Files
agent-estate-wiki/concepts/model-routing.md
T
Hermes 341dda95dd stocktake 2026-11-11: 36 stale pages re-verified (0 stale); fallback chain rewritten 4/6 slots; CT333 'clawtest' naming claim corrected; nous 404 credits signature replaces refresh-token 401; estate shrank (VM 310 + CT601 gone)
- freshness: 36 stale / 93 fresh / 3 no-date -> 0 stale / 129 fresh / 3 no-date of 132 (same-day report regenerated after the corrections, per pattern 13)
- skills: 299 skills / 67 categories (count unchanged; generator arithmetic re-verified by hand against the live tree: 300 SKILL.md files - 1 nested duplicate = 299)
- runbooks/test-openrouter: live fallback chain rewritten (primary back to smart-route/custom:omni-101; mimo-v2.5, stepfun, qwen3.5/ollama-cloud out; guacano + kilo-reflection-beam-free in); ollama.com dropped from the URL test list; guacano/litellm noted as internal + key-gated
- comparisons/hermes-vs-nanobot: the 2026-10-20 'CT333 is zeroclaw in Proxmox naming' claim was itself wrong - pct list and hostname both return clawtest
- incidents/2026-10-22-nous: refresh-token failure gone from logs; new signature 404 insufficient_credits_for_paid_model from auxiliary.pulse_memory (ninth slot, back on nous, paid model) - open item revised
- incidents/2026-08-29-gmail: email crons now ok but root cause unfixed 74 days (himalaya message read still in fetch_newsletters.py; goimapnotify back up PID 441; monitor still paused)
- 30 further pages re-verified live and de-flagged with dated probe notes (ct207 ingress, CT205 guanaco v0.8.15, CT270 nextcloud 33.0.2.2, headless stack, hermes-ui both hosts, systemd units + cgroup limits, terminal/tool_search/mnemosyne/custom_providers config blocks, ops profile still deleted, hosts/pve guest lists)
- log.md addendum + index.md date bump; no page deleted, no secrets written
2026-11-11 03:52:48 +08:00

97 lines
5.3 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
---
title: Model Routing
type: concept
status: active
created: 2026-07-22
updated: 2026-11-11
verified_on: 2026-11-11
confidence: high
tags: [routing, model, inference]
sources: [raw/configs/hermes-config-sanitized.txt, current-profile.md]
---
# Model Routing
## Purpose
How Hermes selects providers, applies combos, and enforces context limits.
## Current Flow
- Main path: `custom:omni-101` / `smart-route` → `http://100.78.63.20:20128` (the OmniRoute node
`omniroute-1`), `api_mode: chat_completions`
- OmniRoute used for the **default chat path** (`custom:omni-101` / `smart-route`, since 2026-10-24 19:06) and for **nine auxiliary slots** (`approval`, `compression`, `curator`, `mcp`, `profile_describer`, `pulse_memory`, `skills_hub`, `title_generation`, `vision` — eight of them on `no-think/freeapi/auto`, plus `pulse_memory`; see the 2026-11-11 note below)
> **Auxiliary slots re-read 2026-11-11 — the set has grown from eight to nine.** Eight slots
> (`approval`, `compression`, `curator`, `mcp`, `profile_describer`, `skills_hub`,
> `title_generation`, `vision`) are unchanged: `provider: custom:omni-101`, `model:
> no-think/freeapi/auto`, each with a two-entry fallback chain
> (`no-think/freeapi/deepseek-v4-flash` → `no-think/kg/kilo-auto/free`, both on `custom:omni-101`).
> **New: `pulse_memory`** — `provider: nous`, `model: anthropic/claude-haiku-4.5`, and it is the
> slot behind the live 404s recorded above: nous is out of credits for that **paid** model, so
> `pulse_memory` is failing right now (3 errors on 2026-11-10 18:02 and 19:19). Also live:
> `auxiliary.free_only: true`.
> **Live `model:` block re-read 2026-10-26 — the default has moved TWICE.**
> `model.default` is **`smart-route`** on provider **`custom:omni-101`**
> (`http://100.78.63.20:20128`), `api_mode: chat_completions`. It moved to `nous` /
> `meituan/longcat-2.5-preview:free` on 2026-10-20 22:01 and **back** on **2026-10-24 19:06:10**.
> A six-provider top-level `fallback_providers` chain is live again, headed by the **dead `nous`**
> credential (`Invalid refresh token` since 2026-10-22).
> → [[systems/current-profile]] · [[incidents/2026-10-22-nous-portal-refresh-token-invalid]]
>
> **Re-read 2026-11-11 — the model block is unchanged, but the `nous` failure signature has moved
> on.** `model.default` is still `smart-route` on `custom:omni-101`
> (`base_url: http://100.78.63.20:20128`, `api_mode: chat_completions`,
> `persist_switch_by_default: false`) and the six-entry `fallback_providers` chain is unchanged in
> membership and order (`nous` → `opencode-zen` → `custom:guacano` → `custom:litellm` →
> `kilocode` → `custom`). The **refresh-token failure is gone from the logs**: `~/.hermes/.env`
> has no `NOUS_API_KEY` and `~/.hermes/auth.json` is live, and no 401/refresh-token error appears
> in `~/.hermes/logs/errors.log`. What *does* appear (3×, most recent **2026-11-10 19:19:33**) is
> **404 `insufficient_credits_for_paid_model`** for `anthropic/claude-haiku-4.5` — a **paid**
> model. So `nous` is no longer dead-at-auth; it is **out of credits for paid models**. Free nous
> models (position 1 is `meituan/longcat-2.5-preview:free`) have not been observed failing. The
> slot actually hit by that paid model is **`auxiliary.pulse_memory`** — see the auxiliary note
> below, which is currently a broken slot.
>
> **The five `custom_providers` and their live base URLs:**
>
> | Name | Base URL | Host |
> |---|---|---|
> | `litellm` | `http://litellm:4000/v1` | LiteLLM proxy (m93p CT500) |
> | `guacano` | `http://ollama:11000/v1` | Guanaco on CT205 |
> | `Omniroute:20128` | `http://omniroute:20128/v1` | OmniRoute on m93p CT101 |
> | `command-code` | `https://api.commandcode.ai/provider/v1` | remote |
> | `omni-101` | `http://100.78.63.20:20128` | OmniRoute node `omniroute-1` (direct IP, not a DNS name) |
>
> Note the **two distinct OmniRoute endpoints**: `Omniroute:20128` (DNS, m93p CT101) and
> `omni-101` (raw IP `100.78.63.20:20128`). `omni-101` is the pin on effectively every
> model-calling cron job, and it has been the single point of failure in the 10-14, 10-16/17 and
> 10-19 incidents — its `/api/health` reads `200 {"status":"ok"}` throughout, so **reachability
> probes do not detect that failure class**; only job-level error text does.
## Context Enforcement
- Three layers: model native → combo `data.config.context_length` → Hermes client cap
- Effective context is the minimum of the three
## Combo Behavior
- `priority` retries same provider up to `maxRetries`, then errors; it does NOT fall through
- `fill-first` walks the list on failure
- `auto`/`lkgp` recalculates per request using health scoring
- Same-provider entries in a combo share rate-limit buckets
- Session affinity caches provider in `session_model_history`
## Aliases and Mappings
- Model aliases resolved via `key_value` table (`modelAliases` namespace)
- Wildcard mappings via `model_combo_mappings`
- OpenCode/Zen entries may resolve back to `opencode-zen` provider internally
## History
- User-provided: OpenRouter was used previously; provider lists now centered on LiteLLM + OmniRoute
- User-provided: `persist_switch_by_default=false`; `/model` is per-session unless `--global` used
- Inferred: Configuration undergoes periodic corrections related to vision, model fallback, and browser backend tuning
## Related
- [[model-providers]]
- [[systems/omniroute]]
- [[openrouter]]