Checkpoint: development → main (PR #96 correction rounds, Gate 0 ADR drafts, MVP-1)

.github/workflows/ci.yml

@@ -4,8 +4,8 @@
 # Turborepo cache, plus the formatting check and the no-vendor-type-across-the-seam fence.
 #
 # Branch protection (set in the GitHub repo settings, not here): `ci`, `coverage` and `windows-concurrency`
-# are the REQUIRED status checks to merge into `main` (the Windows job since 2026-11-04: ADR-0106's
-# acceptance names required CI for the platform-scoped claims it carries). Wire, usage, filesystem, secrets,
+# are the REQUIRED status checks to merge into `main` (the Windows job since the maintainer's 2026-11-04
+# ruling: it is the CI lane that validates ADR-0106's Windows-scoped claims). Wire, usage, filesystem, secrets,
 # packed CLI and product map control jobs feed `ci` through an explicit success-only aggregation; they cannot
 # silently skip or fail while `ci` stays green. The other jobs self-label `(advisory)` — `floor-check` and
 # `peer-dep-gate`.
@@ -226,7 +226,7 @@ jobs:
       # RUN the artifact this job just built, through the SAME `pnpm smoke:cli` script the root `ci` script
       # calls — a check that exists in only one of the two is exactly the #312 divergence this change closes.
       # Until now nothing in the required gate executed
-      # `apps/cli/dist/index.js` — only the advisory Windows leg and the tag-gated release smoke did (#294) —
+      # `apps/cli/dist/index.js` — only the then-advisory Windows leg and the tag-gated release smoke did (#294) —
       # so a bundle that builds but cannot boot merged green. The `run --json` leg additionally proves the
       # migrations resolve beside the bundle, which is the failure `apps/cli/drizzle/**` becoming a declared
       # turbo output (#315) exists to prevent: a cache-hit replay used to leave `dist/` fresh next to a
@@ -424,7 +424,8 @@ jobs:
   # Node-22 prebuild (ABI 127 — one binary for ALL of 22.x, so 22.13.0 loads the same artifact 22.0.0 would;
   # no node-gyp C++ source build). What it does NOT prove is that no RUNTIME dep requires >22.0.0 — that is
   # a property of the prod closure, verified at the ADR (see ADR-0067's amendment note) rather than here.
-  # A SEPARATE job (the required check stays the ubuntu `ci` job on Node 24); promote once confirmed stable.
+  # A SEPARATE, advisory job (the required checks are `ci`, `coverage` and `windows-concurrency`; see the header);
+  # promote once confirmed stable.
   floor-check:
     name: node 22-line floor (22.13.0) · typecheck · test · build (advisory)
     runs-on: ubuntu-latest
@@ -537,8 +538,8 @@ jobs:
   # Runs the @relavium/db concurrency suite (the real Windows native-addon + WAL exercise) + the CLI
   # concurrency/perf harness there, the native Step 2 filesystem boundaries (ADR-0105), plus a headless
   # no-TTY smoke. Native links/containment complement portable Windows lexical vectors; these results are
-  # evidence only after this job actually runs. A SEPARATE, advisory job (the required check
-  # stays the ubuntu `ci` job) — promote to required in branch protection once confirmed stable. POSIX 0600/0700
+  # evidence only after this job actually runs. A SEPARATE job and, since 2026-11-04, a REQUIRED check in
+  # branch protection beside `ci` and `coverage` (see the header). POSIX 0600/0700
   # perm assertions are NOT exercised here (a documented Windows no-op — ADR-0050); nor is ink's raw-mode code,
   # which is TTY-gated (the driver-selection gate picks the plain driver without a TTY — see the smoke below).
   windows-concurrency:
@@ -603,8 +604,10 @@ jobs:
           # `chat` is TTY-gated on STDOUT (selectChatDriver / io.stdoutIsTty): with stdout captured (non-TTY)
           # here, the driver-selection gate picks the PLAIN driver (ink/setRawMode is never entered), and the
           # plain driver must exit cleanly on EOF — never crash. A regression routing to ink without a TTY would
-          # throw "Raw mode is not supported…", which this catches.
-          if out="$(printf '' | node apps/cli/dist/index.js chat 2>&1)"; then
+          # throw "Raw mode is not supported…", which this catches. The runner stores no provider key, and since
+          # MVP-A4 a keyless `chat` refuses with exit 2 before choosing a driver, so the chat gets a fake env key:
+          # an empty stdin sends nothing, so the key is never sent.
+          if out="$(printf '' | RELAVIUM_ANTHROPIC_API_KEY=relavium-headless-smoke-not-a-key node apps/cli/dist/index.js chat 2>&1)"; then
             chat_status=0
           else
             chat_status=$?

.github/workflows/release.yml

@@ -14,7 +14,8 @@
 # release ADR that publishes the smoke-verified package-test tarball, with a sunset at 1.0 or when Steps
 # 9c/9d land. That ADR must explicitly override ADR-0111's "missing custody cannot downgrade to test mode".
 # Until it is Accepted and its guard change passes a security sitting, the guard below stays as it is.
-# `publish` runs in the `release` deployment environment, so the npm token can be scoped to it.
+# `publish` names the `release` deployment environment, so the npm token can be scoped to it. The environment and
+# that scoping are MVP-A23's prerequisite (not done on 2026-11-04: the token is still a repository secret).
 #
 # Flow: `pack` builds the engine-inlined bundle and `pnpm pack`s it (resolving catalog:/workspace: to
 # concrete versions) into a tarball artifact → `smoke` installs THAT SAME tarball outside the workspace on ubuntu /
@@ -30,7 +31,9 @@
 # The `allow_off_main` dispatch input is the maintainer escape hatch for building and smoking an unmerged
 # BRANCH. It cannot reach a publish: `workflow_dispatch` accepts a tag ref as readily as a branch, so the
 # ancestry assertion ignores the override whenever the ref is a `v*` tag, and `publish` additionally requires
-# `github.event_name == 'push'` — a dispatched run never publishes, whatever ref or input it was given.
+# `github.event_name == 'push'` — a dispatched run of this workflow never publishes, whatever ref or input it
+# was given. A dispatch runs the workflow file found at its ref, so an edited copy on another ref is bounded only by
+# where `NPM_TOKEN` lives (MVP-A23).
 #
 # Third-party actions are pinned to a commit SHA (the `# vX.Y.Z` comment tracks the release), matching ci.yml.
 name: Release CLI
@@ -227,8 +230,9 @@ jobs:
     # manually dispatched run publish. Every manual run now stops after the cross-OS smoke.
     if: ${{ github.event_name == 'push' && startsWith(github.ref, 'refs/tags/v') }}
     runs-on: ubuntu-latest
-    # The deployment environment that holds `NPM_TOKEN` and its required reviewer (repo settings); a
-    # repository-level secret is readable by any workflow a writer pushes. It sets no job environment.
+    # The deployment environment meant to hold `NPM_TOKEN` and its required reviewer (repo settings, MVP-A23); a
+    # repository-level secret is readable by any workflow a writer pushes. It does not exist yet: GitHub creates it
+    # unprotected on the first run that names it, so it gates nothing until MVP-A23 sets it up.
     environment: release
     timeout-minutes: 10
     permissions:

AGENTS.md

@@ -24,7 +24,8 @@ gateway; engine stays local, Phase 2) — split across build phase 5 (managed in
 phase 6 (cloud execution + portal); the engine is identical across all three (ADR-0012..0015).
 **Status (2026-11-04): the CLI MVP comes first.** Phases 1, 2 and 2.5 are complete (`relavium@0.1.1` is
 published); Phase 2.6.5 is complete with approved deferrals (49/51, PR #92). Phase 2.5.5's Waves 0–2 and Wave 3
-Steps 1–8 are merged (PR #94, PR #95), and Steps 9a.1–9b.2 and NI are accepted on `development`. On 2026-10-26
+Steps 1–8 are merged (PR #94, PR #95), and Steps 9a.1–9b.2 and NI are accepted and on `main` since the
+`development`→`main` checkpoint (PR #96, 2026-11-04). On 2026-10-26
 the maintainer paused Wave 3 (Steps 9c–19 pending, not declined) and put the CLI MVP first: `relavium` v0.2.0
 (milestone M3.5), ordered as MVP-1…MVP-10 under decisions D58–D72 in
 [current.md](docs/roadmap/current.md#next-work--the-cli-mvp). Phase 2.6 is in progress (2.6.C and 2.6.F done,

CLAUDE.md

@@ -44,13 +44,13 @@ and [docs/architecture/managed-inference.md](docs/architecture/managed-inference
 complete (M2.5-4, PR #69, 2026-07-08); Phase 2.6.5 (Core reliability) is complete with approved deferrals
 (49/51, PR #92, 2026-10-11). Phase 2.5.5 (Hardening and Remediation): Waves 0–2 are merged (Wave 2: PR #94),
 Wave 3 Steps 1–8 are merged (PR #95, 2026-10-18), and PR95, 9a.1, 9a.2a, 9a.2b, NI, 9b.1 and 9b.2 are
-accepted on `development`. **On 2026-10-26 the maintainer paused Wave 3 — Steps 9c–19 are pending, not
+accepted and on `main` since the `development`→`main` checkpoint (PR #96, merged 2026-11-04). **On 2026-10-26 the maintainer paused Wave 3 — Steps 9c–19 are pending, not
 declined — and put the CLI MVP first:** `relavium` v0.2.0 on npm (milestone M3.5), ordered as MVP-1…MVP-10 in
 [current.md](docs/roadmap/current.md#next-work--the-cli-mvp) under decisions D58–D72 (ruled 2026-11-04).
 Phase 2.6 (Conversational Authoring and the First-Class CLI) is in progress: 2.6.C and 2.6.F are done
 (2.6.F's v0.2.0 release clause closes with the MVP) and 2.6.Q P1–P5 landed; the MVP takes slices of the rest.
 
-Two rules shape every MVP increment. Step 9a.1's interim release guard on `development` refuses any npm
+Two rules shape every MVP increment. Step 9a.1's interim release guard (on `main` since PR #96) refuses any npm
 publication until the superseding pre-1.0 release ADR (D59) is Accepted. The acceptance bar is tiered (D64):
 an increment touching an approval or consent path, egress, keys or secrets, filesystem writes, a new terminal
 sink for untrusted content, the lease or resume paths, or release keeps a rigor pass, one NEW complete

README.md

@@ -158,11 +158,12 @@ Relavium is under active development. The **CLI is published as v0.1.1**; the pu
 agent-session entry point, workflow runtime, inbound MCP client, and CLI management surface
 are implemented. **Phase 2.6.5 — Core Reliability Remediation is complete with approved
 deferrals**, with W8 merged in PR #92. The joint Phase 2.5.5/2.6 security remediation
-Wave 2 merged to `main` on 2026-10-15 as PR #94, and the first eight steps of Wave 3
-(“Clear the ground”) followed on 2026-10-18 as PR #95. Since 2026-10-26 the **CLI MVP
-comes first**: Wave 3 is paused, not dropped, while the work a CLI user feels — visible
-run results, a chat agent that can edit and run commands, readable replies, conversational
-authoring, web, the Home hub and MCP management — leads to `relavium` v0.2.0.
+Wave 2 merged to `main` on 2026-10-15 as PR #94. The first eight steps of Wave 3
+(“Clear the ground”) followed on 2026-10-18 as PR #95, and the hardened release tooling and
+the MVP plan on 2026-11-04 as PR #96. Since 2026-10-26 the **CLI MVP comes first**: Wave 3
+is paused, not dropped, while the work a CLI user feels — visible run results, a chat agent
+that can edit and run commands, readable replies, conversational authoring, web, the Home
+hub and MCP management — leads to `relavium` v0.2.0.
 
 Status changes quickly, so this README intentionally stays high-level. The canonical source
 for the exact active wave, completed work, and open reliability obligations is
@@ -178,8 +179,8 @@ for the exact active wave, completed work, and open reliability obligations is
 | [`packages/db`](packages/db/) | Local SQLite persistence with a Postgres-compatible schema and migrations |
 | [`packages/mcp`](packages/mcp/) | SDK-confined inbound MCP client and schema validation |
 | [`apps/cli`](apps/cli/) | Published terminal product and integration harness |
-| [`apps/desktop`](apps/desktop/) | Tauri desktop surface under development |
-| [`apps/vscode-extension`](apps/vscode-extension/) | VS Code surface under development |
+| [`apps/desktop`](apps/desktop/) | Tauri desktop surface — a placeholder, built in Phase 3 |
+| [`apps/vscode-extension`](apps/vscode-extension/) | VS Code surface — a placeholder, built in Phase 4 |
 
 The full dependency graph and ownership rules live in
 [docs/project-structure.md](docs/project-structure.md).

apps/cli/src/authoring/create-prompter.ts

@@ -126,7 +126,8 @@ export function createClackPrompter(deps: ClackCreateDeps = defaultDeps): Create
 
       const toolsRaw = await deps.text({
         message: 'Tools (comma-separated tool ids, optional)',
-        placeholder: 'read_file, web_search',
+        // Tools that also work on `relavium run`, which wires no egress or os arm (2.6.A).
+        placeholder: 'read_file, list_directory',
       });
       if (deps.isCancel(toolsRaw)) return null;
       const tools = toolsRaw

apps/cli/src/chat/agent-source.test.ts

@@ -5,6 +5,7 @@ import { join } from 'node:path';
 import { AgentParseError, MAX_SOURCE_CHARS } from '@relavium/core';
 import { afterEach, beforeEach, describe, expect, it } from 'vitest';
 
+import { KNOWN_PROVIDERS } from '../engine/providers.js';
 import { isCliError } from '../process/errors.js';
 import { DEFAULT_CHAT_AGENT_ID, DEFAULT_CHAT_MODEL } from './default-agent.js';
 import { resolveChatAgent } from './agent-source.js';
@@ -72,6 +73,52 @@ describe('resolveChatAgent', () => {
     expect(agent.model).toBe(DEFAULT_CHAT_MODEL);
   });
 
+  it('MVP-A4: with no configured default, the model comes from the first provider that has a key', () => {
+    const resolve = (keyed: readonly string[]) =>
+      resolveChatAgent(undefined, {
+        cwd: dir,
+        projectConfigDir: undefined,
+        defaultModel: undefined,
+        hasKey: (id) => keyed.includes(id),
+      });
+    // Anthropic keeps the long-standing default; every other provider gets the wizard's starter model.
+    expect(resolve(['anthropic', 'openai'])).toMatchObject({
+      model: DEFAULT_CHAT_MODEL,
+      provider: 'anthropic',
+    });
+    for (const id of ['openai', 'gemini', 'deepseek'] as const) {
+      expect(resolve([id])).toMatchObject({ model: KNOWN_PROVIDERS[id].testModel, provider: id });
+    }
+    // In KNOWN_PROVIDER_IDS order when several are keyed.
+    expect(resolve(['deepseek', 'gemini'])).toMatchObject({ provider: 'gemini' });
+    // No key anywhere: the built-in default, so the session pre-flight names the remedy for it.
+    expect(resolve([])).toMatchObject({ model: DEFAULT_CHAT_MODEL, provider: 'anthropic' });
+  });
+
+  it('MVP-A4: a configured model, or a provider a caller supplies, wins over the keys', () => {
+    const keyedOpenAi = (id: string) => id === 'openai';
+    expect(
+      resolveChatAgent(undefined, {
+        cwd: dir,
+        projectConfigDir: undefined,
+        defaultModel: 'claude-haiku-4-5',
+        defaultProvider: 'anthropic',
+        hasKey: keyedOpenAi,
+      }),
+    ).toMatchObject({ model: 'claude-haiku-4-5', provider: 'anthropic' });
+    // A provider passed without a model gets that provider's starter, never the Anthropic default. Only a direct
+    // caller reaches this: config resolution drops a persisted provider that has no model.
+    expect(
+      resolveChatAgent(undefined, {
+        cwd: dir,
+        projectConfigDir: undefined,
+        defaultModel: undefined,
+        defaultProvider: 'gemini',
+        hasKey: keyedOpenAi,
+      }),
+    ).toMatchObject({ model: KNOWN_PROVIDERS.gemini.testModel, provider: 'gemini' });
+  });
+
   it('bakes opts.reasoningEffort onto the DEFAULT agent (ADR-0066)', () => {
     const agent = resolveChatAgent(undefined, {
       cwd: dir,

apps/cli/src/chat/agent-source.ts

@@ -1,6 +1,7 @@
 import { parseAgent, type AgentDefinition } from '@relavium/core';
 import type { LlmProviderId, ReasoningEffort } from '@relavium/shared';
 
+import { KNOWN_PROVIDER_IDS, KNOWN_PROVIDERS } from '../engine/providers.js';
 import { sourceLabel } from '../process/source-label.js';
 import { resolveYamlSource } from '../workflows/resolve.js';
 import { buildDefaultChatAgent, DEFAULT_CHAT_MODEL } from './default-agent.js';
@@ -16,6 +17,40 @@ export interface ResolveChatAgentOptions {
   /** The resolved `[chat].reasoning_effort` (ADR-0066) — baked onto the DEFAULT agent only (an authored `--agent`
    *  owns its own `reasoning_effort`, never overridden by config). Absent ⇒ no reasoning control. */
   readonly reasoningEffort?: ReasoningEffort;
+  /**
+   * Whether a provider has a resolvable key — a boolean, never the key (MVP-A4). Consulted only for the DEFAULT
+   * agent when neither `defaultModel` nor `defaultProvider` is configured, so the zero-config chat binds a model a
+   * keyed provider serves. Absent ⇒ the built-in default, as before.
+   */
+  readonly hasKey?: (id: LlmProviderId) => boolean;
+}
+
+/**
+ * The default agent's model and provider (MVP-A4).
+ *
+ * A configured `defaultModel` wins, with its persisted provider (ADR-0059). Otherwise the provider is a
+ * `defaultProvider` the caller passed, else the first provider in {@link KNOWN_PROVIDER_IDS} order that has a key,
+ * else Anthropic. Config resolution never passes a provider without a model (it drops a stray one), so from config
+ * the choice is always the key-aware one. Anthropic keeps the long-standing zero-config {@link DEFAULT_CHAT_MODEL}; any other provider gets the
+ * onboarding wizard's starter model (`KNOWN_PROVIDERS[id].testModel`), so a user with only an OpenAI key reaches a
+ * working chat instead of an Anthropic credential failure.
+ */
+export function defaultChatModel(opts: ResolveChatAgentOptions): {
+  readonly model: string;
+  readonly provider: LlmProviderId | undefined;
+} {
+  if (opts.defaultModel !== undefined) {
+    return { model: opts.defaultModel, provider: opts.defaultProvider };
+  }
+  const hasKey = opts.hasKey;
+  const provider =
+    opts.defaultProvider ??
+    (hasKey === undefined ? undefined : KNOWN_PROVIDER_IDS.find((id) => hasKey(id))) ??
+    'anthropic';
+  return {
+    model: provider === 'anthropic' ? DEFAULT_CHAT_MODEL : KNOWN_PROVIDERS[provider].testModel,
+    provider,
+  };
 }
 
 /**
@@ -56,12 +91,9 @@ export function resolveChatAgentSource(
   opts: ResolveChatAgentOptions,
 ): ResolvedChatAgent {
   if (agentRef === undefined) {
+    const { model, provider } = defaultChatModel(opts);
     return {
-      agent: buildDefaultChatAgent(
-        opts.defaultModel ?? DEFAULT_CHAT_MODEL,
-        opts.reasoningEffort,
-        opts.defaultProvider,
-      ),
+      agent: buildDefaultChatAgent(model, opts.reasoningEffort, provider),
       artifact: undefined,
     };
   }

apps/cli/src/chat/compaction-persistence.test.ts

@@ -18,6 +18,7 @@ const chat: ResolvedChatConfig = {
   autoCompact: true,
   compactThreshold: 0.7,
   maxCostMicrocents: undefined,
+  maxTokens: undefined,
   onExceed: undefined,
   strictCostCap: false,
   allowedCommands: undefined,