feat: Prometheus — CommandCode provider, api_mode=responses, goal migration, Honcho prune, profile routing

[ousiaresearch]

What does this PR do?

Adds six features (previously blocked by GitHub push issues) rebased onto current upstream HEAD:

• CommandCode provider plugin — plugins/model-providers/commandcode/ with Anthropic shim + test coverage
• api_mode="responses" (provider: custom always appends /chat/completions to base_url, ignoring api_mode #33600) — generic OpenAI Responses API dispatch for custom providers using /v1/responses
• _normalize_responses_to_chat() — normalizes Responses API output to chat.completions shape
• Nous Portal usage/credits tracking (Feature Request: Check Nous Portal credit balance and usage via CLI #33376) — fetches credit snapshots from Nous inference API
• migrate_goal() (Persistent /goal is lost after context compression rotates session_id #33618) — migrates active goals across session_id rotation (3 sites in gateway)
• Fallback provider notification ([Feature] Notify user when fallback provider is activated (gateway/messaging mode) #33174) — opt-in user-visible notification on fallback activation
• Per-user profile routing (Feature: Per-user profile routing for messaging platforms (user->profile binding) #33548) — profile_routing config maps user IDs to Hermes profiles
• Honcho session pruning (Feature Request: Add observation pruning controls to Honcho memory provider #33436) — hermes honcho prune CLI with --dry-run, --force, --include-active

Related Issue

Fixes #33600, #33618, #33376, #33174, #33548, #33436

Type of Change

• ✨ New feature (non-breaking change that adds functionality)
• ✅ Tests (adding or improving test coverage)

Changes Made

• New files:
  • plugins/model-providers/commandcode/ — provider plugin (4 files)
  • tests/providers/test_commandcode_profile.py — test coverage
• Modified files:
  • run_agent.py — added responses api_mode branch in _call(), import _normalize_responses_to_chat
  • agent/codex_responses_adapter.py — added _normalize_responses_to_chat() for generic Responses API
  • agent/account_usage.py — added Nous credit fetch + connectivity fallback
  • hermes_cli/runtime_provider.py — added "responses" to _VALID_API_MODES
  • hermes_cli/goals.py — added migrate_goal() function
  • gateway/run.py — fallback notification callbacks, migrate_goal() at 3 session rotation sites, _load_profile_routing(), _resolve_profile_for_user(), _set_fallback_status_callback(), _gateway_extra_flag(), _notify_fallback_used()
  • gateway/platforms/base.py — added target_profile field to MessageEvent
  • plugins/memory/honcho/cli.py — added cmd_prune() with --dry-run, --force, --include-active
  • plugins/memory/honcho/client.py — added prune_enabled and prune_max_age_days config fields
  • tests/providers/test_plugin_discovery.py — added "commandcode" to required providers

How to Test

1. CommandCode plugin: pytest tests/providers/test_commandcode_profile.py tests/providers/test_plugin_discovery.py
2. api_mode=responses: Configure a custom provider with api_mode: responses in config.yaml, send a message — verify tool calls and text work
3. Goal migration: Set a /goal, let session compress — verify goal survives session_id rotation
4. Fallback notification: Enable fallback_notifications in gateway.extra, force primary provider to fail — verify notification
5. Profile routing: Add profile_routing: {"<user_id>": "<profile>"} to config.yaml — verify message routes to correct profile
6. Honcho pruning: hermes honcho prune --dry-run — verify stale session list, then hermes honcho prune --force

Checklist

• I have tested these changes locally
• I have written or updated tests
• My changes follow the project code style
• I have read the CONTRIBUTING guidelines

[alt-glitch]

Mega PR bundling 8 features, each also submitted as individual PRs: #33886 (CommandCode), #33889 (api_mode=responses), #33888 (Nous usage), #33890 (goal migration), #33891 (fallback notification), #33892 (profile routing), #33893 (Honcho prune). CommandCode portion duplicates existing #32909. Goal migration competes with #33620, #29955, #18749, #18427. Profile routing competes with #33558. api_mode=responses competes with #33604, #33637, #33765.