hi i need help regarding ollama | Friends of the Crustacean 🦞🤝 | Page 1

wet robin Feb 20, 2026, 2:03 PM

#

Invalid config at /root/.openclaw/openclaw.json:\n- models.providers.ollama.models.0.name: Invalid input: expected string, received undefined

models.providers.ollama.models.1.name: Invalid input: expected string, received undefined
models.providers.ollama.models.2.name: Invalid input: expected string, received undefined
models.providers.ollama.models.3.name: Invalid input: expected string, received undefined
models.providers.ollama.models.4.name: Invalid input: expected string, received undefined
bindings.0.match.peer.id: Invalid input: expected string, received undefined
bindings.1.match.peer.kind: Invalid input

Error: Config validation failed: models.providers.ollama.models.0.name: Invalid input: expected string, received undefined

jaunty willowBOT Feb 20, 2026, 2:03 PM

#

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

wet robin Feb 20, 2026, 2:06 PM

#

  "models": {
    "providers": {
      "ollama": {
        "baseUrl": "http://127.0.0.1:11434/v1",
        "apiKey": "ollama",
        "api": "openai-completions",
        "models": [
          { "id": "llama3.3", "contextWindow": 131072 },
          { "id": "mistral", "contextWindow": 32768 },
          { "id": "codellama", "contextWindow": 16384 },
          { "id": "llama3.1", "contextWindow": 131072 },
          { "id": "qwen2.5", "contextWindow": 131072 }
        ]
      }
    }
  },

lament tundraBOT Feb 20, 2026, 2:07 PM

#

Those errors are straight schema validation (the Gateway refuses to boot until the config matches exactly).

1) `models.providers.ollama.models.*.name` is missing

If you explicitly set models.providers.ollama, auto-discovery is disabled and each entry in models.providers.ollama.models must include both:

id (the Ollama model id, e.g. llama3.3)
name (a human-readable name)

Docs + copy/paste example: https://docs.openclaw.ai/providers/ollama (see “Explicit setup (manual models)”)

Fastest fix: either add name to every model entry, or remove the whole models.providers.ollama block and use auto-discovery (set OLLAMA_API_KEY=ollama-local and let OpenClaw discover models).

2) `bindings.*.match.peer` is malformed

A binding looks like:

#

bindings: [
  {
    agentId: "main",
    match: {
      channel: "discord",
      peer: { kind: "channel", id: "1474406157366988870" },
    },
  },
]

Key points:

match.channel is required
match.peer (if present) must be { kind, id }
id must be a string (quote Discord IDs)
kind must be one of: direct | group | channel (so if you used "dm", change it to "direct")

#

Reference: https://docs.openclaw.ai/channels/channel-routing and https://docs.openclaw.ai/gateway/configuration-reference (Multi-agent routing → binding match fields)

If you paste 2 snippets, I can tell you the exact edits

Please paste (redact tokens/keys):

your models.providers.ollama block
your bindings: [...] block

Also: what does openclaw --version print on that machine?

#

Yep — that block is exactly why you’re getting the models.providers.ollama.models.*.name errors: each model entry needs a name string when you configure Ollama explicitly.

Just change your models list to include name:

"models": {
  "providers": {
    "ollama": {
      "baseUrl": "http://127.0.0.1:11434/v1",
      "apiKey": "ollama",
      "api": "openai-completions",
      "models": [
        { "id": "llama3.3",  "name": "Llama 3.3",  "contextWindow": 131072 },
        { "id": "mistral",   "name": "Mistral",    "contextWindow": 32768 },
        { "id": "codellama", "name": "CodeLlama",  "contextWindow": 16384 },
        { "id": "llama3.1",  "name": "Llama 3.1",  "contextWindow": 131072 },

#

        { "id": "qwen2.5",   "name": "Qwen 2.5",   "contextWindow": 131072 }
      ]
    }
  }
},

After that, your remaining startup failures are the other ones from earlier: the bindings.*.match.peer errors.

Can you paste your bindings: [...] section too? (Redact tokens.)
I’ll point out exactly what field(s) are missing/invalid (it should be peer: { kind: "channel"|"group"|"direct", id: "..." } with the id quoted as a string).

wet robin Feb 20, 2026, 2:12 PM

#


  "bindings": [
    {
      "agentId": "wazuh-triage",
      "match": {
        "channel": "slack",
        "peer": { "kind": "group" }
      }
    },
    {
      "agentId": "wazuh-responder",
      "match": {
        "channel": "slack",
        "peer": { "id": "approvals" }
      }
    }
  ],

lament tundraBOT Feb 20, 2026, 2:15 PM

#

Both of those bindings entries are invalid because match.peer must be complete:

peer.kind must be one of: "direct" | "group" | "channel"
peer.id is required (string)

So:

Binding 1 is missing peer.id
Binding 2 is missing peer.kind (and id: "approvals" is probably a channel name, not the Slack conversation ID)

Fix option A (most common): one specific Slack channel → responder, everything else in the Slack workspace → triage

Use a peer match for the one channel, then a teamId fallback for the rest (put the specific rule first):

"bindings": [
  {

#

    "agentId": "wazuh-responder",
    "match": {
      "channel": "slack",
      "peer": { "kind": "channel", "id": "C1234567890" } // <- approvals channel ID
    }
  },
  {
    "agentId": "wazuh-triage",
    "match": {
      "channel": "slack",
      "teamId": "T1234567890" // <- your Slack workspace/team ID
    }
  }
],

#

How to find these IDs quickly:

Channel ID: open the channel in Slack web → the URL contains /archives/C… (that C… is the channel id). Private channels/group DMs often start with G….
Team/workspace ID: Slack web URL usually contains /client/T…/… (that T… is the team id).

Fix option B: match two specific peers

If you truly want two specific conversations routed:

"bindings": [
  {
    "agentId": "wazuh-triage",
    "match": { "channel": "slack", "peer": { "kind": "channel", "id": "C_TRIAGE_ID" } }
  },
  {
    "agentId": "wazuh-responder",
    "match": { "channel": "slack", "peer": { "kind": "channel", "id": "C_APPROVALS_ID" } }

#

  }
],

If you tell me what the “group” in your first binding actually is (Slack channel? private channel? group DM?), I can tell you whether kind should be "channel" or "group"—but either way you’ll need the real Slack conversation id (C…/G…).

wet robin Feb 20, 2026, 2:22 PM

#

tell me step by step to do this

#

how to fix

lament tundraBOT Feb 20, 2026, 2:28 PM

#

To fix it, you need to correct two schema rules in /root/.openclaw/openclaw.json:

1) Fix Ollama model entries (add `name`)

Right now you have:

models: [
  { id: "llama3.3", contextWindow: 131072 },
  ...
]

Each model must have a name when you define models.providers.ollama.models explicitly:

models: [

#

  { id: "llama3.3",  name: "Llama 3.3",  contextWindow: 131072 },
  { id: "mistral",   name: "Mistral",    contextWindow: 32768 },
  { id: "codellama", name: "CodeLlama",  contextWindow: 16384 },
  { id: "llama3.1",  name: "Llama 3.1",  contextWindow: 131072 },
  { id: "qwen2.5",   name: "Qwen 2.5",   contextWindow: 131072 },
]

2) Fix `bindings` (peer must include both `kind` and `id`)

Your current bindings are invalid because:

first one has peer.kind but no peer.id
second one has peer.id but no peer.kind

Also: for Slack, peer.id is not "approvals" — it must be the real Slack conversation ID like C0123... (or G0123...).

Step-by-step to get the Slack channel ID

#

Open Slack in a browser
Click the #approvals channel
Look at the URL. It will look like:
- .../client/T12345678/C23456789
Copy:
- T12345678 = teamId (optional, but nice for a fallback rule)
- C23456789 = channel id (this is what you need)

Then set bindings like this (recommended)

Put the specific channel rule first:

bindings: [
  {
    agentId: "wazuh-responder",
    match: {

#

      channel: "slack",
      peer: { kind: "channel", id: "C23456789" }, // #approvals channel id
    },
  },
  {
    agentId: "wazuh-triage",
    match: {
      channel: "slack",
      teamId: "T12345678", // slack workspace id
    },
  },
],

3) Restart + verify

After editing:

#

openclaw gateway restart
openclaw doctor

If you paste the Slack URL you see for #approvals (just the T... and C.../G... parts), I’ll tell you exactly what to put for peer.kind and peer.id.

#

Here’s a safe step-by-step to fix your config and get the Gateway booting again.

Step 0 — Make a backup (so you can undo)

On the machine where OpenClaw runs:

cp /root/.openclaw/openclaw.json /root/.openclaw/openclaw.json.bak

Step 1 — Fix the Ollama models (`name` is required)

Edit the config:

nano /root/.openclaw/openclaw.json

Find your block:

#

models: { providers: { ollama: { ... models: [ ... ] } } }

For each entry in models.providers.ollama.models, add a "name": "...".

Example:

models: {
  providers: {
    ollama: {
      baseUrl: "http://127.0.0.1:11434/v1",
      apiKey: "ollama",
      api: "openai-completions",
      models: [

#

        { id: "llama3.3",  name: "Llama 3.3",  contextWindow: 131072 },
        { id: "mistral",   name: "Mistral",    contextWindow: 32768 },
        { id: "codellama", name: "CodeLlama",  contextWindow: 16384 },
        { id: "llama3.1",  name: "Llama 3.1",  contextWindow: 131072 },
        { id: "qwen2.5",   name: "Qwen 2.5",   contextWindow: 131072 },
      ],
    },
  },
},

Step 2 — Decide what you want bindings to do

From what you pasted, you likely want:

route the #approvals Slack channel → wazuh-responder
route everything else on Slack → wazuh-triage

#

That’s the simplest/common pattern.

Step 3 — Get the Slack IDs (channel id + optionally team id)

Open Slack in the browser and go to the #approvals channel.

The URL will look like:

https://app.slack.com/client/T12345678/C23456789

Where:

T12345678 = teamId (workspace)
C23456789 = channel id

(Private channels / some group chats may show G... instead of C... — that’s fine.)

Step 4 — Fix `bindings` (peer must have both kind + id)

Back in openclaw.json, replace your bindings with:

#

Option A (recommended): approvals channel + team-wide fallback

bindings: [
  {
    agentId: "wazuh-responder",
    match: {
      channel: "slack",
      peer: { kind: "channel", id: "C23456789" }, // <-- approvals channel id from URL
    },
  },
  {
    agentId: "wazuh-triage",
    match: {
      channel: "slack",
      teamId: "T12345678", // <-- from URL
    },

#

  },
],

Option B (even simpler): approvals channel + “all Slack” fallback
If you only have one Slack workspace configured, you can skip teamId:

bindings: [
  {
    agentId: "wazuh-responder",
    match: { channel: "slack", peer: { kind: "channel", id: "C23456789" } },
  },
  {
    agentId: "wazuh-triage",
    match: { channel: "slack" },

#

  },
],

Important: put the more specific rule first (approvals first). “First match wins”.

Step 5 — Save + restart

Save in nano (Ctrl+O, Enter, Ctrl+X), then:

openclaw gateway restart

If the gateway still refuses to start, run:

#

openclaw doctor

and paste the output (it will point to the next exact schema error).

If you paste the Slack URL you see for #approvals (you can redact everything except the T... and C.../G... parts), I can tell you exactly what peer.kind/peer.id + teamId should be.

wet robin Feb 20, 2026, 3:04 PM

#

do i need to create new workspace in slack

lament tundraBOT Feb 20, 2026, 3:05 PM

#

No — you do not need to create a new Slack workspace.

What you do need depends on what you mean by “workspace”:

1) Slack “workspace” (the Slack team)

That’s your existing Slack org (the one with a T... id in the URL). You keep using that. You only need:

your Slack app installed to that workspace
the Slack tokens configured in channels.slack (botToken, appToken, etc.)

2) OpenClaw “workspace” (a folder on disk)

OpenClaw uses the word workspace to mean a local directory for an agent’s files + memory.

If you’re routing to agents like wazuh-triage and wazuh-responder, make sure you actually defined those agents in your config, e.g.:

agents: {

#

  list: [
    { id: "wazuh-triage", workspace: "/root/.openclaw/workspace-wazuh-triage" },
    { id: "wazuh-responder", workspace: "/root/.openclaw/workspace-wazuh-responder" },
  ],
},

You don’t have to “create them in Slack” — they’re just folders on your server (OpenClaw can create the folders automatically when it starts, as long as the config is valid and it has permissions).

If you paste your agents: { list: [...] } section (redact secrets), I can sanity-check that the agent IDs used in bindings actually exist.

wet robin Feb 20, 2026, 6:00 PM

#

  // ---------------------------------------------------------------------------
  // CHANNEL CONFIGURATION
  // ---------------------------------------------------------------------------
//  "channels": {
//    "slack": {
//      "enabled": false,
//      "botToken": "${SLACK_BOT_TOKEN}",
//      "appToken": "${SLACK_APP_TOKEN}",
//     "dmPolicy": "allowlist",
//      "allowFrom": [],
//      "groupPolicy": "allowlist"
//    }
//  },

  // ---------------------------------------------------------------------------
  // MULTI-AGENT ROUTING
  // ---------------------------------------------------------------------------
  "bindings": [
    {
      "agentId": "wazuh-triage",
      "match": {
        "channel": "slack",
        "peer": { "kind": "group" }
      }
    },
    {
      "agentId": "wazuh-responder",
      "match": {
        "channel": "slack",
        "peer": { "id": "approvals" }
      }
    }
  ],

what should i do

#

Gateway health:
connect failed: connect ECONNREFUSED 127.0.0.1:18789

#hi i need help regarding ollama

1) models.providers.ollama.models.*.name is missing

2) bindings.*.match.peer is malformed