Setup help needed (Telegram connection, Llama3 connection)

[Stef4fun]

My questions?

1. Why does it not find llama3 inside running ironclaw?
2. There was no question during the onboarding to connect a telegram channel/bot (like written in the telegram.md) how can I do this?
3. Which key has to be in SECRETS_MASTER_KEY just some random key I generate myself?

See following routing I did and the respective logs.
Thanks for your advice and help
Stefan

• Running Ubuntu 24 on local machine.
• Installed ironclaw remotely attached via SSH
• managed to get postrgresql running and the vector extension

Then I ran ironclaw onboard with the following options:

╭─────────────────────────╮
│ IronClaw Setup Wizard │
╰─────────────────────────╯

Step 1/8: Database Connection
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

ℹ Existing database URL: postgres://ironclaw:****@localhost:5432/ironclaw
Use this database? [Y/n]
✓ Database connection successful
Step 2/8: Security
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

ℹ Secrets master key found in SECRETS_MASTER_KEY environment variable.
✓ Security configured (env var)
Step 3/8: Inference Provider
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

ℹ Current provider: Ollama (local)

Keep current provider? [Y/n]
Ollama base URL (default: http://localhost:11434):
✓ Ollama configured (http://localhost:11434)
Step 4/8: Model Selection
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Available models:

Select a model:

[1] llama3:8b
[2] Custom model ID

1
✓ Selected llama3:8b
Step 5/8: Embeddings (Semantic Search)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

ℹ Embeddings enable semantic search in your workspace memory.

Enable semantic search? [Y/n]
ℹ No NEAR AI session or OpenAI key found for embeddings.
ℹ Set OPENAI_API_KEY in your environment to enable embeddings.
Step 6/8: Channel Configuration
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

ℹ Current tunnel configuration:
ℹ Provider: ngrok
ℹ Auth: token configured

Change tunnel configuration? [y/N]

Which channels do you want to enable?

(Use arrow keys to navigate, space to toggle, enter to confirm)

[x] CLI/TUI (always enabled)

[x] HTTP webhook

HTTP Webhook Setup:

ℹ The HTTP webhook allows external services to send messages to the agent.

Port (default: 8080):
Host (default: 0.0.0.0):
Generate a webhook secret for authentication? [Y/n]
✓ Webhook secret generated and saved to database
ℹ Retrieve it later with: ironclaw secret get http_webhook_secret
✓ HTTP webhook will listen on 0.0.0.0:8080
Step 7/8: Extensions
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

ℹ Extension registry not found. Skipping tool installation.
ℹ Install tools manually with: ironclaw tool install
Step 8/8: Background Tasks
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

ℹ Heartbeat runs periodic background tasks (e.g., checking your calendar,
ℹ monitoring for notifications, running scheduled workflows).

Enable heartbeat? [y/N] y
Check interval in minutes (default: 30):
Notify channel on findings (e.g., telegram): telegram
✓ Heartbeat enabled (every 30 minutes)

✓ Configuration saved to database

Configuration Summary:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Database: PostgreSQL (configured)
Security: environment variable
Provider: Ollama
Model: llama3:8b
Embeddings: disabled
Tunnel: ngrok (managed, starts at boot)
Channels:
- CLI/TUI: enabled
- HTTP: enabled (port 8080)
Heartbeat: every 30 minutes

To start the agent, run:
ironclaw

To change settings later:
ironclaw config set
ironclaw onboard

I have the following log when starting ironclaw:

before I started ollama with ollama serve &
Firewall is disabled
ollama serve &
[1] 4815
stefan@stefan-ZBOX-EN1070-1060-EN1070K-1060K:~$ time=2026-02-22T15:13:50.107+01:00 level=INFO source=routes.go:1663 msg="server config" env="map[CUDA_VISIBLE_DEVICES: GGML_VK_VISIBLE_DEVICES: GPU_DEVICE_ORDINAL: HIP_VISIBLE_DEVICES: HSA_OVERRIDE_GFX_VERSION: HTTPS_PROXY: HTTP_PROXY: NO_PROXY: OLLAMA_CONTEXT_LENGTH:0 OLLAMA_DEBUG:INFO OLLAMA_EDITOR: OLLAMA_FLASH_ATTENTION:false OLLAMA_GPU_OVERHEAD:0 OLLAMA_HOST:http://127.0.0.1:11434 OLLAMA_KEEP_ALIVE:5m0s OLLAMA_KV_CACHE_TYPE: OLLAMA_LLM_LIBRARY: OLLAMA_LOAD_TIMEOUT:5m0s OLLAMA_MAX_LOADED_MODELS:0 OLLAMA_MAX_QUEUE:512 OLLAMA_MODELS:/home/stefan/.ollama/models OLLAMA_MULTIUSER_CACHE:false OLLAMA_NEW_ENGINE:false OLLAMA_NOHISTORY:false OLLAMA_NOPRUNE:false OLLAMA_NO_CLOUD:false OLLAMA_NUM_PARALLEL:1 OLLAMA_ORIGINS:[http://localhost https://localhost http://localhost:* https://localhost:* http://127.0.0.1 https://127.0.0.1 http://127.0.0.1:* https://127.0.0.1:* http://0.0.0.0 https://0.0.0.0 http://0.0.0.0:* https://0.0.0.0:* app://* file://* tauri://* vscode-webview://* vscode-file://*] OLLAMA_REMOTES:[ollama.com] OLLAMA_SCHED_SPREAD:false OLLAMA_VULKAN:false ROCR_VISIBLE_DEVICES: http_proxy: https_proxy: no_proxy:]"
time=2026-02-22T15:13:50.107+01:00 level=INFO source=routes.go:1665 msg="Ollama cloud disabled: false"
time=2026-02-22T15:13:50.107+01:00 level=INFO source=images.go:473 msg="total blobs: 5"
time=2026-02-22T15:13:50.108+01:00 level=INFO source=images.go:480 msg="total unused blobs removed: 0"
time=2026-02-22T15:13:50.108+01:00 level=INFO source=routes.go:1718 msg="Listening on 127.0.0.1:11434 (version 0.16.3)"
time=2026-02-22T15:13:50.109+01:00 level=INFO source=runner.go:67 msg="discovering available GPUs..."
time=2026-02-22T15:13:50.109+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 39477"
time=2026-02-22T15:13:50.200+01:00 level=INFO source=runner.go:106 msg="experimental Vulkan support disabled. To enable, set OLLAMA_VULKAN=1"
time=2026-02-22T15:13:50.200+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 38313"
time=2026-02-22T15:13:50.317+01:00 level=INFO source=server.go:431 msg="starting runner" cmd="/usr/local/bin/ollama runner --ollama-engine --port 40797"
time=2026-02-22T15:13:50.433+01:00 level=INFO source=types.go:42 msg="inference compute" id=GPU-a2b3289-ec0d-8fa0-45dd-48a98257ca3e filter_id="" library=CUDA compute=6.1 name=CUDA0 description="NVIDIA GeForce GTX 1060" libdirs=ollama,cuda_v12 driver=12.2 pci_id=0000:01:00.0 type=discrete total="6.0 GiB" available="5.9 GiB"
time=2026-02-22T15:13:50.433+01:00 level=INFO source=routes.go:1768 msg="vram-based default context" total_vram="6.0 GiB" default_num_ctx=4096

---

stefan@stefan-ZBOX-EN1070-1060-EN1070K-1060K:~$ ironclaw
2026-02-22T13:48:50.423779Z INFO Starting IronClaw...
2026-02-22T13:48:50.423869Z INFO Loaded configuration for agent: ironclaw
2026-02-22T13:48:50.423885Z INFO LLM backend: ollama
2026-02-22T13:48:50.435750Z INFO PostgreSQL database connected and migrations applied
2026-02-22T13:48:50.437366Z INFO Configuration reloaded from database
2026-02-22T13:48:50.441246Z INFO Starting ngrok tunnel on 127.0.0.1:3000...
2026-02-22T13:48:50.441486Z ERROR Failed to start tunnel: No such file or directory (os error 2)
2026-02-22T13:48:50.503208Z INFO Using Ollama (base_url: http://localhost:11434, model: llama3)
2026-02-22T13:48:50.503250Z INFO LLM provider initialized: llama3
2026-02-22T13:48:50.503257Z INFO LLM retry wrapper enabled max_retries=3
2026-02-22T13:48:50.514317Z INFO Safety layer initialized
2026-02-22T13:48:50.575002Z INFO Registered 4 built-in tools
2026-02-22T13:48:50.575018Z INFO Embeddings disabled (set OPENAI_API_KEY or EMBEDDING_ENABLED=true)
2026-02-22T13:48:50.575065Z INFO Registered 4 memory tools
2026-02-22T13:48:50.637664Z INFO Registered 6 extension management tools
2026-02-22T13:48:50.637700Z INFO Extension manager initialized with in-chat discovery tools
2026-02-22T13:48:50.637934Z INFO Tool registry initialized with 14 total tools
2026-02-22T13:48:50.637966Z INFO REPL mode enabled
2026-02-22T13:48:50.638082Z INFO Orchestrator internal API listening on 0.0.0.0:50051
2026-02-22T13:48:50.641544Z INFO Lifecycle hooks initialized bundled=1 plugin=0 workspace=0 outbound_webhooks=0 errors=0
2026-02-22T13:48:50.641605Z INFO Registered 6 job management tools
2026-02-22T13:48:50.641729Z INFO Web UI: http://127.0.0.1:3000/

╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶

ironclaw v0.9.0

model llama3 via ollama
database postgres (connected)
tools 20 registered
features heartbeat (30m) sandbox routines
channels repl gateway

gateway http://127.0.0.1:3000/?token=Ntp...

╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶╶

/help for commands, /quit to exit

2026-02-22T13:48:50.642397Z INFO Started channel: gateway
2026-02-22T13:48:50.642517Z INFO Started channel: repl
2026-02-22T13:48:50.642560Z INFO Registered 5 routine management tools
2026-02-22T13:48:50.642611Z INFO Starting heartbeat loop with interval 1800s
› 2026-02-22T13:48:50.643225Z INFO Routines enabled: cron ticker every 15s, max 10 concurrent
2026-02-22T13:48:50.643241Z INFO Agent ironclaw ready and listening
test
› ○ Processing...
...
Error: LLM error: Provider llama3 request failed: HttpError: Invalid status code 404 Not Found with message:
{"error":"model 'llama3' not found"}

[Stef4fun]

ollama cp llama3:8b llama3
This lead to a new error message regaring ollama:
2026-02-22T15:13:56.917274Z WARN Retrying after transient error (tools) provider=llama3 attempt=3 max_retries=3 delay_ms=3828 error=Provider llama3 request failed: HttpError: Invalid status code 400 Bad Request with message: {"error":"registry.ollama.ai/library/llama3:latest does not support tools"}

replacing the model with one which supports tools helps:
ollama pull qwen2.5:7b
But I need to also map the qwen model to llama3:
ollama cp qwen2.5:7b llama3

[Stef4fun]

~/repos/ironclaw/tools-src$ ironclaw tool install ./telegram
Installing the telegram tool from within the repo now shows during ironclaw startup:
INFO Loaded WASM tools from directory count=1 tools=["telegram-tool"]

[Stef4fun]

Running the build.sh of the channels leads to the following error:
:~/repos/ironclaw$ ./channels-src/telegram/build.sh
Building Telegram channel WASM component...
error: duplicate key
--> Cargo.toml:29:2
|
29 | [workspace]
| ^^^^^^^^^

Removing one of the [workspace] from Cargo.toml makes the install successful!

ironclaw startup:
2026-02-22T16:48:29.232884Z INFO Loaded WASM tools from directory count=2 tools=["telegram-tool", "telegram-channel"]
2026-02-22T16:48:29.232932Z INFO Loaded 2 WASM tools from /home/stefan/.ironclaw/tools

[serrrfirat]

#323 I am addressing it here

[Stef4fun]

#323 I am addressing it here

Great @serrrfirat. After digging a bit further I see two main issues:

1. When using a local model and defining it during the onboarding process, the internal state somehow still points to the llama3 model, doing a link inside ollama to the right model is a workaround here.
   ollama cp qwen2.5:7b llama3
2. The onboarding process does not properly ask for the telegram channel configuration and the documentation here is also not clear how to achieve it outside of the onboarding. I still could not solve this yet.

[serrrfirat]

1. I need to look a bit more into this.
2. It actually does but right now without the PR that fixes it, it just skips since it cannot create the channel. When it finds the telegram wasm binary it will ask you about couple of things including bot token, pairing etc.

I'll try to get the fix in today.

[Stef4fun]

Great from source I see now the telegram option!

[Stef4fun]

Sorry to bother you with that, I now managed to build and set the keys, but when starting ironclaw it is not able to start the telegram channel:
ERROR Failed to start channel telegram: Channel telegram failed to start: WASM instantiation error: failed to convert function to given type

I cleaned and rebuild but it didn't fix it using the build-all.sh
The files in /.ironclaw/channels are new from the build.

Are you able to spot the reason?