ollama local setup, claw keeps responding in json format, streaming set to false
Issue: OpenClaw displaying raw tool-calling JSON instead of parsed text with Ollama
Setup:
OpenClaw gateway connected to local Ollama (llama3.2 model)
Docker networking configured correctly
Model responding successfully but UI shows raw JSON
Current behavior:
The chat interface displays:
json{"name":"response","parameters":{"text":"Hi there! Welcome to our conversation...","type":"text"}}
Instead of just showing the actual message text.
Current config (openclaw.json):
json"models": {
"providers": {
"ollama": {
"baseUrl": "http://openclaw-ollama:11434/v1",
"apiKey": "ollama-local",
"api": "openai-completions",
"models": [{
"id": "llama3.2",
"contextWindow": 16384,
"maxTokens": 2048
}]
}
},
"ollama/llama3.2": {
"streaming": false
}
}
Question: Is there a configuration issue causing OpenClaw to not parse the tool-calling response properly? Should I be using a different api value (openai-chat vs openai-completions)? Or is there a model-specific setting needed for proper response parsing?
The model is generating valid responses with actual content, but something in the chain isn't extracting the text from the tool-calling structure.
Setup:
OpenClaw gateway connected to local Ollama (llama3.2 model)
Docker networking configured correctly
Model responding successfully but UI shows raw JSON
Current behavior:
The chat interface displays:
json{"name":"response","parameters":{"text":"Hi there! Welcome to our conversation...","type":"text"}}
Instead of just showing the actual message text.
Current config (openclaw.json):
json"models": {
"providers": {
"ollama": {
"baseUrl": "http://openclaw-ollama:11434/v1",
"apiKey": "ollama-local",
"api": "openai-completions",
"models": [{
"id": "llama3.2",
"contextWindow": 16384,
"maxTokens": 2048
}]
}
},
"ollama/llama3.2": {
"streaming": false
}
}
Question: Is there a configuration issue causing OpenClaw to not parse the tool-calling response properly? Should I be using a different api value (openai-chat vs openai-completions)? Or is there a model-specific setting needed for proper response parsing?
The model is generating valid responses with actual content, but something in the chain isn't extracting the text from the tool-calling structure.
The AI that actually does things. Emails, calendar, home automation - all from your favorite chat app. New shell, ready to help. The lobster way.
134,827Members
Resources
Similar Threads
Was this page helpful?
