,
inputMode: 'disabled',
llmBackend: 'claude-code',
skipLlm: false,
},
io.io,
{
claudeCodeAuthProbe: async () => ({ ok: true as const }),
},
);
expect(result.status).toBe('ready');
expect(io.stderr()).toContain('claude-code ignores llm.promptCaching.systemTtl');
expect(io.stderr()).toContain('Claude Agent SDK does not expose KTX prompt-cache TTL, tool, or history markers');
});
it('returns from Anthropic credential Back to provider selection', async () => {
const prompts = makePromptAdapter({ selectValues: ['anthropic', 'back', 'back'] });
const result = await runKtxSetupAnthropicModelStep(
{ projectDir: tempDir, inputMode: 'auto', skipLlm: false },
makeIo().io,
{ prompts, env: {} },
);
expect(result.status).toBe('back');
expect(prompts.select).toHaveBeenNthCalledWith(
3,
expect.objectContaining({
message: expect.stringContaining('Which LLM provider should KTX use?'),
}),
);
});
2026-05-10 23:12:26 +02:00
it('configures env credentials, selected model, prompt caching, and llm completion state', async () => {
const io = makeIo();
const { events: spinnerEvents, spinner } = makeSpinnerEvents();
2026-05-10 23:51:24 +02:00
const result = await runKtxSetupAnthropicModelStep(
2026-05-10 23:12:26 +02:00
{
projectDir: tempDir,
inputMode: 'disabled',
anthropicApiKeyEnv: 'ANTHROPIC_API_KEY', // pragma: allowlist secret
2026-05-10 23:12:26 +02:00
skipLlm: false,
},
io.io,
{
env: { ANTHROPIC_API_KEY: 'sk-ant-test' }, // pragma: allowlist secret
2026-05-10 23:12:26 +02:00
healthCheck: vi.fn(async () => ({ ok: true as const })),
spinner,
2026-05-10 23:12:26 +02:00
},
);
expect(result.status).toBe('ready');
2026-05-10 23:51:24 +02:00
const config = parseKtxProjectConfig(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8'));
2026-05-10 23:12:26 +02:00
expect(config.llm).toMatchObject({
provider: {
backend: 'anthropic',
anthropic: { api_key: 'env:ANTHROPIC_API_KEY' }, // pragma: allowlist secret
2026-05-10 23:12:26 +02:00
},
models: anthropicPreset,
2026-05-10 23:12:26 +02:00
promptCaching: { enabled: true },
});
expect(config.scan.enrichment.mode).toBe('llm');
expect(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8')).not.toContain('completed_steps:');
expect((await readKtxSetupState(tempDir)).completed_steps).toContain('llm');
expect(spinnerEvents).toEqual([
'start:Checking Anthropic API LLM (claude-sonnet-4-6).',
'stop:LLM test passed (Anthropic API, claude-sonnet-4-6)',
'start:Checking Anthropic API LLM (claude-haiku-4-5).',
'stop:LLM test passed (Anthropic API, claude-haiku-4-5)',
'start:Checking Anthropic API LLM (claude-opus-4-7).',
'stop:LLM test passed (Anthropic API, claude-opus-4-7)',
]);
2026-05-10 23:12:26 +02:00
expect(io.stdout()).toContain('LLM ready: yes');
expect(io.stdout()).not.toContain('sk-ant-test');
});
it('degrades unavailable Anthropic non-anchor models to the anchor before persisting', async () => {
const io = makeIo();
const { events: spinnerEvents, spinner } = makeSpinnerEvents();
const healthCheck = vi
.fn()
.mockResolvedValueOnce({ ok: true as const })
.mockResolvedValueOnce({ ok: false as const, message: 'model not enabled' })
.mockResolvedValueOnce({ ok: true as const });
const result = await runKtxSetupAnthropicModelStep(
{
projectDir: tempDir,
inputMode: 'disabled',
anthropicApiKeyEnv: 'ANTHROPIC_API_KEY', // pragma: allowlist secret
skipLlm: false,
},
io.io,
{
env: { ANTHROPIC_API_KEY: 'sk-ant-test' }, // pragma: allowlist secret
healthCheck,
spinner,
},
);
expect(result.status).toBe('ready');
const config = parseKtxProjectConfig(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8'));
expect(config.llm.models).toMatchObject({
default: 'claude-sonnet-4-6',
triage: 'claude-sonnet-4-6',
candidateExtraction: 'claude-sonnet-4-6',
curator: 'claude-opus-4-7',
reconcile: 'claude-opus-4-7',
repair: 'claude-sonnet-4-6',
});
expect(io.stderr()).toContain(
'LLM model claude-haiku-4-5 is unavailable for triage, repair; using claude-sonnet-4-6 for those roles.',
);
expect(spinnerEvents).toEqual([
'start:Checking Anthropic API LLM (claude-sonnet-4-6).',
'stop:LLM test passed (Anthropic API, claude-sonnet-4-6)',
'start:Checking Anthropic API LLM (claude-haiku-4-5).',
'error:LLM test failed',
'start:Checking Anthropic API LLM (claude-opus-4-7).',
'stop:LLM test passed (Anthropic API, claude-opus-4-7)',
]);
});
it('configures Vertex AI provider, selected model, prompt caching, and llm completion state', async () => {
const io = makeIo();
const healthCheck = vi.fn(async () => ({ ok: true as const }));
const { events: spinnerEvents, spinner } = makeSpinnerEvents();
const result = await runKtxSetupAnthropicModelStep(
{
projectDir: tempDir,
inputMode: 'disabled',
llmBackend: 'vertex',
vertexProject: 'local-gcp-project',
vertexLocation: 'us-east5',
skipLlm: false,
},
io.io,
{ env: {}, healthCheck, spinner },
);
expect(result.status).toBe('ready');
expect(healthCheck).toHaveBeenNthCalledWith(1, {
backend: 'vertex',
vertex: { project: 'local-gcp-project', location: 'us-east5' },
modelSlots: { default: 'claude-sonnet-4-6' },
promptCaching: { enabled: true, vertexFallbackTo5m: true },
});
expect(healthCheck).toHaveBeenNthCalledWith(2, {
backend: 'vertex',
vertex: { project: 'local-gcp-project', location: 'us-east5' },
modelSlots: { default: 'claude-haiku-4-5' },
promptCaching: { enabled: true, vertexFallbackTo5m: true },
});
expect(healthCheck).toHaveBeenNthCalledWith(3, {
backend: 'vertex',
vertex: { project: 'local-gcp-project', location: 'us-east5' },
modelSlots: { default: 'claude-opus-4-7' },
promptCaching: { enabled: true, vertexFallbackTo5m: true },
});
const config = parseKtxProjectConfig(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8'));
expect(config.llm).toMatchObject({
provider: {
backend: 'vertex',
vertex: { project: 'local-gcp-project', location: 'us-east5' },
},
models: anthropicPreset,
promptCaching: { enabled: true, vertexFallbackTo5m: true },
});
expect(config.scan.enrichment.mode).toBe('llm');
expect(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8')).not.toContain('completed_steps:');
expect((await readKtxSetupState(tempDir)).completed_steps).toContain('llm');
expect(spinnerEvents).toEqual([
'start:Checking Vertex AI LLM (claude-sonnet-4-6).',
'stop:LLM test passed (Vertex AI, claude-sonnet-4-6)',
'start:Checking Vertex AI LLM (claude-haiku-4-5).',
'stop:LLM test passed (Vertex AI, claude-haiku-4-5)',
'start:Checking Vertex AI LLM (claude-opus-4-7).',
'stop:LLM test passed (Vertex AI, claude-opus-4-7)',
]);
expect(io.stdout()).toContain('LLM ready: yes (claude-sonnet-4-6)');
});
it('uses existing Vertex AI credentials without an extra auth choice', async () => {
const io = makeIo();
const prompts = makePromptAdapter({ selectValues: ['vertex', 'local-gcp-project'] });
const readGcloudProject = vi.fn(async () => 'local-gcp-project');
const listGcloudProjects = vi.fn(async () => [
{ projectId: 'local-gcp-project', name: 'Local project' },
{ projectId: 'other-gcp-project', name: 'Other project' },
]);
const healthCheck = vi.fn(async () => ({ ok: true as const }));
const result = await runKtxSetupAnthropicModelStep(
{ projectDir: tempDir, inputMode: 'auto', skipLlm: false },
io.io,
{
prompts,
env: {},
readGcloudProject,
listGcloudProjects,
healthCheck,
},
);
expect(result.status).toBe('ready');
expect(prompts.select).not.toHaveBeenCalledWith(
expect.objectContaining({
message: expect.stringContaining('How should KTX authenticate with Google Vertex AI?'),
}),
);
expect(readGcloudProject).toHaveBeenCalled();
expect(listGcloudProjects).toHaveBeenCalled();
expect(prompts.text).not.toHaveBeenCalled();
expect(prompts.autocomplete).toHaveBeenCalledWith(
expect.objectContaining({
message: expect.stringContaining('Which Google Cloud project should KTX use for Vertex AI?'),
options: [
{ value: 'local-gcp-project', label: 'local-gcp-project - Local project (current gcloud project)' },
{ value: 'other-gcp-project', label: 'other-gcp-project - Other project' },
{ value: 'manual', label: 'Enter a project ID manually' },
{ value: 'back', label: 'Back' },
],
}),
);
expect(healthCheck).toHaveBeenCalledWith({
backend: 'vertex',
vertex: { project: 'local-gcp-project', location: 'us-east5' },
modelSlots: { default: 'claude-sonnet-4-6' },
promptCaching: { enabled: true, vertexFallbackTo5m: true },
});
const config = parseKtxProjectConfig(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8'));
expect(config.llm.provider).toMatchObject({
backend: 'vertex',
vertex: { project: 'local-gcp-project', location: 'us-east5' },
});
});
it('skips the Vertex AI auth choice when Application Default Credentials are the only option', async () => {
const io = makeIo();
const prompts = makePromptAdapter({ selectValues: ['vertex', 'local-gcp-project'] });
const healthCheck = vi.fn(async () => ({ ok: true as const }));
const result = await runKtxSetupAnthropicModelStep(
{ projectDir: tempDir, inputMode: 'auto', skipLlm: false },
io.io,
{
prompts,
env: {},
readGcloudProject: vi.fn(async () => 'local-gcp-project'),
listGcloudProjects: vi.fn(async () => [{ projectId: 'local-gcp-project', name: 'Local project' }]),
healthCheck,
},
);
expect(result.status).toBe('ready');
expect(prompts.select).not.toHaveBeenCalledWith(
expect.objectContaining({
message: expect.stringContaining('How should KTX authenticate with Google Vertex AI?'),
}),
);
expect(prompts.autocomplete).toHaveBeenCalledWith(
expect.objectContaining({
message: expect.stringContaining('Which Google Cloud project should KTX use for Vertex AI?'),
}),
);
expect(healthCheck).toHaveBeenCalledWith(
expect.objectContaining({
backend: 'vertex',
vertex: { project: 'local-gcp-project', location: 'us-east5' },
}),
);
});
it('lets users choose a different visible gcloud project for Vertex AI', async () => {
const io = makeIo();
const prompts = makePromptAdapter({ selectValues: ['vertex', 'other-gcp-project'] });
const healthCheck = vi.fn(async () => ({ ok: true as const }));
const result = await runKtxSetupAnthropicModelStep(
{ projectDir: tempDir, inputMode: 'auto', skipLlm: false },
io.io,
{
prompts,
env: {},
readGcloudProject: vi.fn(async () => 'current-gcp-project'),
listGcloudProjects: vi.fn(async () => [
{ projectId: 'current-gcp-project', name: 'Current project' },
{ projectId: 'other-gcp-project', name: 'Other project' },
]),
healthCheck,
},
);
expect(result.status).toBe('ready');
expect(healthCheck).toHaveBeenCalledWith({
backend: 'vertex',
vertex: { project: 'other-gcp-project', location: 'us-east5' },
modelSlots: { default: 'claude-sonnet-4-6' },
promptCaching: { enabled: true, vertexFallbackTo5m: true },
});
const config = parseKtxProjectConfig(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8'));
expect(config.llm.provider).toMatchObject({
backend: 'vertex',
vertex: { project: 'other-gcp-project', location: 'us-east5' },
});
});
it('allows manual Vertex AI project entry when gcloud project listing is empty', async () => {
const io = makeIo();
const prompts = makePromptAdapter({ selectValues: ['vertex', 'manual'], textValues: ['manual-gcp-project'] });
const healthCheck = vi.fn(async () => ({ ok: true as const }));
const result = await runKtxSetupAnthropicModelStep(
{ projectDir: tempDir, inputMode: 'auto', skipLlm: false },
io.io,
{
prompts,
env: {},
readGcloudProject: vi.fn(async () => undefined),
listGcloudProjects: vi.fn(async () => []),
healthCheck,
},
);
expect(result.status).toBe('ready');
expect(prompts.autocomplete).toHaveBeenCalledWith(
expect.objectContaining({
message: expect.stringContaining('Which Google Cloud project should KTX use for Vertex AI?'),
options: [
{ value: 'manual', label: 'Enter a project ID manually' },
{ value: 'back', label: 'Back' },
],
}),
);
expect(prompts.text).toHaveBeenCalledWith(
expect.objectContaining({
message: 'Google Cloud project ID\n│ Press Escape to go back.\n│',
}),
);
expect(healthCheck).toHaveBeenCalledWith(
expect.objectContaining({
vertex: { project: 'manual-gcp-project', location: 'us-east5' },
}),
);
});
it('lets users retry Vertex AI project listing after gcloud auth fails', async () => {
const io = makeIo();
const prompts = makePromptAdapter({ selectValues: ['vertex', 'retry', 'other-gcp-project'] });
const listGcloudProjects = vi
.fn()
.mockRejectedValueOnce(new Error('Reauthentication failed. cannot prompt during non-interactive execution.'))
.mockResolvedValueOnce([
{ projectId: 'local-gcp-project', name: 'Local project' },
{ projectId: 'other-gcp-project', name: 'Other project' },
]);
const healthCheck = vi.fn(async () => ({ ok: true as const }));
const result = await runKtxSetupAnthropicModelStep(
{ projectDir: tempDir, inputMode: 'auto', skipLlm: false },
io.io,
{
prompts,
env: {},
readGcloudProject: vi.fn(async () => 'local-gcp-project'),
listGcloudProjects,
healthCheck,
},
);
expect(result.status).toBe('ready');
expect(listGcloudProjects).toHaveBeenCalledTimes(2);
expect(prompts.autocomplete).toHaveBeenCalledWith(
expect.objectContaining({
message: expect.stringContaining('Could not list Google Cloud projects with gcloud'),
options: expect.arrayContaining([{ value: 'retry', label: 'Retry loading Google Cloud projects' }]),
}),
);
expect(prompts.autocomplete).toHaveBeenCalledWith(
expect.objectContaining({
message: expect.stringContaining(
`${String.fromCharCode(0x1b)}[33mCould not list Google Cloud projects with gcloud`,
),
}),
);
expect(prompts.autocomplete).toHaveBeenCalledWith(
expect.objectContaining({
message: expect.stringContaining('gcloud auth login --update-adc'),
}),
);
expect(prompts.autocomplete).toHaveBeenCalledWith(
expect.objectContaining({
message: expect.stringContaining(
`${String.fromCharCode(0x1b)}[33mRun \`gcloud auth login --update-adc\``,
),
}),
);
expect(healthCheck).toHaveBeenCalledWith(
expect.objectContaining({
vertex: { project: 'other-gcp-project', location: 'us-east5' },
}),
);
});
it('returns from Vertex AI project selection Back to provider selection', async () => {
const prompts = makePromptAdapter({ selectValues: ['vertex', 'back', 'back'] });
const result = await runKtxSetupAnthropicModelStep(
{ projectDir: tempDir, inputMode: 'auto', skipLlm: false },
makeIo().io,
{
prompts,
env: {},
readGcloudProject: vi.fn(async () => 'current-gcp-project'),
listGcloudProjects: vi.fn(async () => [{ projectId: 'current-gcp-project', name: 'Current project' }]),
},
);
expect(result.status).toBe('back');
expect(prompts.select).toHaveBeenNthCalledWith(
2,
expect.objectContaining({
message: expect.stringContaining('Which LLM provider should KTX use?'),
}),
);
});
it('explains common Vertex AI Forbidden health-check causes', async () => {
const io = makeIo();
const result = await runKtxSetupAnthropicModelStep(
{
projectDir: tempDir,
inputMode: 'disabled',
llmBackend: 'vertex',
vertexProject: 'kaelio-orbit-looker-20260430',
vertexLocation: 'us-east5',
skipLlm: false,
},
io.io,
{
env: {},
healthCheck: vi.fn(async () => ({ ok: false as const, message: 'Forbidden' })),
},
);
expect(result.status).toBe('failed');
expect(io.stderr()).toContain('project kaelio-orbit-looker-20260430');
expect(io.stderr()).toContain('Vertex AI API is enabled');
expect(io.stderr()).toContain('Anthropic Claude model access');
expect(io.stderr()).toContain('roles/aiplatform.user');
});
2026-05-10 23:12:26 +02:00
it('resolves --anthropic-api-key-file for health checks and stores a file reference', async () => {
const io = makeIo();
const secretPath = join(tempDir, 'anthropic-api-key');
await writeFile(secretPath, 'sk-ant-file', 'utf-8'); // pragma: allowlist secret
2026-05-10 23:12:26 +02:00
const healthCheck = vi.fn(async () => ({ ok: true as const }));
2026-05-10 23:51:24 +02:00
const result = await runKtxSetupAnthropicModelStep(
2026-05-10 23:12:26 +02:00
{
projectDir: tempDir,
inputMode: 'disabled',
anthropicApiKeyFile: secretPath,
skipLlm: false,
},
io.io,
{ env: {}, healthCheck },
);
expect(result.status).toBe('ready');
expect(healthCheck).toHaveBeenNthCalledWith(
1,
2026-05-10 23:12:26 +02:00
expect.objectContaining({
anthropic: { apiKey: 'sk-ant-file' }, // pragma: allowlist secret
2026-05-10 23:12:26 +02:00
modelSlots: { default: 'claude-sonnet-4-6' },
}),
);
expect(healthCheck).toHaveBeenNthCalledWith(
2,
expect.objectContaining({
anthropic: { apiKey: 'sk-ant-file' }, // pragma: allowlist secret
modelSlots: { default: 'claude-haiku-4-5' },
}),
);
expect(healthCheck).toHaveBeenNthCalledWith(
3,
expect.objectContaining({
anthropic: { apiKey: 'sk-ant-file' }, // pragma: allowlist secret
modelSlots: { default: 'claude-opus-4-7' },
}),
);
2026-05-10 23:51:24 +02:00
const config = parseKtxProjectConfig(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8'));
2026-05-10 23:12:26 +02:00
expect(config.llm).toMatchObject({
provider: {
backend: 'anthropic',
anthropic: { api_key: `file:${secretPath}` }, // pragma: allowlist secret
2026-05-10 23:12:26 +02:00
},
models: anthropicPreset,
2026-05-10 23:12:26 +02:00
});
expect(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8')).not.toContain('completed_steps:');
expect((await readKtxSetupState(tempDir)).completed_steps).toContain('llm');
2026-05-10 23:12:26 +02:00
expect(io.stdout()).not.toContain('sk-ant-file');
});
it('returns missing-input when --anthropic-api-key-file points to a missing file', async () => {
const io = makeIo();
const missingSecretPath = join(tempDir, 'missing-anthropic-api-key');
const healthCheck = vi.fn(async () => ({ ok: true as const }));
2026-05-10 23:51:24 +02:00
const result = await runKtxSetupAnthropicModelStep(
2026-05-10 23:12:26 +02:00
{
projectDir: tempDir,
inputMode: 'disabled',
anthropicApiKeyFile: missingSecretPath,
skipLlm: false,
},
io.io,
{ env: {}, healthCheck },
);
expect(result.status).toBe('missing-input');
expect(healthCheck).not.toHaveBeenCalled();
expect(io.stderr()).toContain(`Missing Anthropic API key file: ${missingSecretPath}`);
});
it('requires an explicit LLM backend in non-interactive setup instead of silently defaulting to anthropic', async () => {
2026-05-10 23:12:26 +02:00
const io = makeIo();
2026-05-10 23:51:24 +02:00
const result = await runKtxSetupAnthropicModelStep(
2026-05-10 23:12:26 +02:00
{ projectDir: tempDir, inputMode: 'disabled', skipLlm: false },
io.io,
);
expect(result.status).toBe('missing-input');
const err = io.stderr();
// Names the flag the user actually needs and lists every valid backend.
expect(err).toContain('--llm-backend');
expect(err).toContain('anthropic');
expect(err).toContain('vertex');
expect(err).toContain('claude-code');
expect(err).toContain('codex');
// No silent anthropic default, so the bare api-key error is not the reason shown.
expect(err).not.toContain('Missing Anthropic API key: pass --anthropic-api-key-env');
expect(err).not.toContain('--skip-llm');
});
it('names --llm-backend when an explicit anthropic backend is missing its API key in non-interactive setup', async () => {
const io = makeIo();
const result = await runKtxSetupAnthropicModelStep(
{ projectDir: tempDir, inputMode: 'disabled', llmBackend: 'anthropic', skipLlm: false },
io.io,
2026-05-10 23:12:26 +02:00
);
expect(result.status).toBe('missing-input');
const err = io.stderr();
expect(err).toContain('--anthropic-api-key-env');
// Reveals that the backend itself is selectable, so a user who wanted a keyless
// backend (claude-code/codex) can discover it from the error.
expect(err).toContain('--llm-backend');
expect(err).not.toContain('--skip-llm');
2026-05-10 23:12:26 +02:00
});
2026-05-10 23:51:24 +02:00
it('writes pasted keys to .ktx/secrets and never prints the key', async () => {
2026-05-10 23:12:26 +02:00
const io = makeIo();
const prompts = makePromptAdapter({
credentialChoice: 'paste',
passwordValue: 'sk-ant-pasted', // pragma: allowlist secret
2026-05-10 23:12:26 +02:00
});
2026-05-10 23:51:24 +02:00
const result = await runKtxSetupAnthropicModelStep(
2026-05-10 23:12:26 +02:00
{ projectDir: tempDir, inputMode: 'auto', skipLlm: false },
io.io,
{
prompts,
env: {},
healthCheck: vi.fn(async () => ({ ok: true as const })),
},
);
expect(result.status).toBe('ready');
await expect(readFile(join(tempDir, '.ktx/secrets/anthropic-api-key'), 'utf-8')).resolves.toBe('sk-ant-pasted\n'); // pragma: allowlist secret
2026-05-10 23:12:26 +02:00
if (process.platform !== 'win32') {
2026-05-10 23:51:24 +02:00
expect((await stat(join(tempDir, '.ktx/secrets/anthropic-api-key'))).mode & 0o777).toBe(0o600);
2026-05-10 23:12:26 +02:00
}
2026-05-10 23:51:24 +02:00
const yaml = await readFile(join(tempDir, 'ktx.yaml'), 'utf-8');
2026-05-10 23:12:26 +02:00
expect(yaml).toContain('api_key: file:');
expect(yaml).not.toContain('sk-ant-pasted');
expect(io.stdout()).not.toContain('sk-ant-pasted');
});
it('opens pasted key entry directly and tells users Escape goes back', async () => {
const prompts = makePromptAdapter({
selectValues: ['paste'],
passwordValue: 'sk-ant-pasted', // pragma: allowlist secret
2026-05-10 23:12:26 +02:00
});
2026-05-10 23:51:24 +02:00
const result = await runKtxSetupAnthropicModelStep(
2026-05-10 23:12:26 +02:00
{ projectDir: tempDir, inputMode: 'auto', skipLlm: false },
makeIo().io,
{
prompts,
env: {},
healthCheck: vi.fn(async () => ({ ok: true as const })),
},
);
expect(result.status).toBe('ready');
expect(prompts.select).not.toHaveBeenCalledWith(expect.objectContaining({ message: 'Paste Anthropic API key now?' }));
expect(prompts.password).toHaveBeenCalledWith({
message: 'Anthropic API key\n│ Press Escape to go back.\n│',
2026-05-10 23:12:26 +02:00
});
});
it('does not offer skipping while choosing an Anthropic credential source', async () => {
const prompts = makePromptAdapter({ credentialChoice: 'back' });
2026-05-10 23:51:24 +02:00
const result = await runKtxSetupAnthropicModelStep(
2026-05-10 23:12:26 +02:00
{ projectDir: tempDir, inputMode: 'auto', skipLlm: false },
makeIo().io,
{ prompts, env: {} },
);
expect(result.status).toBe('back');
expect(prompts.select).toHaveBeenCalledWith(
expect.objectContaining({
2026-05-10 23:51:24 +02:00
message: expect.stringContaining('How should KTX find your Anthropic API key?'),
2026-05-10 23:12:26 +02:00
options: expect.not.arrayContaining([expect.objectContaining({ value: 'skip' })]),
}),
);
});
2026-05-10 23:51:24 +02:00
it('explains why KTX asks for an Anthropic API key', async () => {
2026-05-10 23:12:26 +02:00
const io = makeIo();
const prompts = makePromptAdapter({ credentialChoice: 'back' });
const expectedPromptMessage = [
2026-05-10 23:51:24 +02:00
'How should KTX find your Anthropic API key?',
2026-05-10 23:12:26 +02:00
'',
[
2026-05-10 23:51:24 +02:00
'KTX uses the key to verify Anthropic model access now and to run ingest agents that turn schemas, SQL,',
'BI metadata, and docs into semantic-layer sources and wiki context. ktx.yaml stores an env: or file:',
2026-05-10 23:12:26 +02:00
'reference, not the raw key.',
].join(' '),
].join('\n');
2026-05-10 23:51:24 +02:00
const result = await runKtxSetupAnthropicModelStep(
2026-05-10 23:12:26 +02:00
{ projectDir: tempDir, inputMode: 'auto', skipLlm: false },
io.io,
{ prompts, env: {} },
);
expect(result.status).toBe('back');
expect(prompts.select).toHaveBeenCalledWith(
expect.objectContaining({
message: expectedPromptMessage,
}),
);
2026-05-10 23:51:24 +02:00
expect(io.stdout()).not.toContain('KTX uses the key');
2026-05-10 23:12:26 +02:00
});
it('does not persist llm completion when the health check fails', async () => {
const io = makeIo();
2026-05-10 23:51:24 +02:00
const result = await runKtxSetupAnthropicModelStep(
2026-05-10 23:12:26 +02:00
{
projectDir: tempDir,
inputMode: 'disabled',
anthropicApiKeyEnv: 'ANTHROPIC_API_KEY', // pragma: allowlist secret
2026-05-10 23:12:26 +02:00
skipLlm: false,
},
io.io,
{
env: { ANTHROPIC_API_KEY: 'sk-ant-test' }, // pragma: allowlist secret
2026-05-10 23:12:26 +02:00
healthCheck: vi.fn(async () => ({ ok: false as const, message: '401 invalid x-api-key [redacted]' })),
},
);
expect(result.status).toBe('failed');
expect(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8')).not.toContain('completed_steps:');
2026-05-10 23:12:26 +02:00
expect(io.stderr()).toContain('Anthropic model health check failed: 401 invalid x-api-key [redacted]');
expect(io.stderr()).not.toContain('sk-ant-test');
});
it('re-prompts after an interactive health-check failure and saves after retry success', async () => {
const io = makeIo();
const prompts = makePromptAdapter({ selectValues: ['env', 'env'] });
2026-05-10 23:12:26 +02:00
const healthCheck = vi
.fn()
.mockResolvedValueOnce({ ok: false as const, message: 'model not found' })
.mockResolvedValueOnce({ ok: true as const })
.mockResolvedValueOnce({ ok: true as const })
2026-05-10 23:12:26 +02:00
.mockResolvedValueOnce({ ok: true as const });
2026-05-10 23:51:24 +02:00
const result = await runKtxSetupAnthropicModelStep(
2026-05-10 23:12:26 +02:00
{ projectDir: tempDir, inputMode: 'auto', skipLlm: false },
io.io,
{
prompts,
env: { ANTHROPIC_API_KEY: 'sk-ant-test' }, // pragma: allowlist secret
2026-05-10 23:12:26 +02:00
healthCheck,
},
);
expect(result.status).toBe('ready');
expect(healthCheck).toHaveBeenCalledTimes(4);
expect(prompts.select).toHaveBeenCalledTimes(3);
expect(prompts.autocomplete).not.toHaveBeenCalledWith(
expect.objectContaining({
message: expect.stringContaining('Which Anthropic model should KTX use?'),
}),
);
2026-05-10 23:12:26 +02:00
expect(io.stderr()).toContain('Anthropic model health check failed: model not found');
expect(io.stderr()).toContain('Choose a different credential source or Back.');
2026-05-10 23:51:24 +02:00
const config = parseKtxProjectConfig(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8'));
expect(config.llm.models).toMatchObject(anthropicPreset);
expect(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8')).not.toContain('completed_steps:');
expect((await readKtxSetupState(tempDir)).completed_steps).toContain('llm');
2026-05-10 23:12:26 +02:00
expect(io.stderr()).not.toContain('sk-ant-test');
});
it('leaves setup incomplete when skipped', async () => {
2026-05-10 23:51:24 +02:00
const result = await runKtxSetupAnthropicModelStep(
2026-05-10 23:12:26 +02:00
{ projectDir: tempDir, inputMode: 'disabled', skipLlm: true },
makeIo().io,
);
expect(result.status).toBe('skipped');
expect(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8')).not.toContain('completed_steps:');
2026-05-10 23:12:26 +02:00
});
it('returns back without writing config when Back is selected', async () => {
const prompts = makePromptAdapter({ credentialChoice: 'back' });
2026-05-10 23:51:24 +02:00
const result = await runKtxSetupAnthropicModelStep(
2026-05-10 23:12:26 +02:00
{ projectDir: tempDir, inputMode: 'auto', skipLlm: false },
makeIo().io,
{ prompts, env: {} },
);
expect(result.status).toBe('back');
2026-05-10 23:51:24 +02:00
const config = parseKtxProjectConfig(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8'));
2026-05-10 23:12:26 +02:00
expect(config.llm.provider.backend).toBe('none');
});
it('returns from pasted key entry Escape to credential selection and can use env credentials', async () => {
const prompts = makePromptAdapter({ selectValues: ['paste', 'env'], passwordValues: [undefined] });
2026-05-10 23:12:26 +02:00
2026-05-10 23:51:24 +02:00
const result = await runKtxSetupAnthropicModelStep(
2026-05-10 23:12:26 +02:00
{ projectDir: tempDir, inputMode: 'auto', skipLlm: false },
makeIo().io,
{
prompts,
env: { ANTHROPIC_API_KEY: 'sk-ant-env' }, // pragma: allowlist secret
healthCheck: vi.fn(async () => ({ ok: true as const })),
},
);
expect(result.status).toBe('ready');
expect(prompts.password).toHaveBeenCalledWith({
message: 'Anthropic API key\n│ Press Escape to go back.\n│',
2026-05-10 23:12:26 +02:00
});
2026-05-10 23:51:24 +02:00
await expect(readFile(join(tempDir, '.ktx/secrets/anthropic-api-key'), 'utf-8')).rejects.toMatchObject({
2026-05-10 23:12:26 +02:00
code: 'ENOENT',
});
2026-05-10 23:51:24 +02:00
const config = parseKtxProjectConfig(await readFile(join(tempDir, 'ktx.yaml'), 'utf-8'));
2026-05-10 23:12:26 +02:00
expect(config.llm.provider).toMatchObject({
backend: 'anthropic',
anthropic: { api_key: 'env:ANTHROPIC_API_KEY' }, // pragma: allowlist secret
2026-05-10 23:12:26 +02:00
});
});
it('preserves already completed llm setup when no model args request changes', async () => {
2026-05-10 23:51:24 +02:00
await mkdir(join(tempDir, '.ktx'), { recursive: true });
await initKtxProject({ projectDir: tempDir, force: true });
2026-05-10 23:12:26 +02:00
await writeFile(
2026-05-10 23:51:24 +02:00
join(tempDir, 'ktx.yaml'),
2026-05-10 23:12:26 +02:00
[
'setup:',
' database_connection_ids: []',
'connections: {}',
'llm:',
' provider:',
' backend: anthropic',
' anthropic:',
' api_key: env:ANTHROPIC_API_KEY', // pragma: allowlist secret
2026-05-10 23:12:26 +02:00
' models:',
' default: claude-sonnet-4-6',
'ingest:',
' embeddings:',
' backend: none',
2026-05-10 23:12:26 +02:00
' dimensions: 8',
].join('\n'),
'utf-8',
);
await writeKtxSetupState(tempDir, { completed_steps: ['project', 'llm'] });
2026-05-10 23:12:26 +02:00
const healthCheck = vi.fn(async () => ({ ok: true as const }));
await expect(
2026-05-10 23:51:24 +02:00
runKtxSetupAnthropicModelStep({ projectDir: tempDir, inputMode: 'disabled', skipLlm: false }, makeIo().io, {
env: { ANTHROPIC_API_KEY: 'sk-ant-test' }, // pragma: allowlist secret
2026-05-10 23:12:26 +02:00
healthCheck,
}),
).resolves.toMatchObject({ status: 'ready' });
expect(healthCheck).not.toHaveBeenCalled();
});
it.each([
{
backend: 'vertex',
providerLines: [' backend: vertex', ' vertex:', ' project: kaelio-dev', ' location: us-east5'],
model: 'claude-sonnet-4-6',
},
{
backend: 'gateway',
providerLines: [' backend: gateway', ' gateway:', ' api_key: env:AI_GATEWAY_API_KEY'],
model: 'anthropic/claude-sonnet-4-6',
},
])('preserves already configured $backend llm setup without asking for Anthropic credentials', async (fixture) => {
await writeFile(
join(tempDir, 'ktx.yaml'),
[
'setup:',
' database_connection_ids: []',
'connections: {}',
'llm:',
' provider:',
...fixture.providerLines,
' models:',
` default: ${fixture.model}`,
'ingest:',
' embeddings:',
' backend: none',
' dimensions: 8',
].join('\n'),
'utf-8',
);
await writeKtxSetupState(tempDir, { completed_steps: ['project', 'llm'] });
const healthCheck = vi.fn(async () => ({ ok: true as const }));
const io = makeIo();
await expect(
runKtxSetupAnthropicModelStep({ projectDir: tempDir, inputMode: 'disabled', skipLlm: false }, io.io, {
healthCheck,
}),
).resolves.toMatchObject({ status: 'ready' });
expect(healthCheck).not.toHaveBeenCalled();
expect(io.stdout()).toContain(`LLM ready: yes (${fixture.model})`);
expect(io.stderr()).not.toContain('Anthropic');
});
2026-05-10 23:12:26 +02:00
});