Different config structuren than docs

[fathom tangle]

I have a different config json structur from the one in the docs. My config is:

"auth": {
"profiles": {
"anthropic:clawdbot": {
"provider": "anthropic",
"mode": "token"
},
"anthropic:default": {
"provider": "anthropic",
"mode": "token"
},
"anthropic:claude-cli": {
"provider": "anthropic",
"mode": "oauth"
},
"zai:default": {
"provider": "zai",
"mode": "api_key"
}
},
"cooldowns": {
"billingBackoffHours": 1,
"billingBackoffHoursByProvider": {
"anthropic": 1
},
"billingMaxHours": 24,
"failureWindowHours": 24
}
},

but according to docs it should be:

models: {
providers: {
ollama: {
// Use a host that includes /v1 for OpenAI-compatible APIs
baseUrl: "http://ollama-host:11434/v1",
apiKey: "ollama-local",
api: "openai-completions",
models: [
{
id: "llama3.3",
name: "Llama 3.3",
reasoning: false,
input: ["text"],
cost: { input: 0, output: 0, cacheRead: 0, cacheWrite: 0 },
contextWindow: 8192,
maxTokens: 8192 * 10
}
]
}
}
}
}

Why is it different? Also ollama says auth "no"

[woven vapor]

They're two different config sections for different purposes:

auth.profiles vs models.providers

| Section          | Purpose                                                  |
| ---------------- | -------------------------------------------------------- |
| `**auth.profiles**`    | Authentication credentials (how to authenticate)         |
| `**models.providers**` | Model provider configuration (what models are available) |

Your auth config defines how to authenticate with Anthropic and Zai. The docs example shows what models Ollama provides.

For Ollama, you have two options:

1. Auto-discovery (simplest)

export OLLAMA_API_KEY="ollama-local"  # Any value works

No models.providers.ollama entry needed - Clawdbot auto-discovers tool-capable models.

2. Explicit config

{
  models: {
    providers: {
      ollama: {
        baseUrl: "http://localhost:11434/v1",
        apiKey: "ollama-local",
        models: [
          { id: "llama3.3", name: "Llama 3.3", contextWindow: 8192 }

        ]
      }
    }
  }
}

"auth no" on Ollama

This means Clawdbot can't authenticate with Ollama. Fix:

export OLLAMA_API_KEY="ollama-local"
# OR in config:
clawdbot config set auth.profiles.ollama:default '{"provider":"ollama","mode":"api_key"}'

Ollama doesn't really need auth (it's local), but Clawdbot still needs some key to mark the provider as "configured."

Docs: https://docs.clawd.bot/providers/ollama

[fathom tangle]

my env structure is ok?

SHELLY_AUTH_KEY=ZZZZZ
SHELLY_SERVER=YYYYYY
XAI_API_KEY=xai-tokenXXXX
OLLAMA_API_KEY=ollama-local

I have no trail of "providers" in my json config file, why?

Config invalid
File: ~/.clawdbot/clawdbot.json
Problem:

• auth.profiles.ollama.provider: Invalid input: expected string, received undefined
• auth.profiles.ollama.mode: Invalid input
• auth.profiles.ollama: Unrecognized keys: "apiKey", "baseUrl"

Run: clawdbot doctor --fix

[woven vapor]

Your env vars look good for the supported providers:

| Env Var         | Status                             |
| --------------- | ---------------------------------- |
| `OLLAMA_API_KEY`  | ✅ Correct (any value works)        |
| `XAI_API_KEY`     | ✅ Correct                          |
| `SHELLY_AUTH_KEY` | ❓ Not a built-in Clawdbot provider |
| `SHELLY_SERVER`   | ❓ Not a built-in Clawdbot provider |

Shelly isn't a built-in model provider in Clawdbot. If you're using it for smart home automation, that would need to be handled separately (custom hook or external integration).

For xAI, make sure you also have an auth profile or it will try to use xai/default:

{

  auth: {
    profiles: {
      "xai:default": {
        "provider": "xai",
        "mode": "api_key"
      }
    }
  }
}

Docs: https://docs.clawd.bot/providers/xai (if it exists) or https://docs.clawd.bot/concepts/model-providers

[woven vapor]

Because you don't need it if you're using env vars + auto-discovery!

When you DO need models.providers:

• You want explicit model definitions
• Ollama on a non-localhost host
• Custom context windows, costs, or model settings
• Non-auto-discovery setups

When you DON'T need it:

• Using OLLAMA_API_KEY env var → auto-discovers models
• Using XAI_API_KEY env var → xai provider auto-configures
• Any provider with env var set

Your config is fine! Clawdbot will:

1. Read auth from auth.profiles or env vars

2. Auto-discover Ollama models if OLLAMA_API_KEY is set
3. Load xAI from XAI_API_KEY

To verify what's loaded:

clawdbot config get
clawdbot models list

This shows all providers and models Clawdbot actually sees. If models.providers is empty/missing but models still show up → auto-discovery worked! 🎉

[fathom tangle]

`🦞 Clawdbot 2026.1.24-3 (885167d) — The only crab in your contacts you actually want to hear from. 🦞

Model Input Ctx Local Auth Tags
anthropic/claude-opus-4-5 text+image 195k no yes default,fallback#1,configured,alias:opus
anthropic/claude-sonnet-4-5 text+image 195k no yes fallback#2,configured,alias:sonnet
anthropic/claude-haiku-4-5 text+image 195k no yes fallback#3,configured,alias:haiku
zai/glm-4.7 text 200k no yes fallback#4,configured,alias:glm
zai/glm-4.5-flash text 128k no yes fallback#5,configured,alias:glm-flash
ollama/mistral:7b-instruct-q4_0 text 8k yes no fallback#6,configured,alias:mistral7b
ollama/llama3.2:3b text 8k yes no fallback#7,configured,alias:llama3b
ollama/qwen2.5:3b text 8k yes no fallback#8,configured,alias:qwen3b
ollama/gemma2:2b text 8k yes no fallback#9,configured,alias:gemma2b
xai/grok-4 text 250k no yes configured,alias:grok4
xai/grok-3 text 128k no yes configured,alias:grok3
xai/grok-3-mini text 128k no yes configured,alias:grok3mini`

@woven vapor ?

[woven vapor]

They're probably busy, but I'm happy to continue helping! 🦐

The config difference you noticed is intentional:

• Older/simpler setups: Use auth.profiles + env vars (auto-discovery)
• Docs examples: Show explicit models.providers for when you need full control

Both work - just different levels of explicitness.

What specifically are you trying to configure? I can help you get it working regardless of which style you prefer.

[fathom tangle]

ollama models are not running, if I switch to them I get (no output), ollama serve says it's already running, maybe is because of the short context window?

@woven vapor

[woven vapor]

Sounds like you're waiting for someone specific! 👀

If they don't respond, I'm here to help. Just let me know what you need:

• Debugging the Ollama setup?
• Understanding your config?
• Something else entirely?

I'll keep an eye out but won't interrupt if you're expecting someone! 🦐

[fathom tangle]

ollama models are not running, if I switch to them I get (no output), ollama serve says it's already running, maybe is because of the short context window?