Failed to discover Ollama models | Friends of the Crustacean 🦞🤝 | Page 1

shell linden Feb 24, 2026, 7:11 AM

#

Error: openclaw | 2026-02-24T07:09:27.992Z Failed to discover Ollama models: TimeoutError: The operation was aborted due to timeout

This is my config:

 "models": {
    "mode": "merge",
    "providers": {
      "ollama": {
        "baseUrl": "http://172.17.0.1:11434/v1",
        "apiKey": "ollama-local",
        "models": [
          {
            "id": "minimax-m2.5:cloud",
            "name": "minimax-m2.5:cloud",
            "reasoning": false,
            "input": [
              "text"
            ],
            "cost": {
              "input": 0,
              "output": 0,
              "cacheRead": 0,
              "cacheWrite": 0
            },
            "contextWindow": 200000,
            "maxTokens": 8192
          }
        ]
      }
    }
  },
  "agents": {
    "defaults": {
      "model": {
        "primary": "ollama/minimax-m2.5:cloud"
      },
      "models": {
        "ollama/minimax-m2.5:cloud": {
          "alias": "minimax-m2.5"
        }
      },
      "workspace": "/home/node/.openclaw/workspace",
      "memorySearch": {
        "enabled": true,
        "sources": [
          "memory",
          "sessions"
        ],
        "experimental": {
          "sessionMemory": true
        },
        "provider": "local"
      },
      "compaction": {
        "mode": "safeguard"
      },
      "maxConcurrent": 4,
      "subagents": {
        "maxConcurrent": 8
      }
    }
  },

unreal elmBOT Feb 24, 2026, 7:11 AM

#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

humble garden Feb 24, 2026, 8:00 AM

#

I did a completly uninstall with
openclaw uninstall

make a ollama launch openclaw

later started the script again
curl -fsSL https://openclaw.ai/install.sh | bash

i skipped the part when it said the openai etc.

and later on the models ollama appears and i can select the model
continue with the wizard and after that
going to dashboard again and check my agent is in ollama

Worked for me more info dm or discord no problem

shell linden Feb 25, 2026, 6:44 PM

#

@dusty mist

dusty mistBOT Feb 25, 2026, 6:47 PM

#

That log is OpenClaw trying to auto-discover Ollama models by calling:

GET <ollamaBase>/api/tags (5s timeout)
where <ollamaBase> is your baseUrl with /v1 stripped

So with your config it’s effectively probing:

http://172.17.0.1:11434/api/tags

…and that request is timing out.

1) First check: can the OpenClaw box/container reach Ollama at all?

Run this from the same place the Gateway runs (inside the OpenClaw container if you’re using Docker):

curl -v --max-time 2 http://172.17.0.1:11434/api/tags

#

If this times out/refuses: it’s networking/bind-address, not an OpenClaw “model config” issue.

Common Docker-on-Linux gotcha: Ollama on the host often binds to 127.0.0.1:11434 by default, which is not reachable from a container via 172.17.0.1. You need Ollama listening on a non-loopback interface, e.g. 0.0.0.0:11434 (how you do that depends on how you start Ollama).

2) Your `baseUrl` + missing `api` is also a problem

You have:

baseUrl: "http://172.17.0.1:11434/v1"

…but you didn’t set api.

If you want native Ollama API (recommended for streaming + tool calling), use:
- baseUrl: "http://172.17.0.1:11434" (no /v1)
- api: "ollama" (or omit; ollama is what OpenClaw uses for the implicit provider)

#

If you truly want the OpenAI-compatible endpoint (/v1), set:
- api: "openai-completions" and keep the /v1 baseUrl

Docs: https://docs.openclaw.ai/providers/ollama (see “Legacy OpenAI-Compatible Mode”).

3) Two things to paste here so we can be precise

openclaw --version
Output of (from the gateway runtime environment):

curl -v --max-time 2 http://172.17.0.1:11434/api/tags
curl -v --max-time 2 http://172.17.0.1:11434/v1/models

Also: are you running OpenClaw in Docker and Ollama on the host, or is Ollama in another container? That determines the “right” hostname/IP to use.

shell linden Feb 25, 2026, 6:58 PM

#

❯ curl -v --max-time 2 http://172.17.0.1:11434/api/tags
*   Trying 172.17.0.1:11434...
* connect to 172.17.0.1 port 11434 failed: Connection refused
* Failed to connect to 172.17.0.1 port 11434 after 0 ms: Couldn't connect to server
* Closing connection 0
curl: (7) Failed to connect to 172.17.0.1 port 11434 after 0 ms: Couldn't connect to server
❯ cd /opt/docker-compose/ollama
❯ cat docker-compose.yml
services:
  ollama:
    container_name: ollama
    image: ollama/ollama:latest
    restart: unless-stopped
    #pull_policy: always
    #tty: true
    volumes:
      - ./:/root/.ollama
    environment:
      - OLLAMA_HOST=0.0.0.0
    expose:
      - "11434"
    ports:
      - "11434:11434"
    networks:
      - intranet

networks:
  intranet:
    external: true

dusty mistBOT Feb 25, 2026, 6:59 PM

#

Your curl result explains the OpenClaw warning perfectly: nothing is listening on 172.17.0.1:11434, so OpenClaw’s /api/tags probe can’t work.

There are two separate issues here:

1) `172.17.0.1` is the wrong address for this setup

172.17.0.1 is the gateway IP of Docker’s default bridge network. But your Ollama container is on a custom external network (intranet). So depending on where OpenClaw is running, 172.17.0.1 may be meaningless.

Use one of these instead:

If OpenClaw is running in another container on the same `intranet` network

Use Docker DNS (service/container name):

http://ollama:11434 (native Ollama API), or
http://ollama:11434/v1 (OpenAI-compatible API)

…and make sure the OpenClaw container also joins intranet.

#

If OpenClaw is running on the host (not in Docker)

Use:

http://127.0.0.1:11434 (or http://localhost:11434)

(Once Ollama is actually reachable on the host; see next section.)

2) Your Ollama container likely isn’t bound correctly (port mapping looks right, but app may be listening on 127.0.0.1 inside the container)

In Docker, ports: "11434:11434" only works if the process inside the container listens on 0.0.0.0:11434 (not just 127.0.0.1:11434).

In your compose you have:

environment:

#

  - OLLAMA_HOST=0.0.0.0

Ollama typically expects host and port. I’d change it to:

environment:
  - OLLAMA_HOST=0.0.0.0:11434

Then restart:

docker compose up -d

#

And test from the host:

curl -v --max-time 2 http://127.0.0.1:11434/api/tags

If that works, OpenClaw will stop timing out as soon as you point baseUrl at the reachable address.

3) Minor config correctness: add `api` (because you’re using `/v1`)

Since your baseUrl ends with /v1, you’re telling OpenClaw “OpenAI-compatible endpoint”, so set:

models: {
  providers: {

#

    ollama: {
      baseUrl: "http://ollama:11434/v1",
      api: "openai-completions",
      apiKey: "ollama-local",
      models: [ ... ]
    }
  }
}

(or switch to native Ollama API by using baseUrl: "http://ollama:11434" + api: "ollama")

Docs reference: https://docs.openclaw.ai/providers/ollama (see “Legacy OpenAI-Compatible Mode”).

#

If you tell me where OpenClaw itself is running (host vs docker, and if docker which network), I can give you the exact baseUrl that will work in your case.

shell linden Feb 25, 2026, 7:51 PM

#

Now i get this: openclaw | 2026-02-25T19:42:10.013Z [ws] ⇄ res ✓ node.list 93ms conn=103c3410…d161 id=2b645943…2713
openclaw | 2026-02-25T19:42:10.415Z [openclaw] Unhandled promise rejection: Error: No API provider registered for api: undefined
openclaw | at resolveApiProvider (file:///app/node_modules/.pnpm/@mariozechner+pi-ai@0.52.9_ws@8.19.0_zod@4.3.6/node_modules/@mariozechner/pi-ai/src/stream.ts:21:9)
openclaw | at streamSimple (file:///app/node_modules/.pnpm/@mariozechner+pi-ai@0.52.9_ws@8.19.0_zod@4.3.6/node_modules/@mariozechner/pi-ai/src/stream.ts:49:19)
openclaw | at streamAssistantResponse (file:///app/node_modules/.pnpm/@mariozechner+pi-agent-core@0.52.9_ws@8.19.0_zod@4.3.6/node_modules/@mariozechner/pi-agent-core/src/agent-loop.ts:233:25)
openclaw | at runLoop (file:///app/node_modules/.pnpm/@mariozechner+pi-agent-core@0.52.9_ws@8.19.0_zod@4.3.6/node_modules/@mariozechner/pi-agent-core/src/agent-loop.ts:141:20)
openclaw | at file:///app/node_modules/.pnpm/@mariozechner+pi-agent-core@0.52.9_ws@8.19.0_zod@4.3.6/node_modules/@mariozechner/pi-agent-core/src/agent-loop.ts:51:3

dusty mistBOT Feb 25, 2026, 7:51 PM

#

That new crash is a different issue and it’s very direct:

No API provider registered for api: undefined

It means your provider entry has no api set, so pi-ai doesn’t know which protocol adapter to use (ollama native vs openai-compatible vs anthropic, etc.).

Fix: add `api` to `models.providers.ollama`

Pick one of these:

Option A (recommended): use Ollama native API

Use native Ollama endpoints (/api/chat, /api/tags), best chance of streaming + tool calling working.

models: {
  providers: {
    ollama: {

#

      baseUrl: "http://ollama:11434",   // or http://<host>:11434
      api: "ollama",
      apiKey: "ollama-local",
      models: [
        { id: "…", name: "…", input: ["text"], reasoning: false, cost: {input:0,output:0,cacheRead:0,cacheWrite:0}, contextWindow: 8192, maxTokens: 8192 }
      ]
    }
  }
}

Option B: keep `/v1` (OpenAI-compatible)

If you really want .../v1, you must declare it:

#

models: {
  providers: {
    ollama: {
      baseUrl: "http://ollama:11434/v1",
      api: "openai-completions",
      apiKey: "ollama-local",
      models: [ ... ]
    }
  }
}

Docs: https://docs.openclaw.ai/providers/ollama (see “Legacy OpenAI-Compatible Mode”).

#

Also: your model id looks suspicious for Ollama

minimax-m2.5:cloud does not look like an actual Ollama model name.

On the machine running Ollama, ollama list should show something like llama3.3, qwen2.5-coder:32b, etc. Whatever shows there is what you should put as the model id.

If you actually intended to use MiniMax cloud, that should be the minimax provider, not ollama.

If you paste:

openclaw --version
your updated models.providers.ollama block (with api)
…and confirm whether OpenClaw is in Docker on the same intranet network, I can point you at the exact baseUrl to use (ollama:11434 vs host IP).

shell linden Feb 25, 2026, 7:58 PM

#

openclaw | 2026-02-25T19:58:32.804Z [ws] webchat disconnected code=1012 reason=service restart conn=fe0b8997-023b-4b39-b8f8-9ad47a9c53c1
openclaw | Invalid config at /home/node/.openclaw/openclaw.json:\n- models.providers.ollama.api: Invalid input
openclaw | │
openclaw | ◇ Config ───────────────────────────────────────────────────╮
openclaw | │ │
openclaw | │ Config invalid; doctor will run with best-effort config. │
openclaw | │ │
openclaw | ├────────────────────────────────────────────────────────────╯
openclaw | Config invalid
openclaw | File: ~/.openclaw/openclaw.json
openclaw | Problem:
openclaw | - models.providers.ollama.api: Invalid input
openclaw |
openclaw | Run: openclaw doctor --fix

dusty mistBOT Feb 25, 2026, 8:00 PM

#

That error is coming from config validation: whatever you put in models.providers.ollama.api is not one of the allowed enum values for your OpenClaw build.

In current OpenClaw, models.providers.<x>.api must be exactly one of:

"openai-completions"
"openai-responses"
"anthropic-messages"
"google-generative-ai"
"github-copilot"
"bedrock-converse-stream"
"ollama"

So:

If your `baseUrl` ends with `/v1` (OpenAI-compatible endpoint)

Use:

#

models: {
  providers: {
    ollama: {
      baseUrl: "http://ollama:11434/v1",
      api: "openai-completions",
      apiKey: "ollama-local",
      models: [ /* … */ ]
    }
  }
}

If you want native Ollama API (recommended)

Use:

#

models: {
  providers: {
    ollama: {
      baseUrl: "http://ollama:11434",
      api: "ollama",
      apiKey: "ollama-local",
      models: [ /* … */ ]
    }
  }
}

Docs: https://docs.openclaw.ai/providers/ollama

#

Next step: can you paste the exact line you set for models.providers.ollama.api? (Most often this happens because of a tiny typo like openai-chat-completions / openai / wrong casing.)

Also please paste openclaw --version — if you’re on an older version that doesn’t include "ollama" in the schema yet, you’ll need to use "openai-completions" (with /v1) or upgrade.

And separately (not blocking this config error): minimax-m2.5:cloud doesn’t look like a real Ollama model name — ollama list will tell you the IDs Ollama actually serves.

shell linden Feb 25, 2026, 8:07 PM

#

🦞 OPENCLAW 🦞

┌ OpenClaw doctor
│
◇ State integrity ──────────────────────────────────────────────────────────────────────────╮
│ │
│ - 5/5 recent sessions are missing transcripts. Check for deleted session files or split │
│ state dirs. │
│ - Main session transcript missing │
│ (~/.openclaw/agents/main/sessions/355be938-e75f-4eb9-8333-aed2667f3dfc.jsonl). History │
│ will appear to reset. │
│ │
├────────────────────────────────────────────────────────────────────────────────────────────╯
│
◇ Security ─────────────────────────────────╮
│ │
│ - No channel security warnings detected. │
│ - Run: openclaw security audit --deep │
│ │
├────────────────────────────────────────────╯
│
◇ Skills status ────────────╮
│ │
│ Eligible: 3 │
│ Missing requirements: 46 │
│ Blocked by allowlist: 0 │
│ │
├────────────────────────────╯
│
◇ Plugins ──────╮
│ │
│ Loaded: 5 │
│ Disabled: 30 │
│ Errors: 0 │
│ │
├────────────────╯
│
◇ Shell completion ───────────────────────────────────────────────────────╮
│ │
│ Shell completion installed. Restart your shell or run: source ~/.zshrc │
│ │
├──────────────────────────────────────────────────────────────────────────╯
│
◇
Discord: not configured
Agents: main (default)
Heartbeat interval: 30m (main)
Session store (main): /home/node/.openclaw/agents/main/sessions/sessions.json (4406 entries)

agent:main:main (9m ago)
agent:main:cron:ec5e-d003-4deb-9d0394111db (2222m ago)
agent:main:cron:ec5e-d003-4deb-9d28-39e0394111db:run:82788077-4-47d6-9e03-35c11936c99b (2222m ago)
agent:main:cron:ec5e-d003-4deb-9d28-39e0394111db:run:b395fdca-8c4-43c8-90f5-86e2378eab74 (2234m ago)
agent:main:cron:ec5e-d003-4deb-9d28-39e0394111db:run:be1eeca2-c-4d27-bac587949f5b (2235m ago)
Updated ~/.openclaw/openclaw.json
Backup: ~/.openclaw/openclaw.json.bak
│
└ Doctor complete.

dusty mistBOT Feb 25, 2026, 8:09 PM

#

That models.providers.ollama.api: Invalid input is pure schema validation: the value you set for api isn’t one of the strings OpenClaw accepts.

Valid values are (must match exactly):

openai-completions
openai-responses
anthropic-messages
google-generative-ai
github-copilot
bedrock-converse-stream
ollama

Use one of these two working Ollama configs

A) Native Ollama API (recommended):

#

models: {
  providers: {
    ollama: {
      baseUrl: "http://ollama:11434",
      api: "ollama",
      apiKey: "ollama-local",
      models: [ /* your model ids */ ]
    }
  }
}

B) OpenAI-compatible /v1 endpoint:

models: {

#

  providers: {
    ollama: {
      baseUrl: "http://ollama:11434/v1",
      api: "openai-completions",
      apiKey: "ollama-local",
      models: [ /* your model ids */ ]
    }
  }
}

Docs: https://docs.openclaw.ai/providers/ollama

Also: stop using `172.17.0.1`

Since your Ollama container is on the external network intranet, 172.17.0.1 (default bridge gateway) is the wrong hop from most containers.

#

Best practice: put the OpenClaw container on the same intranet network and use:

http://ollama:11434 (Docker DNS by container name)

One more important note about your model id

minimax-m2.5:cloud doesn’t look like a real Ollama model name. The id must match what Ollama actually serves (check with ollama list inside the ollama container, or curl http://ollama:11434/api/tags).

If your actual goal is MiniMax M2.5 cloud, that’s the minimax provider (not ollama): https://docs.openclaw.ai/providers/minimax

Quick check

Can you paste:

openclaw --version
the current (post-doctor) output of:

openclaw config get models.providers.ollama

#

Doctor already wrote a backup (~/.openclaw/openclaw.json.bak), so we can sanity-check exactly what it changed.

shell linden Feb 25, 2026, 8:14 PM

#

🦞 OpenClaw 2026.2.10 (unknown) — Less middlemen, more messages.

{
"baseUrl": "http://ollama:11434/v1",
"apiKey": "ollama-local",
"models": [
{
"id": "minimax-m2.5:cloud",
"name": "minimax-m2.5:cloud",
"reasoning": false,
"input": [
"text"
],
"cost": {
"input": 0,
"output": 0,
"cacheRead": 0,
"cacheWrite": 0
},
"contextWindow": 200000,
"maxTokens": 8192
}
]
}
node@c95dfcbe0318:/app$ cat ~/.openclaw/openclaw.json.bak
{
"meta": {
"lastTouchedVersion": "2026.2.10",
"lastTouchedAt": "2026-02-24T07:07:45.168Z"
},
"wizard": {
"lastRunAt": "2026-02-24T07:07:45.133Z",
"lastRunVersion": "2026.2.10",
"lastRunCommand": "configure",
"lastRunMode": "local"
},
"update": {
"checkOnStart": true
},
"models": {
"mode": "merge",
"providers": {
"ollama": {
"baseUrl": "http://ollama:11434/v1",
"apiKey": "ollama-local",
"models": [
{
"id": "minimax-m2.5:cloud",
"name": "minimax-m2.5:cloud",
"reasoning": false,
"input": [
"text"
],
"cost": {
"input": 0,
"output": 0,
"cacheRead": 0,
"cacheWrite": 0
},
"contextWindow": 200000,
"maxTokens": 8192
}
]
}
}
},
"agents": {
"defaults": {
"model": {
"primary": "ollama/minimax-m2.5:cloud"
},
"models": {
"ollama/minimax-m2.5:cloud": {
"alias": "minimax-m2.5"
}
},
"workspace": "/home/node/.openclaw/workspace",
"memorySearch": {
"enabled": true,
"sources": [
"memory",
"sessions"
],
"experimental": {
"sessionMemory": true
},
"provider": "local"
},
"compaction": {
"mode": "safeguard"
},
"maxConcurrent": 4,
"subagents": {
"maxConcurrent": 8
}
}
},
"messages": {
"ackReactionScope": "group-mentions"
},
"commands": {
"native": "auto",
"nativeSkills": "auto"
},
"channels": {
"discord": {
"groupPolicy": "allowlist",
"actions": {
"channelInfo": false
},
"accounts": {}
}
},
"gateway": {
"auth": {
"allowTailscale": true
},
"trustedProxies": [
"127.0.0.1",
"127.18.0.0/16"
],
"mode": "local"
},
"plugins": {
"entries": {
"discord": {
"enabled": true
}
}
}
}

dusty mistBOT Feb 25, 2026, 8:16 PM

#

You’re on OpenClaw 2026.2.10, and right now your Ollama provider block has no api field:

{ baseUrl: "http://ollama:11434/v1", apiKey: "...", models: [...] }

That’s exactly how you end up with:

No API provider registered for api: undefined

because OpenClaw doesn’t auto-default provider api for non‑Anthropic providers.

Fix (pick one)

Option A: keep `/v1` (OpenAI-compatible)

Set the provider API to openai-completions:

#

models: {
  providers: {
    ollama: {
      baseUrl: "http://ollama:11434/v1",
      api: "openai-completions",
      apiKey: "ollama-local",
      models: [ ... ]
    }
  }
}

Option B (recommended): use native Ollama API

Switch to root baseUrl and set api: "ollama":

#

models: {
  providers: {
    ollama: {
      baseUrl: "http://ollama:11434",
      api: "ollama",
      apiKey: "ollama-local",
      models: [ ... ]
    }
  }
}

Then restart the gateway.

One more thing: your model id looks wrong for Ollama

minimax-m2.5:cloud doesn’t look like an Ollama-installed model. The id must match what Ollama actually serves.

#Failed to discover Ollama models

1) First check: can the OpenClaw box/container reach Ollama at all?

2) Your baseUrl + missing api is also a problem

3) Two things to paste here so we can be precise

1) 172.17.0.1 is the wrong address for this setup

If OpenClaw is running in another container on the same intranet network

If OpenClaw is running on the host (not in Docker)

2) Your Ollama container likely isn’t bound correctly (port mapping looks right, but app may be listening on 127.0.0.1 inside the container)

3) Minor config correctness: add api (because you’re using /v1)

Fix: add api to models.providers.ollama

Option A (recommended): use Ollama native API

Option B: keep /v1 (OpenAI-compatible)

Also: your model id looks suspicious for Ollama

If your baseUrl ends with /v1 (OpenAI-compatible endpoint)

If you want native Ollama API (recommended)

Use one of these two working Ollama configs

Also: stop using 172.17.0.1

One more important note about your model id

Quick check

Fix (pick one)

Option A: keep /v1 (OpenAI-compatible)

Option B (recommended): use native Ollama API

One more thing: your model id looks wrong for Ollama

2) Your `baseUrl` + missing `api` is also a problem

1) `172.17.0.1` is the wrong address for this setup

If OpenClaw is running in another container on the same `intranet` network

3) Minor config correctness: add `api` (because you’re using `/v1`)

Fix: add `api` to `models.providers.ollama`

Option B: keep `/v1` (OpenAI-compatible)

If your `baseUrl` ends with `/v1` (OpenAI-compatible endpoint)

Also: stop using `172.17.0.1`

Option A: keep `/v1` (OpenAI-compatible)