Phase 003: remove the coding-agent and ekko-agent runtimes

Every chat session now runs through the Hermes bridge, so the three chat
chains become feature-equivalent instead of three partially-maintained
dispatch paths. Keeping the coding-agent and in-house agent runtimes alive
meant every bridge change had to be mirrored into two more runtimes that
had already drifted.

Historical rows are read, not migrated: sessions stored with the retired
source normalize to cli on load, workflow graphs naming the retired agents
keep loading and run on the bridge, and the visible-source check plus
SOURCE_LABELS still recognize the old value so those sessions stay listed.

The public surface follows the code: the OpenAPI generator and the MCP
chat-run tool schema drop the six retired entry fields so external callers
are not told about runtimes that no longer exist.

harness-check gains a RETIRED_RUNTIME scan so the residual check is a
mechanical gate instead of a manual grep, with the two surviving read-side
compat sites recorded with reasons.

Refs: docs/phases/phase_003.md

tests/client/chat-store-reasoning-tool-boundary.test.ts

@@ -294,12 +294,9 @@ describe('chat store reasoning/tool boundaries', () => {
     ])
   })
 
-  it('settles running coding-agent tools when the run completes without a tool.completed event', async () => {
+  it('settles running tools when the run completes without a tool.completed event', async () => {
     const store = useChatStore()
     const session = makeSession()
-    session.source = 'coding_agent'
-    session.agent = 'claude'
-    session.codingAgentId = 'claude-code'
     store.sessions = [session]
     store.activeSessionId = 'session-1'
     store.activeSession = session
@@ -351,49 +348,6 @@ describe('chat store reasoning/tool boundaries', () => {
     ].join('\n'))
   })
 
-  it('queues active coding agent follow-up messages without command styling', async () => {
-    const store = useChatStore()
-    const session = makeSession()
-    session.source = 'coding_agent'
-    session.agent = 'claude'
-    session.codingAgentId = 'claude-code'
-    store.sessions = [session]
-    store.activeSessionId = 'session-1'
-    store.activeSession = session
-
-    await store.sendMessage('first input')
-    const onEvent = chatApi.startRunViaSocket.mock.calls[0][1] as (event: RunEvent) => void
-    onEvent({
-      event: 'agent.event',
-      session_id: 'session-1',
-      source: 'coding_agent',
-      kind: 'status',
-      text: 'Input sent to coding agent.',
-    })
-    expect(store.messages).toHaveLength(1)
-
-    onEvent({ event: 'run.started', session_id: 'session-1' })
-
-    await store.sendMessage('/not-a-hermes-command')
-
-    expect(chatApi.startRunViaSocket).toHaveBeenCalledTimes(2)
-    expect(store.queuedUserMessages.get('session-1')).toEqual([
-      expect.objectContaining({
-        role: 'user',
-        content: '/not-a-hermes-command',
-        queued: true,
-        systemType: undefined,
-      }),
-    ])
-    expect(store.messages).toEqual([
-      expect.objectContaining({
-        role: 'user',
-        content: 'first input',
-        queued: false,
-      }),
-    ])
-  })
-
   it('queues unknown slash commands in bridge sessions as normal user input', async () => {
     const store = useChatStore()
     const session = makeSession()
@@ -484,111 +438,14 @@ describe('chat store reasoning/tool boundaries', () => {
     ])
   })
 
-  it('starts global coding-agent runs without provider credentials', async () => {
+  it('sends the selected workspace and reasoning effort when starting a bridge run', async () => {
     const store = useChatStore()
     const session = makeSession()
-    session.source = 'coding_agent'
-    session.agent = 'codex'
-    session.codingAgentId = 'codex'
-    session.codingAgentMode = 'global'
-    session.provider = 'should-not-send'
-    session.model = 'should-not-send'
-    session.baseUrl = 'http://example.invalid'
-    session.apiKey = 'secret'
-    session.apiMode = 'chat_completions'
-    session.reasoningEffort = 'high'
-    store.sessions = [session]
-    store.activeSessionId = 'session-1'
-    store.activeSession = session
-
-    await store.sendMessage('use global codex auth')
-
-    const body = chatApi.startRunViaSocket.mock.calls[0][0]
-    expect(body).toEqual(expect.objectContaining({
-      source: 'coding_agent',
-      coding_agent_id: 'codex',
-      mode: 'global',
-    }))
-    expect(body.provider).toBeUndefined()
-    expect(body.model).toBeUndefined()
-    expect(body.baseUrl).toBeUndefined()
-    expect(body.apiKey).toBeUndefined()
-    expect(body.apiMode).toBeUndefined()
-    expect(body.reasoning_effort).toBeUndefined()
-  })
-
-  it('clears stale coding-agent runtime credentials when switching providers', async () => {
-    const store = useChatStore()
-    const session = makeSession()
-    session.source = 'coding_agent'
-    session.agent = 'codex'
-    session.codingAgentId = 'codex'
-    session.codingAgentMode = 'scoped'
-    session.provider = 'xiaomi'
-    session.model = 'mimo-v2.5-pro'
-    session.baseUrl = 'https://api.xiaomimimo.com/v1'
-    session.apiKey = 'sk-xiaomi'
-    session.apiMode = 'chat_completions'
-    store.sessions = [session]
-    store.activeSessionId = 'session-1'
-    store.activeSession = session
-
-    const ok = await store.switchSessionModel('deepseek-v4-pro', 'deepseek', 'session-1', 'chat_completions')
-
-    expect(ok).toBe(true)
-    expect(sessionsApi.setSessionModel).toHaveBeenCalledWith(
-      'session-1',
-      'deepseek-v4-pro',
-      'deepseek',
-      'chat_completions',
-    )
-    expect(session.provider).toBe('deepseek')
-    expect(session.model).toBe('deepseek-v4-pro')
-    expect(session.baseUrl).toBeUndefined()
-    expect(session.apiKey).toBeUndefined()
-    expect(session.apiMode).toBe('chat_completions')
-  })
-
-
-  it('preserves a scoped coding-agent API mode when reselecting the same provider', async () => {
-    const store = useChatStore()
-    const session = makeSession()
-    session.source = 'coding_agent'
-    session.agent = 'codex'
-    session.codingAgentId = 'codex'
-    session.codingAgentMode = 'scoped'
-    session.provider = 'fun-codex'
-    session.model = 'gpt-5.4'
-    session.apiMode = 'chat_completions'
-    store.sessions = [session]
-    store.activeSessionId = 'session-1'
-    store.activeSession = session
-
-    const ok = await store.switchSessionModel('gpt-5.5', 'fun-codex', 'session-1')
-
-    expect(ok).toBe(true)
-    expect(sessionsApi.setSessionModel).toHaveBeenCalledWith(
-      'session-1',
-      'gpt-5.5',
-      'fun-codex',
-      'chat_completions',
-    )
-    expect(session.apiMode).toBe('chat_completions')
-  })
-
-  it('sends the selected workspace when starting a coding-agent run', async () => {
-    const store = useChatStore()
-    const session = makeSession()
-    session.source = 'coding_agent'
-    session.agent = 'claude'
-    session.codingAgentId = 'claude-code'
-    session.codingAgentMode = 'scoped'
+    session.source = 'cli'
     session.provider = 'openrouter'
     session.model = 'anthropic/claude-sonnet-4.6'
-    session.baseUrl = 'https://openrouter.ai/api/v1'
-    session.apiKey = 'sk-test'
-    session.apiMode = 'anthropic_messages'
     session.workspace = '/workspace/project-a'
+    session.reasoningEffort = 'high'
     store.sessions = [session]
     store.activeSessionId = 'session-1'
     store.activeSession = session
@@ -596,21 +453,62 @@ describe('chat store reasoning/tool boundaries', () => {
     await store.sendMessage('use this project')
 
     expect(chatApi.startRunViaSocket.mock.calls[0][0]).toEqual(expect.objectContaining({
-      source: 'coding_agent',
-      coding_agent_id: 'claude-code',
-      mode: 'scoped',
+      source: 'cli',
       workspace: '/workspace/project-a',
+      reasoning_effort: 'high',
     }))
   })
 
-  it('keeps a local stream controller for active coding-agent runs', async () => {
+  it('sends only the bridge run fields on the run payload', async () => {
+    const store = useChatStore()
+    const session = makeSession()
+    store.sessions = [session]
+    store.activeSessionId = 'session-1'
+    store.activeSession = session
+
+    await store.sendMessage('hello')
+
+    expect(Object.keys(chatApi.startRunViaSocket.mock.calls[0][0] as Record<string, unknown>).sort())
+      .toEqual([
+        'input',
+        'model',
+        'model_groups',
+        'profile',
+        'provider',
+        'queue_id',
+        'reasoning_effort',
+        'session_id',
+        'source',
+        'workspace',
+      ])
+  })
+
+  it('switches the session model without touching runtime credentials', async () => {
+    const store = useChatStore()
+    const session = makeSession()
+    session.provider = 'xiaomi'
+    session.model = 'mimo-v2.5-pro'
+    store.sessions = [session]
+    store.activeSessionId = 'session-1'
+    store.activeSession = session
+
+    const ok = await store.switchSessionModel('deepseek-v4-pro', 'deepseek', 'session-1')
+
+    expect(ok).toBe(true)
+    expect(sessionsApi.setSessionModel).toHaveBeenCalledWith(
+      'session-1',
+      'deepseek-v4-pro',
+      'deepseek',
+    )
+    expect(session.provider).toBe('deepseek')
+    expect(session.model).toBe('deepseek-v4-pro')
+  })
+
+  it('keeps a local stream controller for active runs', async () => {
     const abort = vi.fn()
     chatApi.startRunViaSocket.mockReturnValue({ abort })
     const store = useChatStore()
     const session = makeSession()
-    session.source = 'coding_agent'
-    session.agent = 'claude'
-    session.codingAgentId = 'claude-code'
     store.sessions = [session]
     store.activeSessionId = 'session-1'
     store.activeSession = session