Align OpenClaw Ocabra roster with the ctx-tagged model set

Bring OpenClaw's models.providers.ocabra and agents.defaults.models in
line with the OpenCode roster (the 10 ctx-tagged ids, matching
contextWindow/maxTokens). Default model and its fallback move from
ocabra/qwen3:14b to ocabra/gemma4:12b-ctx256k. README updated.

Also fixes a latent bug: the ocabra_provider_json Python block used
lowercase true/false, which raised NameError at runtime, so the provider
metadata was never actually applied. Switched to Python True/False so
json.dumps emits valid JSON.

continue-config.yaml is intentionally left untouched: Continue needs its
own autocomplete/embedding models that are not part of the chat roster.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: successbyfailure

workspaces/OpenClaw/README.md

@@ -32,7 +32,7 @@ Template inicial para ejecutar OpenClaw sobre la imagen `ghcr.io/makespacemadrid
 - Puedes entrar por KasmVNC (escritorio) o terminal.
 - `[OpenClaw] Auto-iniciar servicio`: arranca OpenClaw al iniciar.
 - `[OpenClaw] Directorio de trabajo`: directorio desde el que se ejecuta OpenClaw.
-- `[OpenClaw] Modelo por defecto`: modelo por defecto para OpenClaw (por defecto `ocabra/qwen3:14b`).
+- `[OpenClaw] Modelo por defecto`: modelo por defecto para OpenClaw (por defecto `ocabra/gemma4:12b-ctx256k`).
 - `Provisionar API key Ocabra automáticamente`: genera una key de 30 días si no aportas una.
 - `Provisionar API key FreeAPI automáticamente`: genera y precarga una key de FreeAPI al crear el workspace.
 - `TF_VAR_ocabra_endpoint_base_url`: base URL OpenAI-compatible de Ocabra por defecto.
@@ -57,7 +57,7 @@ Template inicial para ejecutar OpenClaw sobre la imagen `ghcr.io/makespacemadrid
 - El template desactiva el pairing de dispositivos (`gateway.controlUi.dangerouslyDisableDeviceAuth=true`) para entrar directamente con token.
 - El template puede autoprovisionar 2 credenciales API (si hay endpoints): Ocabra (`OCABRA_KEY`) y FreeAPI (`FREEAPI_KEY_ENDPOINT`), cada una con su propio toggle en el formulario.
 - El template crea `auth-profiles.json` con perfiles `ocabra:manual` y/o `freeapi:manual`.
-- El template define `models.providers.ocabra` con `qwen3:14b`, `qwen3:32b`, `qwen3-coder:30b`, `gpt-oss:20b`.
+- El template define `models.providers.ocabra` con el roster ctx-tagged: `gemma4:12b-ctx256k`, `qwen3-coder:30b-ctx256k`, `nemotron-3-nano:30b-ctx384k`, `nemotron-3-nano:30b-ctx256k`, `qwen3.6:ctx128k`, `gemma4:26b-ctx160k`, `ministral-3:14b-ctx128k`, `qwen3.5:27b-ctx96k`, `nemotron3:33b-ctx64k`, `ravenx-256k`.
 - El template detecta modelos FreeAPI acabados en `-ha` y, cuando está disponible `/model/info` (LiteLLM), enriquece automáticamente capacidades, tokens y costes por modelo.
 - El template rellena `agents.defaults.models` con esos modelos para que aparezcan en el selector de agentes.
 - El template asegura `agents.list` con entrada `id: "main"` para que los cambios de modelo en la UI de agentes se marquen como modificados y el botón `Save` se habilite.

workspaces/OpenClaw/main.tf

@@ -84,9 +84,9 @@ data "coder_parameter" "openclaw_workdir" {
 data "coder_parameter" "openclaw_default_model" {
   name         = "05_openclaw_default_model"
   display_name = "[OpenClaw] Modelo por defecto"
-  description  = "Modelo por defecto de OpenClaw (ej. ocabra/qwen3:14b). Si se deja vacío, no se fuerza."
+  description  = "Modelo por defecto de OpenClaw (ej. ocabra/gemma4:12b-ctx256k). Si se deja vacío, no se fuerza."
   type         = "string"
-  default      = "ocabra/qwen3:14b"
+  default      = "ocabra/gemma4:12b-ctx256k"
   mutable      = true
 }
 
@@ -756,12 +756,16 @@ cfg = {
     "auth": "api-key",
     "api": "openai-completions",
     "models": [
-        {"id": "qwen3-coder:30b", "name": "Qwen3 Coder 30b", "reasoning": false, "input": ["text"], "contextWindow": 262144, "maxTokens": 8192},
-        {"id": "devstral-small-2:24b", "name": "Devstral Small 2 24b", "reasoning": false, "input": ["text"], "contextWindow": 393216, "maxTokens": 8192},
-        {"id": "qwen3.5:27b", "name": "Qwen3.5 27b", "reasoning": true, "input": ["text"], "contextWindow": 262144, "maxTokens": 8192},
-        {"id": "qwen3.6:latest", "name": "Qwen3.6", "reasoning": true, "input": ["text"], "contextWindow": 262144, "maxTokens": 8192},
-        {"id": "gemma4:26b", "name": "Gemma 4 26b", "reasoning": true, "input": ["text"], "contextWindow": 262144, "maxTokens": 8192},
-        {"id": "qwen3:8b", "name": "Qwen3 8b", "reasoning": true, "input": ["text"], "contextWindow": 40960, "maxTokens": 8192},
+        {"id": "gemma4:12b-ctx256k", "name": "Gemma 4 12b (256K)", "reasoning": True, "input": ["text"], "contextWindow": 262144, "maxTokens": 16384},
+        {"id": "qwen3-coder:30b-ctx256k", "name": "Qwen3 Coder 30b (256K)", "reasoning": False, "input": ["text"], "contextWindow": 262144, "maxTokens": 32768},
+        {"id": "nemotron-3-nano:30b-ctx384k", "name": "Nemotron 3 Nano 30b (384K)", "reasoning": True, "input": ["text"], "contextWindow": 393216, "maxTokens": 16384},
+        {"id": "nemotron-3-nano:30b-ctx256k", "name": "Nemotron 3 Nano 30b (256K)", "reasoning": True, "input": ["text"], "contextWindow": 262144, "maxTokens": 16384},
+        {"id": "qwen3.6:ctx128k", "name": "Qwen3.6 (128K)", "reasoning": True, "input": ["text"], "contextWindow": 131072, "maxTokens": 16384},
+        {"id": "gemma4:26b-ctx160k", "name": "Gemma 4 26b (160K)", "reasoning": True, "input": ["text"], "contextWindow": 163840, "maxTokens": 16384},
+        {"id": "ministral-3:14b-ctx128k", "name": "Ministral 3 14b (128K)", "reasoning": True, "input": ["text"], "contextWindow": 131072, "maxTokens": 8192},
+        {"id": "qwen3.5:27b-ctx96k", "name": "Qwen3.5 27b (96K)", "reasoning": True, "input": ["text"], "contextWindow": 98304, "maxTokens": 16384},
+        {"id": "nemotron3:33b-ctx64k", "name": "Nemotron3 33b (64K)", "reasoning": True, "input": ["text"], "contextWindow": 65536, "maxTokens": 8192},
+        {"id": "ravenx-256k", "name": "RavenX CyberAgent (256K)", "reasoning": False, "input": ["text"], "contextWindow": 262144, "maxTokens": 32768},
     ],
 }
 print(json.dumps(cfg, separators=(",", ":")))
@@ -948,7 +952,7 @@ PY
 import json, os
 allowed = {}
 if os.environ.get("HAS_OCABRA") == "1":
-    for model in ("qwen3-coder:30b", "devstral-small-2:24b", "qwen3.5:27b", "qwen3.6:latest", "gemma4:26b", "qwen3:8b"):
+    for model in ("gemma4:12b-ctx256k", "qwen3-coder:30b-ctx256k", "nemotron-3-nano:30b-ctx384k", "nemotron-3-nano:30b-ctx256k", "qwen3.6:ctx128k", "gemma4:26b-ctx160k", "ministral-3:14b-ctx128k", "qwen3.5:27b-ctx96k", "nemotron3:33b-ctx64k", "ravenx-256k"):
         allowed[f"ocabra/{model}"] = {}
 if os.environ.get("HAS_FREEAPI") == "1":
     try:
@@ -1002,7 +1006,7 @@ PY
 OPENCLAW_PORT="$${OPENCLAW_PORT:-3333}"
 OPENCLAW_GATEWAY_TOKEN="$${OPENCLAW_GATEWAY_TOKEN:-}"
 OPENCLAW_WORKDIR="$${OPENCLAW_WORKDIR:-$HOME/Projects}"
-OPENCLAW_DEFAULT_MODEL="$${OPENCLAW_DEFAULT_MODEL:-ocabra/qwen3:14b}"
+OPENCLAW_DEFAULT_MODEL="$${OPENCLAW_DEFAULT_MODEL:-ocabra/gemma4:12b-ctx256k}"
 EOF
     chmod 600 "$HOME/.local/state/openclaw/runtime.env"